diff --git a/doc/sphinx/Pacemaker_Explained/advanced-resources.rst b/doc/sphinx/Pacemaker_Explained/advanced-resources.rst index 5d32685a3c..c63a1b0d45 100644 --- a/doc/sphinx/Pacemaker_Explained/advanced-resources.rst +++ b/doc/sphinx/Pacemaker_Explained/advanced-resources.rst @@ -1,1643 +1,1592 @@ Advanced Resource Types ----------------------- .. index: single: group resource single: resource; group .. _group-resources: Groups - A Syntactic Shortcut ############################# One of the most common elements of a cluster is a set of resources that need to be located together, start sequentially, and stop in the reverse order. To simplify this configuration, we support the concept of groups. .. topic:: A group of two primitive resources .. code-block:: xml Although the example above contains only two resources, there is no limit to the number of resources a group can contain. The example is also sufficient to explain the fundamental properties of a group: * Resources are started in the order they appear in (**Public-IP** first, then **Email**) * Resources are stopped in the reverse order to which they appear in (**Email** first, then **Public-IP**) If a resource in the group can't run anywhere, then nothing after that is allowed to run, too. * If **Public-IP** can't run anywhere, neither can **Email**; * but if **Email** can't run anywhere, this does not affect **Public-IP** in any way The group above is logically equivalent to writing: .. topic:: How the cluster sees a group resource .. code-block:: xml Obviously as the group grows bigger, the reduced configuration effort can become significant. Another (typical) example of a group is a DRBD volume, the filesystem mount, an IP address, and an application that uses them. .. index:: pair: XML element; group Group Properties ________________ .. table:: **Properties of a Group Resource** +-------+--------------------------------------+ | Field | Description | +=======+======================================+ | id | .. index:: | | | single: group; property, id | | | single: property; id (group) | | | single: id; group property | | | | | | A unique name for the group | +-------+--------------------------------------+ Group Options _____________ Groups inherit the ``priority``, ``target-role``, and ``is-managed`` properties from primitive resources. See :ref:`resource_options` for information about those properties. Group Instance Attributes _________________________ Groups have no instance attributes. However, any that are set for the group object will be inherited by the group's children. Group Contents ______________ Groups may only contain a collection of cluster resources (see :ref:`primitive-resource`). To refer to a child of a group resource, just use the child's ``id`` instead of the group's. Group Constraints _________________ Although it is possible to reference a group's children in constraints, it is usually preferable to reference the group itself. .. topic:: Some constraints involving groups .. code-block:: xml .. index:: pair: resource-stickiness; group Group Stickiness ________________ Stickiness, the measure of how much a resource wants to stay where it is, is additive in groups. Every active resource of the group will contribute its stickiness value to the group's total. So if the default ``resource-stickiness`` is 100, and a group has seven members, five of which are active, then the group as a whole will prefer its current location with a score of 500. .. index:: - single: clone resource + single: clone single: resource; clone .. _s-resource-clone: Clones - Resources That Can Have Multiple Active Instances ########################################################## *Clone* resources are resources that can have more than one copy active at the same time. This allows you, for example, to run a copy of a daemon on every node. You can clone any primitive or group resource [#]_. Anonymous versus Unique Clones ______________________________ A clone resource is configured to be either *anonymous* or *globally unique*. Anonymous clones are the simplest. These behave completely identically everywhere they are running. Because of this, there can be only one instance of an anonymous clone active per node. The instances of globally unique clones are distinct entities. All instances are launched identically, but one instance of the clone is not identical to any other instance, whether running on the same node or a different node. As an example, a cloned IP address can use special kernel functionality such that each instance handles a subset of requests for the same IP address. .. index:: - single: Promotable Clone Resources + single: promotable clone single: resource; promotable .. _s-resource-promotable: Promotable clones _________________ If a clone is *promotable*, its instances can perform a special role that Pacemaker will manage via the ``promote`` and ``demote`` actions of the resource agent. Services that support such a special role have various terms for the special role and the default role: primary and secondary, master and replica, controller and worker, etc. Pacemaker uses the terms *master* and *slave* [#]_, but is agnostic to what the service calls them or what they do. All that Pacemaker cares about is that an instance comes up in the default role when started, and the resource agent supports the ``promote`` and ``demote`` actions to manage entering and exiting the special role. .. index:: pair: XML element; clone Clone Properties ________________ .. table:: **Properties of a Clone Resource** +-------+--------------------------------------+ | Field | Description | +=======+======================================+ | id | .. index:: | | | single: clone; property, id | | | single: property; id (clone) | | | single: id; clone property | | | | | | A unique name for the clone | +-------+--------------------------------------+ .. index:: pair: options; clone Clone Options _____________ :ref:`Options ` inherited from primitive resources: ``priority, target-role, is-managed`` .. table:: **Clone-specific configuration options** +-------------------+-----------------+-------------------------------------------------------+ | Field | Default | Description | +===================+=================+=======================================================+ | globally-unique | false | .. index:: | | | | single: clone; option, globally-unique | | | | single: option; globally-unique (clone) | | | | single: globally-unique; clone option | | | | | | | | If **true**, each clone instance performs a | | | | distinct function | +-------------------+-----------------+-------------------------------------------------------+ | clone-max | number of nodes | .. index:: | | | in the cluster | single: clone; option, clone-max | | | | single: option; clone-max (clone) | | | | single: clone-max; clone option | | | | | | | | The maximum number of clone instances that can | | | | be started across the entire cluster | +-------------------+-----------------+-------------------------------------------------------+ | clone-node-max | 1 | .. index:: | | | | single: clone; option, clone-node-max | | | | single: option; clone-node-max (clone) | | | | single: clone-node-max; clone option | | | | | | | | If ``globally-unique`` is **true**, the maximum | | | | number of clone instances that can be started | | | | on a single node | +-------------------+-----------------+-------------------------------------------------------+ | clone-min | 0 | .. index:: | | | | single: clone; option, clone-min | | | | single: option; clone-min (clone) | | | | single: clone-min; clone option | | | | | | | | Require at least this number of clone instances | | | | to be runnable before allowing resources | | | | depending on the clone to be runnable. A value | | | | of 0 means require all clone instances to be | | | | runnable. | +-------------------+-----------------+-------------------------------------------------------+ | notify | false | .. index:: | | | | single: clone; option, notify | | | | single: option; notify (clone) | | | | single: notify; clone option | | | | | | | | Call the resource agent's **notify** action for | | | | all active instances, before and after starting | | | | or stopping any clone instance. The resource | | | | agent must support this action. | | | | Allowed values: **false**, **true** | +-------------------+-----------------+-------------------------------------------------------+ | ordered | false | .. index:: | | | | single: clone; option, ordered | | | | single: option; ordered (clone) | | | | single: ordered; clone option | | | | | | | | If **true**, clone instances must be started | | | | sequentially instead of in parallel. | | | | Allowed values: **false**, **true** | +-------------------+-----------------+-------------------------------------------------------+ | interleave | false | .. index:: | | | | single: clone; option, interleave | | | | single: option; interleave (clone) | | | | single: interleave; clone option | | | | | | | | When this clone is ordered relative to another | | | | clone, if this option is **false** (the default), | | | | the ordering is relative to *all* instances of | | | | the other clone, whereas if this option is | | | | **true**, the ordering is relative only to | | | | instances on the same node. | | | | Allowed values: **false**, **true** | +-------------------+-----------------+-------------------------------------------------------+ | promotable | false | .. index:: | | | | single: clone; option, promotable | | | | single: option; promotable (clone) | | | | single: promotable; clone option | | | | | | | | If **true**, clone instances can perform a | | | | special role that Pacemaker will manage via the | | | | resource agent's **promote** and **demote** | | | | actions. The resource agent must support these | | | | actions. | | | | Allowed values: **false**, **true** | +-------------------+-----------------+-------------------------------------------------------+ | promoted-max | 1 | .. index:: | | | | single: clone; option, promoted-max | | | | single: option; promoted-max (clone) | | | | single: promoted-max; clone option | | | | | | | | If ``promotable`` is **true**, the number of | | | | instances that can be promoted at one time | | | | across the entire cluster | +-------------------+-----------------+-------------------------------------------------------+ | promoted-node-max | 1 | .. index:: | | | | single: clone; option, promoted-node-max | | | | single: option; promoted-node-max (clone) | | | | single: promoted-node-max; clone option | | | | | | | | If ``promotable`` is **true** and ``globally-unique`` | | | | is **false**, the number of clone instances can be | | | | promoted at one time on a single node | +-------------------+-----------------+-------------------------------------------------------+ For backward compatibility, ``master-max`` and ``master-node-max`` are accepted as aliases for ``promoted-max`` and ``promoted-node-max``, but are deprecated since 2.0.0, and support for them will be removed in a future version. Clone Contents ______________ Clones must contain exactly one primitive or group resource. .. topic:: A clone that runs a web server on all nodes .. code-block:: xml .. warning:: You should never reference the name of a clone's child (the primitive or group resource being cloned). If you think you need to do this, you probably need to re-evaluate your design. Clone Instance Attribute ________________________ Clones have no instance attributes; however, any that are set here will be inherited by the clone's child. +.. index:: + single: clone; constraint + Clone Constraints _________________ In most cases, a clone will have a single instance on each active cluster node. If this is not the case, you can indicate which nodes the cluster should preferentially assign copies to with resource location constraints. These constraints are written no differently from those for primitive resources except that the clone's **id** is used. .. topic:: Some constraints involving clones .. code-block:: xml Ordering constraints behave slightly differently for clones. In the example above, ``apache-stats`` will wait until all copies of ``apache-clone`` that need to be started have done so before being started itself. Only if *no* copies can be started will ``apache-stats`` be prevented from being active. Additionally, the clone will wait for ``apache-stats`` to be stopped before stopping itself. Colocation of a primitive or group resource with a clone means that the resource can run on any node with an active instance of the clone. The cluster will choose an instance based on where the clone is running and the resource's own location preferences. Colocation between clones is also possible. If one clone **A** is colocated with another clone **B**, the set of allowed locations for **A** is limited to nodes on which **B** is (or will be) active. Placement is then performed normally. +.. index:: + single: promotable clone; constraint + +.. _promotable-clone-constraints: + Promotable Clone Constraints ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ For promotable clone resources, the ``first-action`` and/or ``then-action`` fields for ordering constraints may be set to ``promote`` or ``demote`` to constrain the master role, and colocation constraints may contain ``rsc-role`` and/or ``with-rsc-role`` fields. -.. index:: - single: constraint; colocation - -.. table:: **Additional colocation constraint options for promotable clone resources** - - +---------------+---------+-------------------------------------------------------+ - | Field | Default | Description | - +===============+=========+=======================================================+ - | rsc-role | Started | .. index:: | - | | | single: clone; ordering constraint, rsc-role | - | | | single: ordering constraint; rsc-role (clone) | - | | | single: rsc-role; clone ordering constraint | - | | | | - | | | An additional attribute of colocation constraints | - | | | that specifies the role that ``rsc`` must be in. | - | | | Allowed values: **Started**, **Master**, **Slave**. | - +---------------+---------+-------------------------------------------------------+ - | with-rsc-role | Started | .. index:: | - | | | single: clone; ordering constraint, with-rsc-role | - | | | single: ordering constraint; with-rsc-role (clone) | - | | | single: with-rsc-role; clone ordering constraint | - | | | | - | | | An additional attribute of colocation constraints | - | | | that specifies the role that ``with-rsc`` must be in. | - | | | Allowed values: **Started**, **Master**, **Slave**. | - +---------------+---------+-------------------------------------------------------+ - .. topic:: Constraints involving promotable clone resources .. code-block:: xml In the example above, **myApp** will wait until one of the database copies has been started and promoted to master before being started itself on the same node. Only if no copies can be promoted will **myApp** be prevented from being active. Additionally, the cluster will wait for **myApp** to be stopped before demoting the database. Colocation of a primitive or group resource with a promotable clone resource means that it can run on any node with an active instance of the promotable clone resource that has the specified role (**master** or **slave**). In the example above, the cluster will choose a location based on where database is running as a **master**, and if there are multiple **master** instances it will also factor in **myApp**'s own location preferences when deciding which location to choose. Colocation with regular clones and other promotable clone resources is also possible. In such cases, the set of allowed locations for the **rsc** clone is (after role filtering) limited to nodes on which the ``with-rsc`` promotable clone resource is (or will be) in the specified role. Placement is then performed as normal. Using Promotable Clone Resources in Colocation Sets ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ -.. index:: - single: constraint; colocation - single: constraint; resource set - -.. table:: **Additional colocation set options relevant to promotable clone resources** - - +-------+---------+-----------------------------------------------------+ - | Field | Default | Description | - +=======+=========+=====================================================+ - | role | Started | .. index:: | - | | | single: clone; ordering constraint; role | - | | | single: ordering constraint; role (clone) | - | | | single: role; clone ordering constraint | - | | | | - | | | The role that *all members* of the set must be in. | - | | | Allowed values: **Started**, **Master**, **Slave**. | - +-------+---------+-----------------------------------------------------+ +When a promotable clone is used in a :ref:`resource set ` +inside a colocation constraint, the resource set may take a ``role`` attribute. In the following example **B**'s master must be located on the same node as **A**'s master. Additionally resources **C** and **D** must be located on the same node as **A**'s and **B**'s masters. .. topic:: Colocate C and D with A's and B's master instances .. code-block:: xml Using Promotable Clone Resources in Ordered Sets ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ -.. index:: - single: constraint; colocation - single: constraint; resource set - -.. table:: **Additional ordered set options relevant to promotable clone resources** - - +--------+------------------+-----------------------------------------------------+ - | Field | Default | Description | - +=======++==================+=====================================================+ - | action | value of | .. index:: | - | | ``first-action`` | single: clone; ordering constraint; action | - | | | single: ordering constraintl action (clone) | - | | | single: action; clone ordering constraint | - | | | | - | | | An additional attribute of ordering constraint | - | | | sets that specifies the action that applies to | - | | | *all members* of the set. | - | | | Allowed values: **start**, **stop**, **promote**, | - | | | **demote**. | - +--------+------------------+-----------------------------------------------------+ - +When a promotable clone is used in a :ref:`resource set ` +inside an ordering constraint, the resource set may take an ``action`` +attribute. + .. topic:: Start C and D after first promoting A and B .. code-block:: xml In the above example, **B** cannot be promoted to a master role until **A** has been promoted. Additionally, resources **C** and **D** must wait until **A** and **B** have been promoted before they can start. .. index:: pair: resource-stickiness; clone .. _s-clone-stickiness: Clone Stickiness ________________ To achieve a stable allocation pattern, clones are slightly sticky by default. If no value for ``resource-stickiness`` is provided, the clone will use a value of 1. Being a small value, it causes minimal disturbance to the score calculations of other resources but is enough to prevent Pacemaker from needlessly moving copies around the cluster. .. note:: For globally unique clones, this may result in multiple instances of the clone staying on a single node, even after another eligible node becomes active (for example, after being put into standby mode then made active again). If you do not want this behavior, specify a ``resource-stickiness`` of 0 for the clone temporarily and let the cluster adjust, then set it back to 1 if you want the default behavior to apply again. .. important:: If ``resource-stickiness`` is set in the ``rsc_defaults`` section, it will apply to clone instances as well. This means an explicit ``resource-stickiness`` of 0 in ``rsc_defaults`` works differently from the implicit default used when ``resource-stickiness`` is not specified. Clone Resource Agent Requirements _________________________________ Any resource can be used as an anonymous clone, as it requires no additional support from the resource agent. Whether it makes sense to do so depends on your resource and its resource agent. Resource Agent Requirements for Globally Unique Clones ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Globally unique clones require additional support in the resource agent. In particular, it must only respond with ``${OCF_SUCCESS}`` if the node has that exact instance active. All other probes for instances of the clone should result in ``${OCF_NOT_RUNNING}`` (or one of the other OCF error codes if they are failed). Individual instances of a clone are identified by appending a colon and a numerical offset, e.g. **apache:2**. Resource agents can find out how many copies there are by examining the ``OCF_RESKEY_CRM_meta_clone_max`` environment variable and which instance it is by examining ``OCF_RESKEY_CRM_meta_clone``. The resource agent must not make any assumptions (based on ``OCF_RESKEY_CRM_meta_clone``) about which numerical instances are active. In particular, the list of active copies will not always be an unbroken sequence, nor always start at 0. Resource Agent Requirements for Promotable Clones ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Promotable clone resources require two extra actions, ``demote`` and ``promote``, which are responsible for changing the state of the resource. Like **start** and **stop**, they should return ``${OCF_SUCCESS}`` if they completed successfully or a relevant error code if they did not. The states can mean whatever you wish, but when the resource is started, it must come up in the mode called **slave**. From there the cluster will decide which instances to promote to **master**. In addition to the clone requirements for monitor actions, agents must also *accurately* report which state they are in. The cluster relies on the agent to report its status (including role) accurately and does not indicate to the agent what role it currently believes it to be in. .. table:: **Role implications of OCF return codes** +---------------------+------------------------------------------------+ | Monitor Return Code | Description | +=====================+================================================+ | OCF_NOT_RUNNING | .. index:: | | | single: OCF_NOT_RUNNING | | | single: OCF return code; OCF_NOT_RUNNING | | | | | | Stopped | +---------------------+------------------------------------------------+ | OCF_SUCCESS | .. index:: | | | single: OCF_SUCCESS | | | single: OCF return code; OCF_SUCCESS | | | | | | Running (Slave) | +---------------------+------------------------------------------------+ | OCF_RUNNING_MASTER | .. index:: | | | single: OCF_RUNNING_MASTER | | | single: OCF return code; OCF_RUNNING_MASTER | | | | | | Running (Master) | +---------------------+------------------------------------------------+ | OCF_FAILED_MASTER | .. index:: | | | single: OCF_FAILED_MASTER | | | single: OCF return code; OCF_FAILED_MASTER | | | | | | Failed (Master) | +---------------------+------------------------------------------------+ | Other | .. index:: | | | single: return code | | | | | | Failed (Slave) | +---------------------+------------------------------------------------+ Clone Notifications ~~~~~~~~~~~~~~~~~~~ If the clone has the ``notify`` meta-attribute set to **true**, and the resource agent supports the ``notify`` action, Pacemaker will call the action when appropriate, passing a number of extra variables which, when combined with additional context, can be used to calculate the current state of the cluster and what is about to happen to it. .. index:: single: clone; environment variables single: notify; environment variables .. table:: **Environment variables supplied with Clone notify actions** +----------------------------------------------+-------------------------------------------------------------------------------+ | Variable | Description | +==============================================+===============================================================================+ | OCF_RESKEY_CRM_meta_notify_type | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_type | | | single: OCF_RESKEY_CRM_meta_notify_type | | | | | | Allowed values: **pre**, **post** | +----------------------------------------------+-------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_operation | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_operation | | | single: OCF_RESKEY_CRM_meta_notify_operation | | | | | | Allowed values: **start**, **stop** | +----------------------------------------------+-------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_start_resource | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_start_resource | | | single: OCF_RESKEY_CRM_meta_notify_start_resource | | | | | | Resources to be started | +----------------------------------------------+-------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_stop_resource | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_stop_resource | | | single: OCF_RESKEY_CRM_meta_notify_stop_resource | | | | | | Resources to be stopped | +----------------------------------------------+-------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_active_resource | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_active_resource | | | single: OCF_RESKEY_CRM_meta_notify_active_resource | | | | | | Resources that are running | +----------------------------------------------+-------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_inactive_resource | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_inactive_resource | | | single: OCF_RESKEY_CRM_meta_notify_inactive_resource | | | | | | Resources that are not running | +----------------------------------------------+-------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_start_uname | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_start_uname | | | single: OCF_RESKEY_CRM_meta_notify_start_uname | | | | | | Nodes on which resources will be started | +----------------------------------------------+-------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_stop_uname | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_stop_uname | | | single: OCF_RESKEY_CRM_meta_notify_stop_uname | | | | | | Nodes on which resources will be stopped | +----------------------------------------------+-------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_active_uname | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_active_uname | | | single: OCF_RESKEY_CRM_meta_notify_active_uname | | | | | | Nodes on which resources are running | +----------------------------------------------+-------------------------------------------------------------------------------+ The variables come in pairs, such as ``OCF_RESKEY_CRM_meta_notify_start_resource`` and ``OCF_RESKEY_CRM_meta_notify_start_uname``, and should be treated as an array of whitespace-separated elements. ``OCF_RESKEY_CRM_meta_notify_inactive_resource`` is an exception, as the matching **uname** variable does not exist since inactive resources are not running on any node. Thus, in order to indicate that **clone:0** will be started on **sles-1**, **clone:2** will be started on **sles-3**, and **clone:3** will be started on **sles-2**, the cluster would set: .. topic:: Notification variables .. code-block:: none OCF_RESKEY_CRM_meta_notify_start_resource="clone:0 clone:2 clone:3" OCF_RESKEY_CRM_meta_notify_start_uname="sles-1 sles-3 sles-2" .. note:: Pacemaker will log but otherwise ignore failures of notify actions. Interpretation of Notification Variables ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ **Pre-notification (stop):** * Active resources: ``$OCF_RESKEY_CRM_meta_notify_active_resource`` * Inactive resources: ``$OCF_RESKEY_CRM_meta_notify_inactive_resource`` * Resources to be started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources to be stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` **Post-notification (stop) / Pre-notification (start):** * Active resources * ``$OCF_RESKEY_CRM_meta_notify_active_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * Inactive resources * ``$OCF_RESKEY_CRM_meta_notify_inactive_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * Resources that were started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources that were stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` **Post-notification (start):** * Active resources: * ``$OCF_RESKEY_CRM_meta_notify_active_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Inactive resources: * ``$OCF_RESKEY_CRM_meta_notify_inactive_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources that were started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources that were stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` Extra Notifications for Promotable Clones ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ .. index:: single: clone; environment variables single: promotable; environment variables .. table:: **Extra environment variables supplied for promotable clones** +---------------------------------------------+------------------------------------------------------------------------------+ | Variable | Description | +=============================================+==============================================================================+ | OCF_RESKEY_CRM_meta_notify_master_resource | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_master_resource | | | single: OCF_RESKEY_CRM_meta_notify_master_resource | | | | | | Resources that are running in **Master** mode | +---------------------------------------------+------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_slave_resource | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_slave_resource | | | single: OCF_RESKEY_CRM_meta_notify_slave_resource | | | | | | Resources that are running in **Slave** mode | +---------------------------------------------+------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_promote_resource | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_promote_resource | | | single: OCF_RESKEY_CRM_meta_notify_promote_resource | | | | | | Resources to be promoted | +---------------------------------------------+------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_demote_resource | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_demote_resource | | | single: OCF_RESKEY_CRM_meta_notify_demote_resource | | | | | | Resources to be demoted | +---------------------------------------------+------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_promote_uname | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_promote_uname | | | single: OCF_RESKEY_CRM_meta_notify_promote_uname | | | | | | Nodes on which resources will be promoted | +---------------------------------------------+------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_demote_uname | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_demote_uname | | | single: OCF_RESKEY_CRM_meta_notify_demote_uname | | | | | | Nodes on which resources will be demoted | +---------------------------------------------+------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_master_uname | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_master_uname | | | single: OCF_RESKEY_CRM_meta_notify_master_uname | | | | | | Nodes on which resources are running in **Master** mode | +---------------------------------------------+------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_slave_uname | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_slave_uname | | | single: OCF_RESKEY_CRM_meta_notify_slave_uname | | | | | | Nodes on which resources are running in **Slave** mode | +---------------------------------------------+------------------------------------------------------------------------------+ Interpretation of Promotable Notification Variables ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ **Pre-notification (demote):** * **Active** resources: ``$OCF_RESKEY_CRM_meta_notify_active_resource`` * **Master** resources: ``$OCF_RESKEY_CRM_meta_notify_master_resource`` * **Slave** resources: ``$OCF_RESKEY_CRM_meta_notify_slave_resource`` * Inactive resources: ``$OCF_RESKEY_CRM_meta_notify_inactive_resource`` * Resources to be started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources to be promoted: ``$OCF_RESKEY_CRM_meta_notify_promote_resource`` * Resources to be demoted: ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Resources to be stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` **Post-notification (demote) / Pre-notification (stop):** * **Active** resources: ``$OCF_RESKEY_CRM_meta_notify_active_resource`` * **Master** resources: * ``$OCF_RESKEY_CRM_meta_notify_master_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * **Slave** resources: ``$OCF_RESKEY_CRM_meta_notify_slave_resource`` * Inactive resources: ``$OCF_RESKEY_CRM_meta_notify_inactive_resource`` * Resources to be started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources to be promoted: ``$OCF_RESKEY_CRM_meta_notify_promote_resource`` * Resources to be demoted: ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Resources to be stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * Resources that were demoted: ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` **Post-notification (stop) / Pre-notification (start)** * **Active** resources: * ``$OCF_RESKEY_CRM_meta_notify_active_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * **Master** resources: * ``$OCF_RESKEY_CRM_meta_notify_master_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * **Slave** resources: * ``$OCF_RESKEY_CRM_meta_notify_slave_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * Inactive resources: * ``$OCF_RESKEY_CRM_meta_notify_inactive_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * Resources to be started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources to be promoted: ``$OCF_RESKEY_CRM_meta_notify_promote_resource`` * Resources to be demoted: ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Resources to be stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * Resources that were demoted: ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Resources that were stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` **Post-notification (start) / Pre-notification (promote)** * **Active** resources: * ``$OCF_RESKEY_CRM_meta_notify_active_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * **Master** resources: * ``$OCF_RESKEY_CRM_meta_notify_master_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * **Slave** resources: * ``$OCF_RESKEY_CRM_meta_notify_slave_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Inactive resources: * ``$OCF_RESKEY_CRM_meta_notify_inactive_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources to be started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources to be promoted: ``$OCF_RESKEY_CRM_meta_notify_promote_resource`` * Resources to be demoted: ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Resources to be stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * Resources that were started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources that were demoted: ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Resources that were stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` **Post-notification (promote)** * **Active** resources: * ``$OCF_RESKEY_CRM_meta_notify_active_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * **Master** resources: * ``$OCF_RESKEY_CRM_meta_notify_master_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_promote_resource`` * **Slave** resources: * ``$OCF_RESKEY_CRM_meta_notify_slave_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_promote_resource`` * Inactive resources: * ``$OCF_RESKEY_CRM_meta_notify_inactive_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources to be started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources to be promoted: ``$OCF_RESKEY_CRM_meta_notify_promote_resource`` * Resources to be demoted: ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Resources to be stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * Resources that were started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources that were promoted: ``$OCF_RESKEY_CRM_meta_notify_promote_resource`` * Resources that were demoted: ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Resources that were stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` Monitoring Promotable Clone Resources _____________________________________ The usual monitor actions are insufficient to monitor a promotable clone resource, because Pacemaker needs to verify not only that the resource is active, but also that its actual role matches its intended one. Define two monitoring actions: the usual one will cover the slave role, and an additional one with ``role="master"`` will cover the master role. .. topic:: Monitoring both states of a promotable clone resource .. code-block:: xml .. important:: It is crucial that *every* monitor operation has a different interval! Pacemaker currently differentiates between operations only by resource and interval; so if (for example) a promotable clone resource had the same monitor interval for both roles, Pacemaker would ignore the role when checking the status -- which would cause unexpected return codes, and therefore unnecessary complications. .. _s-promotion-scores: Determining Which Instance is Promoted ______________________________________ Pacemaker can choose a promotable clone instance to be promoted in one of two ways: * Promotion scores: These are node attributes set via the ``crm_master`` utility, which generally would be called by the resource agent's start action if it supports promotable clones. This tool automatically detects both the resource and host, and should be used to set a preference for being promoted. Based on this, ``promoted-max``, and ``promoted-node-max``, the instance(s) with the highest preference will be promoted. * Constraints: Location constraints can indicate which nodes are most preferred as masters. .. topic:: Explicitly preferring node1 to be promoted to master .. code-block:: xml .. index: single: bundle resource single: resource; bundle pair: container; Docker pair: container; podman pair: container; rkt .. _s-resource-bundle: Bundles - Isolated Environments ############################### Pacemaker supports a special syntax for launching a `container `_ with any infrastructure it requires: the *bundle*. Pacemaker bundles support `Docker `_, `podman `_ *(since 2.0.1)*, and `rkt `_ container technologies. [#]_ .. topic:: A bundle for a containerized web server .. code-block:: xml .. index: single: bundle resource single: resource; bundle Bundle Prerequisites ____________________ Before configuring a bundle in Pacemaker, the user must install the appropriate container launch technology (Docker, podman, or rkt), and supply a fully configured container image, on every node allowed to run the bundle. Pacemaker will create an implicit resource of type **ocf:heartbeat:docker**, **ocf:heartbeat:podman**, or **ocf:heartbeat:rkt** to manage a bundle's container. The user must ensure that the appropriate resource agent is installed on every node allowed to run the bundle. .. index:: pair: XML element; bundle Bundle Properties _________________ .. table:: **XML Attributes of a bundle Element** +-------------+-----------------------------------------------+ | Attribute | Description | +=============+===============================================+ | id | .. index:: | | | single: bundle; attribute, id | | | single: attribute; id (bundle) | | | single: id; bundle attribute | | | | | | A unique name for the bundle (required) | +-------------+-----------------------------------------------+ | description | .. index:: | | | single: bundle; attribute, description | | | single: attribute; description (bundle) | | | single: description; bundle attribute | | | | | | Arbitrary text (not used by Pacemaker) | +-------------+-----------------------------------------------+ A bundle must contain exactly one ``docker``, ``podman``, or ``rkt`` element. .. index:: pair: XML element; docker pair: XML element; podman pair: XML element; rkt single: resource; bundle Bundle Container Properties ___________________________ .. table:: **XML attributes of a docker, podman, or rkt Element** +-------------------+------------------------------------+---------------------------------------------------+ | Attribute | Default | Description | +===================+====================================+===================================================+ | image | | .. index:: | | | | single: docker; attribute, image | | | | single: attribute; image (docker) | | | | single: image; docker attribute | | | | single: podman; attribute, image | | | | single: attribute; image (podman) | | | | single: image; podman attribute | | | | single: rkt; attribute, image | | | | single: attribute; image (rkt) | | | | single: image; rkt attribute | | | | | | | | Container image tag (required) | +-------------------+------------------------------------+---------------------------------------------------+ | replicas | Value of ``promoted-max`` | .. index:: | | | if that is positive, else 1 | single: docker; attribute, replicas | | | | single: attribute; replicas (docker) | | | | single: replicas; docker attribute | | | | single: podman; attribute, replicas | | | | single: attribute; replicas (podman) | | | | single: replicas; podman attribute | | | | single: rkt; attribute, replicas | | | | single: attribute; replicas (rkt) | | | | single: replicas; rkt attribute | | | | | | | | A positive integer specifying the number of | | | | container instances to launch | +-------------------+------------------------------------+---------------------------------------------------+ | replicas-per-host | 1 | .. index:: | | | | single: docker; attribute, replicas-per-host | | | | single: attribute; replicas-per-host (docker) | | | | single: replicas-per-host; docker attribute | | | | single: podman; attribute, replicas-per-host | | | | single: attribute; replicas-per-host (podman) | | | | single: replicas-per-host; podman attribute | | | | single: rkt; attribute, replicas-per-host | | | | single: attribute; replicas-per-host (rkt) | | | | single: replicas-per-host; rkt attribute | | | | | | | | A positive integer specifying the number of | | | | container instances allowed to run on a | | | | single node | +-------------------+------------------------------------+---------------------------------------------------+ | promoted-max | 0 | .. index:: | | | | single: docker; attribute, promoted-max | | | | single: attribute; promoted-max (docker) | | | | single: promoted-max; docker attribute | | | | single: podman; attribute, promoted-max | | | | single: attribute; promoted-max (podman) | | | | single: promoted-max; podman attribute | | | | single: rkt; attribute, promoted-max | | | | single: attribute; promoted-max (rkt) | | | | single: promoted-max; rkt attribute | | | | | | | | A non-negative integer that, if positive, | | | | indicates that the containerized service | | | | should be treated as a promotable service, | | | | with this many replicas allowed to run the | | | | service in the master role | +-------------------+------------------------------------+---------------------------------------------------+ | network | | .. index:: | | | | single: docker; attribute, network | | | | single: attribute; network (docker) | | | | single: network; docker attribute | | | | single: podman; attribute, network | | | | single: attribute; network (podman) | | | | single: network; podman attribute | | | | single: rkt; attribute, network | | | | single: attribute; network (rkt) | | | | single: network; rkt attribute | | | | | | | | If specified, this will be passed to the | | | | ``docker run``, ``podman run``, or | | | | ``rkt run`` command as the network setting | | | | for the container. | +-------------------+------------------------------------+---------------------------------------------------+ | run-command | ``/usr/sbin/pacemaker-remoted`` if | .. index:: | | | bundle contains a **primitive**, | single: docker; attribute, run-command | | | otherwise none | single: attribute; run-command (docker) | | | | single: run-command; docker attribute | | | | single: podman; attribute, run-command | | | | single: attribute; run-command (podman) | | | | single: run-command; podman attribute | | | | single: rkt; attribute, run-command | | | | single: attribute; run-command (rkt) | | | | single: run-command; rkt attribute | | | | | | | | This command will be run inside the container | | | | when launching it ("PID 1"). If the bundle | | | | contains a **primitive**, this command *must* | | | | start ``pacemaker-remoted`` (but could, for | | | | example, be a script that does other stuff, too). | +-------------------+------------------------------------+---------------------------------------------------+ | options | | .. index:: | | | | single: docker; attribute, options | | | | single: attribute; options (docker) | | | | single: options; docker attribute | | | | single: podman; attribute, options | | | | single: attribute; options (podman) | | | | single: options; podman attribute | | | | single: rkt; attribute, options | | | | single: attribute; options (rkt) | | | | single: options; rkt attribute | | | | | | | | Extra command-line options to pass to the | | | | ``docker run``, ``podman run``, or ``rkt run`` | | | | command | +-------------------+------------------------------------+---------------------------------------------------+ .. note:: Considerations when using cluster configurations or container images from Pacemaker 1.1: * If the container image has a pre-2.0.0 version of Pacemaker, set ``run-command`` to ``/usr/sbin/pacemaker_remoted`` (note the underbar instead of dash). * ``masters`` is accepted as an alias for ``promoted-max``, but is deprecated since 2.0.0, and support for it will be removed in a future version. Bundle Network Properties _________________________ A bundle may optionally contain one ```` element. .. index:: pair: XML element; network single: resource; bundle single: bundle; networking .. topic:: **XML attributes of a network Element** +----------------+---------+------------------------------------------------------------+ | Attribute | Default | Description | +================+=========+============================================================+ | add-host | TRUE | .. index:: | | | | single: network; attribute, add-host | | | | single: attribute; add-host (network) | | | | single: add-host; network attribute | | | | | | | | If TRUE, and ``ip-range-start`` is used, Pacemaker will | | | | automatically ensure that ``/etc/hosts`` inside the | | | | containers has entries for each | | | | :ref:`replica name ` | | | | and its assigned IP. | +----------------+---------+------------------------------------------------------------+ | ip-range-start | | .. index:: | | | | single: network; attribute, ip-range-start | | | | single: attribute; ip-range-start (network) | | | | single: ip-range-start; network attribute | | | | | | | | If specified, Pacemaker will create an implicit | | | | ``ocf:heartbeat:IPaddr2`` resource for each container | | | | instance, starting with this IP address, using up to | | | | ``replicas`` sequential addresses. These addresses can be | | | | used from the host's network to reach the service inside | | | | the container, though it is not visible within the | | | | container itself. Only IPv4 addresses are currently | | | | supported. | +----------------+---------+------------------------------------------------------------+ | host-netmask | 32 | .. index:: | | | | single: network; attribute; host-netmask | | | | single: attribute; host-netmask (network) | | | | single: host-netmask; network attribute | | | | | | | | If ``ip-range-start`` is specified, the IP addresses | | | | are created with this CIDR netmask (as a number of bits). | +----------------+---------+------------------------------------------------------------+ | host-interface | | .. index:: | | | | single: network; attribute; host-interface | | | | single: attribute; host-interface (network) | | | | single: host-interface; network attribute | | | | | | | | If ``ip-range-start`` is specified, the IP addresses are | | | | created on this host interface (by default, it will be | | | | determined from the IP address). | +----------------+---------+------------------------------------------------------------+ | control-port | 3121 | .. index:: | | | | single: network; attribute; control-port | | | | single: attribute; control-port (network) | | | | single: control-port; network attribute | | | | | | | | If the bundle contains a ``primitive``, the cluster will | | | | use this integer TCP port for communication with | | | | Pacemaker Remote inside the container. Changing this is | | | | useful when the container is unable to listen on the | | | | default port, for example, when the container uses the | | | | host's network rather than ``ip-range-start`` (in which | | | | case ``replicas-per-host`` must be 1), or when the bundle | | | | may run on a Pacemaker Remote node that is already | | | | listening on the default port. Any ``PCMK_remote_port`` | | | | environment variable set on the host or in the container | | | | is ignored for bundle connections. | +----------------+---------+------------------------------------------------------------+ .. _s-resource-bundle-note-replica-names: .. note:: Replicas are named by the bundle id plus a dash and an integer counter starting with zero. For example, if a bundle named **httpd-bundle** has **replicas=2**, its containers will be named **httpd-bundle-0** and **httpd-bundle-1**. .. index:: pair: XML element; port-mapping Additionally, a ``network`` element may optionally contain one or more ``port-mapping`` elements. .. table:: **Attributes of a port-mapping Element** +---------------+-------------------+------------------------------------------------------+ | Attribute | Default | Description | +===============+===================+======================================================+ | id | | .. index:: | | | | single: port-mapping; attribute, id | | | | single: attribute; id (port-mapping) | | | | single: id; port-mapping attribute | | | | | | | | A unique name for the port mapping (required) | +---------------+-------------------+------------------------------------------------------+ | port | | .. index:: | | | | single: port-mapping; attribute, port | | | | single: attribute; port (port-mapping) | | | | single: port; port-mapping attribute | | | | | | | | If this is specified, connections to this TCP port | | | | number on the host network (on the container's | | | | assigned IP address, if ``ip-range-start`` is | | | | specified) will be forwarded to the container | | | | network. Exactly one of ``port`` or ``range`` | | | | must be specified in a ``port-mapping``. | +---------------+-------------------+------------------------------------------------------+ | internal-port | value of ``port`` | .. index:: | | | | single: port-mapping; attribute, internal-port | | | | single: attribute; internal-port (port-mapping) | | | | single: internal-port; port-mapping attribute | | | | | | | | If ``port`` and this are specified, connections | | | | to ``port`` on the host's network will be | | | | forwarded to this port on the container network. | +---------------+-------------------+------------------------------------------------------+ | range | | .. index:: | | | | single: port-mapping; attribute, range | | | | single: attribute; range (port-mapping) | | | | single: range; port-mapping attribute | | | | | | | | If this is specified, connections to these TCP | | | | port numbers (expressed as *first_port*-*last_port*) | | | | on the host network (on the container's assigned IP | | | | address, if ``ip-range-start`` is specified) will | | | | be forwarded to the same ports in the container | | | | network. Exactly one of ``port`` or ``range`` | | | | must be specified in a ``port-mapping``. | +---------------+-------------------+------------------------------------------------------+ .. note:: If the bundle contains a ``primitive``, Pacemaker will automatically map the ``control-port``, so it is not necessary to specify that port in a ``port-mapping``. .. index: pair: XML element; storage pair: XML element; storage-mapping single: resource; bundle .. _s-bundle-storage: Bundle Storage Properties _________________________ A bundle may optionally contain one ``storage`` element. A ``storage`` element has no properties of its own, but may contain one or more ``storage-mapping`` elements. .. table:: **Attributes of a storage-mapping Element** +-----------------+---------+-------------------------------------------------------------+ | Attribute | Default | Description | +=================+=========+=============================================================+ | id | | .. index:: | | | | single: storage-mapping; attribute, id | | | | single: attribute; id (storage-mapping) | | | | single: id; storage-mapping attribute | | | | | | | | A unique name for the storage mapping (required) | +-----------------+---------+-------------------------------------------------------------+ | source-dir | | .. index:: | | | | single: storage-mapping; attribute, source-dir | | | | single: attribute; source-dir (storage-mapping) | | | | single: source-dir; storage-mapping attribute | | | | | | | | The absolute path on the host's filesystem that will be | | | | mapped into the container. Exactly one of ``source-dir`` | | | | and ``source-dir-root`` must be specified in a | | | | ``storage-mapping``. | +-----------------+---------+-------------------------------------------------------------+ | source-dir-root | | .. index:: | | | | single: storage-mapping; attribute, source-dir-root | | | | single: attribute; source-dir-root (storage-mapping) | | | | single: source-dir-root; storage-mapping attribute | | | | | | | | The start of a path on the host's filesystem that will | | | | be mapped into the container, using a different | | | | subdirectory on the host for each container instance. | | | | The subdirectory will be named the same as the | | | | :ref:`replica name `. | | | | Exactly one of ``source-dir`` and ``source-dir-root`` | | | | must be specified in a ``storage-mapping``. | +-----------------+---------+-------------------------------------------------------------+ | target-dir | | .. index:: | | | | single: storage-mapping; attribute, target-dir | | | | single: attribute; target-dir (storage-mapping) | | | | single: target-dir; storage-mapping attribute | | | | | | | | The path name within the container where the host | | | | storage will be mapped (required) | +-----------------+---------+-------------------------------------------------------------+ | options | | .. index:: | | | | single: storage-mapping; attribute, options | | | | single: attribute; options (storage-mapping) | | | | single: options; storage-mapping attribute | | | | | | | | A comma-separated list of file system mount | | | | options to use when mapping the storage | +-----------------+---------+-------------------------------------------------------------+ .. note:: Pacemaker does not define the behavior if the source directory does not already exist on the host. However, it is expected that the container technology and/or its resource agent will create the source directory in that case. .. note:: If the bundle contains a ``primitive``, Pacemaker will automatically map the equivalent of ``source-dir=/etc/pacemaker/authkey target-dir=/etc/pacemaker/authkey`` and ``source-dir-root=/var/log/pacemaker/bundles target-dir=/var/log`` into the container, so it is not necessary to specify those paths in a ``storage-mapping``. .. important:: The ``PCMK_authkey_location`` environment variable must not be set to anything other than the default of ``/etc/pacemaker/authkey`` on any node in the cluster. .. important:: If SELinux is used in enforcing mode on the host, you must ensure the container is allowed to use any storage you mount into it. For Docker and podman bundles, adding "Z" to the mount options will create a container-specific label for the mount that allows the container access. .. index:: single: resource; bundle Bundle Primitive ________________ A bundle may optionally contain one :ref:`primitive ` resource. The primitive may have operations, instance attributes, and meta-attributes defined, as usual. If a bundle contains a primitive resource, the container image must include the Pacemaker Remote daemon, and at least one of ``ip-range-start`` or ``control-port`` must be configured in the bundle. Pacemaker will create an implicit **ocf:pacemaker:remote** resource for the connection, launch Pacemaker Remote within the container, and monitor and manage the primitive resource via Pacemaker Remote. If the bundle has more than one container instance (replica), the primitive resource will function as an implicit :ref:`clone ` -- a :ref:`promotable clone ` if the bundle has ``promoted-max`` greater than zero. .. note:: If you want to pass environment variables to a bundle's Pacemaker Remote connection or primitive, you have two options: * Environment variables whose value is the same regardless of the underlying host may be set using the container element's ``options`` attribute. * If you want variables to have host-specific values, you can use the :ref:`storage-mapping ` element to map a file on the host as ``/etc/pacemaker/pcmk-init.env`` in the container *(since 2.0.3)*. Pacemaker Remote will parse this file as a shell-like format, with variables set as NAME=VALUE, ignoring blank lines and comments starting with "#". .. important:: When a bundle has a ``primitive``, Pacemaker on all cluster nodes must be able to contact Pacemaker Remote inside the bundle's containers. * The containers must have an accessible network (for example, ``network`` should not be set to "none" with a ``primitive``). * The default, using a distinct network space inside the container, works in combination with ``ip-range-start``. Any firewall must allow access from all cluster nodes to the ``control-port`` on the container IPs. * If the container shares the host's network space (for example, by setting ``network`` to "host"), a unique ``control-port`` should be specified for each bundle. Any firewall must allow access from all cluster nodes to the ``control-port`` on all cluster and remote node IPs. .. index:: single: resource; bundle .. _s-bundle-attributes: Bundle Node Attributes ______________________ If the bundle has a ``primitive``, the primitive's resource agent may want to set node attributes such as :ref:`promotion scores `. However, with containers, it is not apparent which node should get the attribute. If the container uses shared storage that is the same no matter which node the container is hosted on, then it is appropriate to use the promotion score on the bundle node itself. On the other hand, if the container uses storage exported from the underlying host, then it may be more appropriate to use the promotion score on the underlying host. Since this depends on the particular situation, the ``container-attribute-target`` resource meta-attribute allows the user to specify which approach to use. If it is set to ``host``, then user-defined node attributes will be checked on the underlying host. If it is anything else, the local node (in this case the bundle node) is used as usual. This only applies to user-defined attributes; the cluster will always check the local node for cluster-defined attributes such as ``#uname``. If ``container-attribute-target`` is ``host``, the cluster will pass additional environment variables to the primitive's resource agent that allow it to set node attributes appropriately: ``CRM_meta_container_attribute_target`` (identical to the meta-attribute value) and ``CRM_meta_physical_host`` (the name of the underlying host). .. note:: When called by a resource agent, the ``attrd_updater`` and ``crm_attribute`` commands will automatically check those environment variables and set attributes appropriately. .. index:: single: resource; bundle Bundle Meta-Attributes ______________________ Any meta-attribute set on a bundle will be inherited by the bundle's primitive and any resources implicitly created by Pacemaker for the bundle. This includes options such as ``priority``, ``target-role``, and ``is-managed``. See :ref:`resource_options` for more information. Limitations of Bundles ______________________ Restarting pacemaker while a bundle is unmanaged or the cluster is in maintenance mode may cause the bundle to fail. Bundles may not be explicitly cloned or included in groups. This includes the bundle's primitive and any resources implicitly created by Pacemaker for the bundle. (If ``replicas`` is greater than 1, the bundle will behave like a clone implicitly.) Bundles do not have instance attributes, utilization attributes, or operations, though a bundle's primitive may have them. A bundle with a primitive can run on a Pacemaker Remote node only if the bundle uses a distinct ``control-port``. .. [#] Of course, the service must support running multiple instances. .. [#] These are historical terms that will eventually be replaced, but the extensive use of them and the need for backward compatibility makes it a long process. You may see examples using a **master** tag instead of a **clone** tag with the **promotable** meta-attribute set to **true**; the **master** tag is supported, but deprecated, and will be removed in a future version. You may also see such services referred to as *multi-state* or *stateful*; these mean the same thing as *promotable*. .. [#] Docker is a trademark of Docker, Inc. No endorsement by or association with Docker, Inc. is implied. diff --git a/doc/sphinx/Pacemaker_Explained/constraints.rst b/doc/sphinx/Pacemaker_Explained/constraints.rst index c321eb971b..c917c79966 100644 --- a/doc/sphinx/Pacemaker_Explained/constraints.rst +++ b/doc/sphinx/Pacemaker_Explained/constraints.rst @@ -1,986 +1,1027 @@ .. index:: single: constraint single: resource; constraint .. _constraints: Resource Constraints -------------------- .. index:: single: resource; score single: node; score Scores ###### Scores of all kinds are integral to how the cluster works. Practically everything from moving a resource to deciding which resource to stop in a degraded cluster is achieved by manipulating scores in some way. Scores are calculated per resource and node. Any node with a negative score for a resource can't run that resource. The cluster places a resource on the node with the highest score for it. Infinity Math _____________ Pacemaker implements **INFINITY** (or equivalently, **+INFINITY**) internally as a score of 1,000,000. Addition and subtraction with it follow these three basic rules: * Any value + **INFINITY** = **INFINITY** * Any value - **INFINITY** = -**INFINITY** * **INFINITY** - **INFINITY** = **-INFINITY** .. note:: What if you want to use a score higher than 1,000,000? Typically this possibility arises when someone wants to base the score on some external metric that might go above 1,000,000. The short answer is you can't. The long answer is it is sometimes possible work around this limitation creatively. You may be able to set the score to some computed value based on the external metric rather than use the metric directly. For nodes, you can store the metric as a node attribute, and query the attribute when computing the score (possibly as part of a custom resource agent). .. _location-constraint: .. index:: single: location constraint single: constraint; location Deciding Which Nodes a Resource Can Run On ########################################## *Location constraints* tell the cluster which nodes a resource can run on. There are two alternative strategies. One way is to say that, by default, resources can run anywhere, and then the location constraints specify nodes that are not allowed (an *opt-out* cluster). The other way is to start with nothing able to run anywhere, and use location constraints to selectively enable allowed nodes (an *opt-in* cluster). Whether you should choose opt-in or opt-out depends on your personal preference and the make-up of your cluster. If most of your resources can run on most of the nodes, then an opt-out arrangement is likely to result in a simpler configuration. On the other-hand, if most resources can only run on a small subset of nodes, an opt-in configuration might be simpler. .. index:: pair: XML element; rsc_location - single: constraint; location single: constraint; rsc_location Location Properties ___________________ .. table:: **Attributes of a rsc_location Element** +--------------------+---------+----------------------------------------------------------------------------------------------+ | Attribute | Default | Description | +====================+=========+==============================================================================================+ | id | | .. index:: | | | | single: rsc_location; attribute, id | | | | single: attribute; id (rsc_location) | | | | single: id; rsc_location attribute | | | | | | | | A unique name for the constraint (required) | +--------------------+---------+----------------------------------------------------------------------------------------------+ | rsc | | .. index:: | | | | single: rsc_location; attribute, rsc | | | | single: attribute; rsc (rsc_location) | | | | single: rsc; rsc_location attribute | | | | | | | | The name of the resource to which this constraint | | | | applies. A location constraint must either have a | | | | ``rsc``, have a ``rsc-pattern``, or contain at | | | | least one resource set. | +--------------------+---------+----------------------------------------------------------------------------------------------+ | rsc-pattern | | .. index:: | | | | single: rsc_location; attribute, rsc-pattern | | | | single: attribute; rsc-pattern (rsc_location) | | | | single: rsc-pattern; rsc_location attribute | | | | | | | | A pattern matching the names of resources to which | | | | this constraint applies. The syntax is the same as | | | | `POSIX `_ | | | | extended regular expressions, with the addition of an | | | | initial *!* indicating that resources *not* matching | | | | the pattern are selected. If the regular expression | | | | contains submatches, and the constraint is governed by | | | | a :ref:`rule `, the submatches can be | | | | referenced as **%0** through **%9** in the rule's | | | | ``score-attribute`` or a rule expression's ``attribute``. | | | | A location constraint must either have a ``rsc``, have a | | | | ``rsc-pattern``, or contain at least one resource set. | +--------------------+---------+----------------------------------------------------------------------------------------------+ | node | | .. index:: | | | | single: rsc_location; attribute, node | | | | single: attribute; node (rsc_location) | | | | single: node; rsc_location attribute | | | | | | | | The name of the node to which this constraint applies. | | | | A location constraint must either have a ``node`` and | | | | ``score``, or contain at least one rule. | +--------------------+---------+----------------------------------------------------------------------------------------------+ | score | | .. index:: | | | | single: rsc_location; attribute, score | | | | single: attribute; score (rsc_location) | | | | single: score; rsc_location attribute | | | | | | | | Positive values indicate a preference for running the | | | | affected resource(s) on ``node`` -- the higher the value, | | | | the stronger the preference. Negative values indicate | | | | the resource(s) should avoid this node (a value of | | | | **-INFINITY** changes "should" to "must"). A location | | | | constraint must either have a ``node`` and ``score``, | | | | or contain at least one rule. | +--------------------+---------+----------------------------------------------------------------------------------------------+ | resource-discovery | always | .. index:: | | | | single: rsc_location; attribute, resource-discovery | | | | single: attribute; resource-discovery (rsc_location) | | | | single: resource-discovery; rsc_location attribute | | | | | | | | Whether Pacemaker should perform resource discovery | | | | (that is, check whether the resource is already running) | | | | for this resource on this node. This should normally be | | | | left as the default, so that rogue instances of a | | | | service can be stopped when they are running where they | | | | are not supposed to be. However, there are two | | | | situations where disabling resource discovery is a good | | | | idea: when a service is not installed on a node, | | | | discovery might return an error (properly written OCF | | | | agents will not, so this is usually only seen with other | | | | agent types); and when Pacemaker Remote is used to scale | | | | a cluster to hundreds of nodes, limiting resource | | | | discovery to allowed nodes can significantly boost | | | | performance. | | | | | | | | * ``always:`` Always perform resource discovery for | | | | the specified resource on this node. | | | | | | | | * ``never:`` Never perform resource discovery for the | | | | specified resource on this node. This option should | | | | generally be used with a -INFINITY score, although | | | | that is not strictly required. | | | | | | | | * ``exclusive:`` Perform resource discovery for the | | | | specified resource only on this node (and other nodes | | | | similarly marked as ``exclusive``). Multiple location | | | | constraints using ``exclusive`` discovery for the | | | | same resource across different nodes creates a subset | | | | of nodes resource-discovery is exclusive to. If a | | | | resource is marked for ``exclusive`` discovery on one | | | | or more nodes, that resource is only allowed to be | | | | placed within that subset of nodes. | +--------------------+---------+----------------------------------------------------------------------------------------------+ .. warning:: Setting ``resource-discovery`` to ``never`` or ``exclusive`` removes Pacemaker's ability to detect and stop unwanted instances of a service running where it's not supposed to be. It is up to the system administrator (you!) to make sure that the service can *never* be active on nodes without ``resource-discovery`` (such as by leaving the relevant software uninstalled). .. index:: single: Asymmetrical Clusters single: Opt-In Clusters Asymmetrical "Opt-In" Clusters ______________________________ To create an opt-in cluster, start by preventing resources from running anywhere by default: .. code-block:: none # crm_attribute --name symmetric-cluster --update false Then start enabling nodes. The following fragment says that the web server prefers **sles-1**, the database prefers **sles-2** and both can fail over to **sles-3** if their most preferred node fails. .. topic:: Opt-in location constraints for two resources .. code-block:: xml .. index:: single: Symmetrical Clusters single: Opt-Out Clusters Symmetrical "Opt-Out" Clusters ______________________________ To create an opt-out cluster, start by allowing resources to run anywhere by default: .. code-block:: none # crm_attribute --name symmetric-cluster --update true Then start disabling nodes. The following fragment is the equivalent of the above opt-in configuration. .. topic:: Opt-out location constraints for two resources .. code-block:: xml .. _node-score-equal: What if Two Nodes Have the Same Score _____________________________________ If two nodes have the same score, then the cluster will choose one. This choice may seem random and may not be what was intended, however the cluster was not given enough information to know any better. .. topic:: Constraints where a resource prefers two nodes equally .. code-block:: xml In the example above, assuming no other constraints and an inactive cluster, **Webserver** would probably be placed on **sles-1** and **Database** on **sles-2**. It would likely have placed **Webserver** based on the node's uname and **Database** based on the desire to spread the resource load evenly across the cluster. However other factors can also be involved in more complex configurations. .. index:: single: constraint; ordering single: resource; start order .. _s-resource-ordering: Specifying the Order in which Resources Should Start/Stop ######################################################### *Ordering constraints* tell the cluster the order in which certain resource actions should occur. .. important:: Ordering constraints affect *only* the ordering of resource actions; they do *not* require that the resources be placed on the same node. If you want resources to be started on the same node *and* in a specific order, you need both an ordering constraint *and* a colocation constraint (see :ref:`s-resource-colocation`), or alternatively, a group (see :ref:`group-resources`). .. index:: pair: XML element; rsc_order - pair: constraint; ordering + pair: constraint; rsc_order Ordering Properties ___________________ .. table:: **Attributes of a rsc_order Element** +--------------+----------------------------+-------------------------------------------------------------------+ | Field | Default | Description | +==============+============================+===================================================================+ | id | | .. index:: | | | | single: rsc_order; attribute, id | | | | single: attribute; id (rsc_order) | | | | single: id; rsc_order attribute | | | | | | | | A unique name for the constraint | +--------------+----------------------------+-------------------------------------------------------------------+ | first | | .. index:: | | | | single: rsc_order; attribute, first | | | | single: attribute; first (rsc_order) | | | | single: first; rsc_order attribute | | | | | | | | Name of the resource that the ``then`` resource | | | | depends on | +--------------+----------------------------+-------------------------------------------------------------------+ | then | | .. index:: | | | | single: rsc_order; attribute, then | | | | single: attribute; then (rsc_order) | | | | single: then; rsc_order attribute | | | | | | | | Name of the dependent resource | +--------------+----------------------------+-------------------------------------------------------------------+ | first-action | start | .. index:: | | | | single: rsc_order; attribute, first-action | | | | single: attribute; first-action (rsc_order) | | | | single: first-action; rsc_order attribute | | | | | | | | The action that the ``first`` resource must complete | | | | before ``then-action`` can be initiated for the ``then`` | | | | resource. Allowed values: ``start``, ``stop``, | | | | ``promote``, ``demote``. | +--------------+----------------------------+-------------------------------------------------------------------+ | then-action | value of ``first-action`` | .. index:: | | | | single: rsc_order; attribute, then-action | | | | single: attribute; then-action (rsc_order) | | | | single: first-action; rsc_order attribute | | | | | | | | The action that the ``then`` resource can execute only | | | | after the ``first-action`` on the ``first`` resource has | | | | completed. Allowed values: ``start``, ``stop``, | | | | ``promote``, ``demote``. | +--------------+----------------------------+-------------------------------------------------------------------+ | kind | Mandatory | .. index:: | | | | single: rsc_order; attribute, kind | | | | single: attribute; kind (rsc_order) | | | | single: kind; rsc_order attribute | | | | | | | | How to enforce the constraint. Allowed values: | | | | | | | | * ``Mandatory:`` ``then-action`` will never be initiated | | | | for the ``then`` resource unless and until ``first-action`` | | | | successfully completes for the ``first`` resource. | | | | | | | | * ``Optional:`` The constraint applies only if both specified | | | | resource actions are scheduled in the same transition | | | | (that is, in response to the same cluster state). This | | | | means that ``then-action`` is allowed on the ``then`` | | | | resource regardless of the state of the ``first`` resource, | | | | but if both actions happen to be scheduled at the same time, | | | | they will be ordered. | | | | | | | | * ``Serialize:`` Ensure that the specified actions are never | | | | performed concurrently for the specified resources. | | | | ``First-action`` and ``then-action`` can be executed in either | | | | order, but one must complete before the other can be initiated. | | | | An example use case is when resource start-up puts a high load | | | | on the host. | +--------------+----------------------------+-------------------------------------------------------------------+ | symmetrical | TRUE for ``Mandatory`` and | .. index:: | | | ``Optional`` kinds. FALSE | single: rsc_order; attribute, symmetrical | | | for ``Serialize`` kind. | single: attribute; symmetrical (rsc)order) | | | | single: symmetrical; rsc_order attribute | | | | | | | | If true, the reverse of the constraint applies for the | | | | opposite action (for example, if B starts after A starts, | | | | then B stops before A stops). ``Serialize`` orders cannot | | | | be symmetrical. | +--------------+----------------------------+-------------------------------------------------------------------+ ``Promote`` and ``demote`` apply to the master role of :ref:`promotable ` resources. Optional and mandatory ordering _______________________________ Here is an example of ordering constraints where **Database** *must* start before **Webserver**, and **IP** *should* start before **Webserver** if they both need to be started: .. topic:: Optional and mandatory ordering constraints .. code-block:: xml Because the above example lets ``symmetrical`` default to TRUE, **Webserver** must be stopped before **Database** can be stopped, and **Webserver** should be stopped before **IP** if they both need to be stopped. .. index:: + single: colocation single: constraint; colocation single: resource; location relative to other resources .. _s-resource-colocation: Placing Resources Relative to other Resources ############################################# *Colocation constraints* tell the cluster that the location of one resource depends on the location of another one. Colocation has an important side-effect: it affects the order in which resources are assigned to a node. Think about it: You can't place A relative to B unless you know where B is [#]_. So when you are creating colocation constraints, it is important to consider whether you should colocate A with B, or B with A. -Another thing to keep in mind is that, assuming A is colocated with -B, the cluster will take into account A's preferences when -deciding which node to choose for B. - -For a detailed look at exactly how this occurs, see -`Colocation Explained `_. - .. important:: Colocation constraints affect *only* the placement of resources; they do *not* require that the resources be started in a particular order. If you want resources to be started on the same node *and* in a specific order, you need both an ordering constraint (see :ref:`s-resource-ordering`) *and* a colocation constraint, or alternatively, a group (see :ref:`group-resources`). .. index:: pair: XML element; rsc_colocation - pair: constraint; colocation + single: constraint; rsc_colocation Colocation Properties _____________________ .. table:: **Attributes of a rsc_colocation Constraint** +----------------+---------+--------------------------------------------------------+ | Field | Default | Description | +================+=========+========================================================+ | id | | .. index:: | | | | single: rsc_colocation; attribute, id | | | | single: attribute; id (rsc_colocation) | | | | single: id; rsc_colocation attribute | | | | | | | | A unique name for the constraint (required). | +----------------+---------+--------------------------------------------------------+ | rsc | | .. index:: | | | | single: rsc_colocation; attribute, rsc | | | | single: attribute; rsc (rsc_colocation) | | | | single: rsc; rsc_colocation attribute | | | | | | | | The name of a resource that should be located | - | | | relative to ``with-rsc`` (required). | + | | | relative to ``with-rsc``. A colocation constraint must | + | | | either contain at least one | + | | | :ref:`resource set `, or specify both | + | | | ``rsc`` and ``with-rsc``. | +----------------+---------+--------------------------------------------------------+ | with-rsc | | .. index:: | | | | single: rsc_colocation; attribute, with-rsc | | | | single: attribute; with-rsc (rsc_colocation) | | | | single: with-rsc; rsc_colocation attribute | | | | | | | | The name of the resource used as the colocation | | | | target. The cluster will decide where to put this | - | | | resource first and then decide where to put | - | | | ``rsc`` (required). | + | | | resource first and then decide where to put ``rsc``. | + | | | A colocation constraint must either contain at least | + | | | one :ref:`resource set `, or specify | + | | | both ``rsc`` and ``with-rsc``. | +----------------+---------+--------------------------------------------------------+ | node-attribute | #uname | .. index:: | | | | single: rsc_colocation; attribute, node-attribute | | | | single: attribute; node-attribute (rsc_colocation) | | | | single: node-attribute; rsc_colocation attribute | | | | | - | | | The node attribute that must be the same on the | - | | | node running ``rsc`` and the node running ``with-rsc`` | - | | | for the constraint to be satisfied. (For details, | - | | | see :ref:`s-coloc-attribute`.) | + | | | If ``rsc`` and ``with-rsc`` are specified, this node | + | | | attribute must be the same on the node running ``rsc`` | + | | | and the node running ``with-rsc`` for the constraint | + | | | to be satisfied. (For details, see | + | | | :ref:`s-coloc-attribute`.) | +----------------+---------+--------------------------------------------------------+ - | score | | .. index:: | + | score | 0 | .. index:: | | | | single: rsc_colocation; attribute, score | | | | single: attribute; score (rsc_colocation) | | | | single: score; rsc_colocation attribute | | | | | | | | Positive values indicate the resources should run on | | | | the same node. Negative values indicate the resources | | | | should run on different nodes. Values of | - | | | +/- **INFINITY** change "should" to "must". | + | | | +/- ``INFINITY`` change "should" to "must". | + +----------------+---------+--------------------------------------------------------+ + | rsc-role | Started | .. index:: | + | | | single: clone; ordering constraint, rsc-role | + | | | single: ordering constraint; rsc-role (clone) | + | | | single: rsc-role; clone ordering constraint | + | | | | + | | | If ``rsc`` and ``with-rsc`` are specified, and ``rsc`` | + | | | is a :ref:`promotable clone `, | + | | | the constraint applies only to ``rsc`` instances in | + | | | this role. Allowed values: ``Started``, ``Master``, | + | | | ``Slave``. For details, see | + | | | :ref:`promotable-clone-constraints`. | + +----------------+---------+--------------------------------------------------------+ + | with-rsc-role | Started | .. index:: | + | | | single: clone; ordering constraint, with-rsc-role | + | | | single: ordering constraint; with-rsc-role (clone) | + | | | single: with-rsc-role; clone ordering constraint | + | | | | + | | | If ``rsc`` and ``with-rsc`` are specified, and | + | | | ``with-rsc`` is a | + | | | :ref:`promotable clone `, the | + | | | constraint applies only to ``with-rsc`` instances in | + | | | this role. Allowed values: ``Started``, ``Master``, | + | | | ``Slave``. For details, see | + | | | :ref:`promotable-clone-constraints`. | +----------------+---------+--------------------------------------------------------+ Mandatory Placement ___________________ Mandatory placement occurs when the constraint's score is **+INFINITY** or **-INFINITY**. In such cases, if the constraint can't be satisfied, then the **rsc** resource is not permitted to run. For ``score=INFINITY``, this includes cases where the ``with-rsc`` resource is not active. If you need resource **A** to always run on the same machine as resource **B**, you would add the following constraint: .. topic:: Mandatory colocation constraint for two resources .. code-block:: xml Remember, because **INFINITY** was used, if **B** can't run on any of the cluster nodes (for whatever reason) then **A** will not be allowed to run. Whether **A** is running or not has no effect on **B**. Alternatively, you may want the opposite -- that **A** *cannot* run on the same machine as **B**. In this case, use ``score="-INFINITY"``. .. topic:: Mandatory anti-colocation constraint for two resources .. code-block:: xml Again, by specifying **-INFINITY**, the constraint is binding. So if the only place left to run is where **B** already is, then **A** may not run anywhere. As with **INFINITY**, **B** can run even if **A** is stopped. However, in this case **A** also can run if **B** is stopped, because it still meets the constraint of **A** and **B** not running on the same node. Advisory Placement __________________ If mandatory placement is about "must" and "must not", then advisory placement is the "I'd prefer if" alternative. For constraints with scores greater than **-INFINITY** and less than **INFINITY**, the cluster will try to accommodate your wishes but may ignore them if the alternative is to stop some of the cluster resources. As in life, where if enough people prefer something it effectively becomes mandatory, advisory colocation constraints can combine with other elements of the configuration to behave as if they were mandatory. .. topic:: Advisory colocation constraint for two resources .. code-block:: xml .. _s-coloc-attribute: Colocation by Node Attribute ____________________________ The ``node-attribute`` property of a colocation constraints allows you to express the requirement, "these resources must be on similar nodes". As an example, imagine that you have two Storage Area Networks (SANs) that are not controlled by the cluster, and each node is connected to one or the other. You may have two resources **r1** and **r2** such that **r2** needs to use the same SAN as **r1**, but doesn't necessarily have to be on the same exact node. In such a case, you could define a :ref:`node attribute ` named **san**, with the value **san1** or **san2** on each node as appropriate. Then, you could colocate **r2** with **r1** using ``node-attribute`` set to **san**. +.. _s-coloc-influence: + +Colocation Influence +____________________ + +If A is colocated with B, the cluster will take into account A's preferences +when deciding where to place B, to maximize the chance that both resources can +run. + +For a detailed look at exactly how this occurs, see +`Colocation Explained `_. + + .. _s-resource-sets: Resource Sets ############# .. index:: single: constraint; resource set single: resource; resource set *Resource sets* allow multiple resources to be affected by a single constraint. .. topic:: A set of 3 resources .. code-block:: xml Resource sets are valid inside ``rsc_location``, ``rsc_order`` (see :ref:`s-resource-sets-ordering`), ``rsc_colocation`` (see :ref:`s-resource-sets-colocation`), and ``rsc_ticket`` (see :ref:`ticket-constraints`) constraints. A resource set has a number of properties that can be set, though not all have an effect in all contexts. .. index:: pair: XML element; resource_set .. topic:: **Attributes of a resource_set Element** - +-------------+---------+--------------------------------------------------------+ - | Field | Default | Description | - +=============+=========+========================================================+ - | id | | .. index:: | - | | | single: resource_set; attribute, id | - | | | single: attribute; id (resource_set) | - | | | single: id; resource_set attribute | - | | | | - | | | A unique name for the set | - +-------------+---------+--------------------------------------------------------+ - | sequential | true | .. index:: | - | | | single: resource_set; attribute, sequential | - | | | single: attribute; sequential (resource_set) | - | | | single: sequential; resource_set attribute | - | | | | - | | | Whether the members of the set must be acted on in | - | | | order. Meaningful within ``rsc_order`` and | - | | | ``rsc_colocation``. | - +-------------+---------+--------------------------------------------------------+ - | require-all | true | .. index:: | - | | | single: resource_set; attribute, require-all | - | | | single: attribute; require-all (resource_set) | - | | | single: require-all; resource_set attribute | - | | | | - | | | Whether all members of the set must be active before | - | | | continuing. With the current implementation, the | - | | | cluster may continue even if only one member of the | - | | | set is started, but if more than one member of the set | - | | | is starting at the same time, the cluster will still | - | | | wait until all of those have started before continuing | - | | | (this may change in future versions). Meaningful | - | | | within ``rsc_order``. | - +-------------+---------+--------------------------------------------------------+ - | role | | .. index:: | - | | | single: resource_set; attribute, role | - | | | single: attribute; role (resource_set) | - | | | single: role; resource_set attribute | - | | | | - | | | Limit the effect of the constraint to the specified | - | | | role. Meaningful within ``rsc_location``, | - | | | ``rsc_colocation`` and ``rsc_ticket``. | - +-------------+---------+--------------------------------------------------------+ - | action | | .. index:: | - | | | single: resource_set; attribute, action | - | | | single: attribute; action (resource_set) | - | | | single: action; resource_set attribute | - | | | | - | | | Limit the effect of the constraint to the specified | - | | | action. Meaningful within ``rsc_order``. | - +-------------+---------+--------------------------------------------------------+ - | score | | .. index:: | - | | | single: resource_set; attribute, score | - | | | single: attribute; score (resource_set) | - | | | single: score; resource_set attribute | - | | | | - | | | *Advanced use only.* Use a specific score for this | - | | | set within the constraint. | - +-------------+---------+--------------------------------------------------------+ + +-------------+------------------+--------------------------------------------------------+ + | Field | Default | Description | + +=============+==================+========================================================+ + | id | | .. index:: | + | | | single: resource_set; attribute, id | + | | | single: attribute; id (resource_set) | + | | | single: id; resource_set attribute | + | | | | + | | | A unique name for the set (required) | + +-------------+------------------+--------------------------------------------------------+ + | sequential | true | .. index:: | + | | | single: resource_set; attribute, sequential | + | | | single: attribute; sequential (resource_set) | + | | | single: sequential; resource_set attribute | + | | | | + | | | Whether the members of the set must be acted on in | + | | | order. Meaningful within ``rsc_order`` and | + | | | ``rsc_colocation``. | + +-------------+------------------+--------------------------------------------------------+ + | require-all | true | .. index:: | + | | | single: resource_set; attribute, require-all | + | | | single: attribute; require-all (resource_set) | + | | | single: require-all; resource_set attribute | + | | | | + | | | Whether all members of the set must be active before | + | | | continuing. With the current implementation, the | + | | | cluster may continue even if only one member of the | + | | | set is started, but if more than one member of the set | + | | | is starting at the same time, the cluster will still | + | | | wait until all of those have started before continuing | + | | | (this may change in future versions). Meaningful | + | | | within ``rsc_order``. | + +-------------+------------------+--------------------------------------------------------+ + | role | | .. index:: | + | | | single: resource_set; attribute, role | + | | | single: attribute; role (resource_set) | + | | | single: role; resource_set attribute | + | | | | + | | | The constraint applies only to resource set members | + | | | that are :ref:`s-resource-promotable` in this | + | | | role. Meaningful within ``rsc_location``, | + | | | ``rsc_colocation`` and ``rsc_ticket``. | + | | | Allowed values: ``Started``, ``Master``, ``Slave``. | + | | | For details, see :ref:`promotable-clone-constraints`. | + +-------------+------------------+--------------------------------------------------------+ + | action | value of | .. index:: | + | | ``first-action`` | single: resource_set; attribute, action | + | | in the enclosing | single: attribute; action (resource_set) | + | | ordering | single: action; resource_set attribute | + | | constraint | | + | | | The action that applies to *all members* of the set. | + | | | Meaningful within ``rsc_order``. Allowed values: | + | | | ``start``, ``stop``, ``promote``, ``demote``. | + +-------------+------------------+--------------------------------------------------------+ + | score | | .. index:: | + | | | single: resource_set; attribute, score | + | | | single: attribute; score (resource_set) | + | | | single: score; resource_set attribute | + | | | | + | | | *Advanced use only.* Use a specific score for this | + | | | set within the constraint. | + +-------------+------------------+--------------------------------------------------------+ .. _s-resource-sets-ordering: Ordering Sets of Resources ########################## A common situation is for an administrator to create a chain of ordered resources, such as: .. topic:: A chain of ordered resources .. code-block:: xml .. topic:: Visual representation of the four resources' start order for the above constraints .. image:: images/resource-set.png :alt: Ordered set Ordered Set ___________ -To simplify this situation, resource sets (see :ref:`s-resource-sets`) can be used -within ordering constraints: +To simplify this situation, :ref:`s-resource-sets` can be used within ordering +constraints: .. topic:: A chain of ordered resources expressed as a set .. code-block:: xml While the set-based format is not less verbose, it is significantly easier to get right and maintain. .. important:: If you use a higher-level tool, pay attention to how it exposes this functionality. Depending on the tool, creating a set **A B** may be equivalent to **A then B**, or **B then A**. Ordering Multiple Sets ______________________ The syntax can be expanded to allow sets of resources to be ordered relative to each other, where the members of each individual set may be ordered or unordered (controlled by the ``sequential`` property). In the example below, **A** and **B** can both start in parallel, as can **C** and **D**, however **C** and **D** can only start once *both* **A** *and* **B** are active. .. topic:: Ordered sets of unordered resources .. code-block:: xml .. topic:: Visual representation of the start order for two ordered sets of unordered resources .. image:: images/two-sets.png :alt: Two ordered sets Of course either set -- or both sets -- of resources can also be internally ordered (by setting ``sequential="true"``) and there is no limit to the number of sets that can be specified. .. topic:: Advanced use of set ordering - Three ordered sets, two of which are internally unordered .. code-block:: xml .. topic:: Visual representation of the start order for the three sets defined above .. image:: images/three-sets.png :alt: Three ordered sets .. important:: An ordered set with ``sequential=false`` makes sense only if there is another set in the constraint. Otherwise, the constraint has no effect. Resource Set OR Logic _____________________ The unordered set logic discussed so far has all been "AND" logic. To illustrate this take the 3 resource set figure in the previous section. Those sets can be expressed, **(A and B) then (C) then (D) then (E and F)**. Say for example we want to change the first set, **(A and B)**, to use "OR" logic so the sets look like this: **(A or B) then (C) then (D) then (E and F)**. This functionality can be achieved through the use of the ``require-all`` option. This option defaults to TRUE which is why the "AND" logic is used by default. Setting ``require-all=false`` means only one resource in the set needs to be started before continuing on to the next set. .. topic:: Resource Set "OR" logic: Three ordered sets, where the first set is internally unordered with "OR" logic .. code-block:: xml .. important:: An ordered set with ``require-all=false`` makes sense only in conjunction with ``sequential=false``. Think of it like this: ``sequential=false`` modifies the set to be an unordered set using "AND" logic by default, and adding ``require-all=false`` flips the unordered set's "AND" logic to "OR" logic. .. _s-resource-sets-colocation: Colocating Sets of Resources ############################ Another common situation is for an administrator to create a set of colocated resources. The simplest way to do this is to define a resource group (see :ref:`group-resources`), but that cannot always accurately express the desired relationships. For example, maybe the resources do not need to be ordered. Another way would be to define each relationship as an individual constraint, but that causes a difficult-to-follow constraint explosion as the number of resources and combinations grow. .. topic:: Colocation chain as individual constraints, where A is placed first, then B, then C, then D .. code-block:: xml To express complicated relationships with a simplified syntax [#]_, :ref:`resource sets ` can be used within colocation constraints. .. topic:: Equivalent colocation chain expressed using **resource_set** .. code-block:: xml .. note:: Within a ``resource_set``, the resources are listed in the order they are *placed*, which is the reverse of the order in which they are *colocated*. In the above example, resource **A** is placed before resource **B**, which is the same as saying resource **B** is colocated with resource **A**. As with individual constraints, a resource that can't be active prevents any resource that must be colocated with it from being active. In both of the two previous examples, if **B** is unable to run, then both **C** and by inference **D** must remain stopped. .. important:: If you use a higher-level tool, pay attention to how it exposes this functionality. Depending on the tool, creating a set **A B** may be equivalent to **A with B**, or **B with A**. Resource sets can also be used to tell the cluster that entire *sets* of resources must be colocated relative to each other, while the individual members within any one set may or may not be colocated relative to each other (determined by the set's ``sequential`` property). In the following example, resources **B**, **C**, and **D** will each be colocated with **A** (which will be placed first). **A** must be able to run in order for any of the resources to run, but any of **B**, **C**, or **D** may be stopped without affecting any of the others. .. topic:: Using colocated sets to specify a shared dependency .. code-block:: xml .. note:: Pay close attention to the order in which resources and sets are listed. While the members of any one sequential set are placed first to last (i.e., the colocation dependency is last with first), multiple sets are placed last to first (i.e. the colocation dependency is first with last). .. important:: A colocated set with ``sequential="false"`` makes sense only if there is another set in the constraint. Otherwise, the constraint has no effect. There is no inherent limit to the number and size of the sets used. The only thing that matters is that in order for any member of one set in the constraint to be active, all members of sets listed after it must also be active (and naturally on the same node); and if a set has ``sequential="true"``, then in order for one member of that set to be active, all members listed before it must also be active. If desired, you can restrict the dependency to instances of promotable clone resources that are in a specific role, using the set's ``role`` property. .. topic:: Colocation in which the members of the middle set have no interdependencies, and the last set listed applies only to instances in the master role .. code-block:: xml .. topic:: Visual representation of the above example (resources are placed from left to right) .. image:: ../shared/images/pcmk-colocated-sets.png :alt: Colocation chain .. note:: Unlike ordered sets, colocated sets do not use the ``require-all`` option. .. [#] While the human brain is sophisticated enough to read the constraint in any order and choose the correct one depending on the situation, the cluster is not quite so smart. Yet. .. [#] which is not the same as saying easy to follow diff --git a/include/crm/pengine/pe_types.h b/include/crm/pengine/pe_types.h index 9fea637ced..59d5ce8b49 100644 --- a/include/crm/pengine/pe_types.h +++ b/include/crm/pengine/pe_types.h @@ -1,539 +1,539 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef PE_TYPES__H # define PE_TYPES__H #ifdef __cplusplus extern "C" { #endif /*! * \file * \brief Data types for cluster status * \ingroup pengine */ # include // bool # include // time_t # include // gboolean, guint, GList, GHashTable # include // GListPtr # include # include typedef struct pe_node_s pe_node_t; typedef struct pe_action_s pe_action_t; typedef struct pe_resource_s pe_resource_t; typedef struct pe_working_set_s pe_working_set_t; enum pe_obj_types { pe_unknown = -1, pe_native = 0, pe_group = 1, pe_clone = 2, pe_container = 3, }; typedef struct resource_object_functions_s { gboolean (*unpack) (pe_resource_t*, pe_working_set_t*); pe_resource_t *(*find_rsc) (pe_resource_t *parent, const char *search, const pe_node_t *node, int flags); /* parameter result must be free'd */ char *(*parameter) (pe_resource_t*, pe_node_t*, gboolean, const char*, pe_working_set_t*); void (*print) (pe_resource_t*, const char*, long, void*); gboolean (*active) (pe_resource_t*, gboolean); enum rsc_role_e (*state) (const pe_resource_t*, gboolean); pe_node_t *(*location) (const pe_resource_t*, GList**, int); void (*free) (pe_resource_t*); void (*count) (pe_resource_t*); gboolean (*is_filtered) (pe_resource_t*, GListPtr, gboolean); } resource_object_functions_t; typedef struct resource_alloc_functions_s resource_alloc_functions_t; enum pe_quorum_policy { no_quorum_freeze, no_quorum_stop, no_quorum_ignore, no_quorum_suicide, no_quorum_demote }; enum node_type { node_ping, node_member, node_remote }; //! \deprecated will be removed in a future release enum pe_restart { pe_restart_restart, pe_restart_ignore }; //! Determine behavior of pe_find_resource_with_flags() enum pe_find { pe_find_renamed = 0x001, //!< match resource ID or LRM history ID pe_find_anon = 0x002, //!< match base name of anonymous clone instances pe_find_clone = 0x004, //!< match only clone instances pe_find_current = 0x008, //!< match resource active on specified node pe_find_inactive = 0x010, //!< match resource not running anywhere pe_find_any = 0x020, //!< match base name of any clone instance }; // @TODO Make these an enum # define pe_flag_have_quorum 0x00000001ULL # define pe_flag_symmetric_cluster 0x00000002ULL # define pe_flag_maintenance_mode 0x00000008ULL # define pe_flag_stonith_enabled 0x00000010ULL # define pe_flag_have_stonith_resource 0x00000020ULL # define pe_flag_enable_unfencing 0x00000040ULL # define pe_flag_concurrent_fencing 0x00000080ULL # define pe_flag_stop_rsc_orphans 0x00000100ULL # define pe_flag_stop_action_orphans 0x00000200ULL # define pe_flag_stop_everything 0x00000400ULL # define pe_flag_start_failure_fatal 0x00001000ULL # define pe_flag_remove_after_stop 0x00002000ULL # define pe_flag_startup_fencing 0x00004000ULL # define pe_flag_shutdown_lock 0x00008000ULL # define pe_flag_startup_probes 0x00010000ULL # define pe_flag_have_status 0x00020000ULL # define pe_flag_have_remote_nodes 0x00040000ULL # define pe_flag_quick_location 0x00100000ULL # define pe_flag_sanitized 0x00200000ULL # define pe_flag_stdout 0x00400000ULL //! Don't count total, disabled and blocked resource instances # define pe_flag_no_counts 0x00800000ULL /*! Skip deprecated code that is kept solely for backward API compatibility. * (Internal code should always set this.) */ # define pe_flag_no_compat 0x01000000ULL struct pe_working_set_s { xmlNode *input; crm_time_t *now; /* options extracted from the input */ char *dc_uuid; pe_node_t *dc_node; const char *stonith_action; const char *placement_strategy; unsigned long long flags; int stonith_timeout; enum pe_quorum_policy no_quorum_policy; GHashTable *config_hash; GHashTable *tickets; // Actions for which there can be only one (e.g. fence nodeX) GHashTable *singletons; GListPtr nodes; GListPtr resources; GListPtr placement_constraints; GListPtr ordering_constraints; GListPtr colocation_constraints; GListPtr ticket_constraints; GListPtr actions; xmlNode *failed; xmlNode *op_defaults; xmlNode *rsc_defaults; /* stats */ int num_synapse; int max_valid_nodes; //! Deprecated (will be removed in a future release) int order_id; int action_id; /* final output */ xmlNode *graph; GHashTable *template_rsc_sets; const char *localhost; GHashTable *tags; int blocked_resources; int disabled_resources; GList *param_check; // History entries that need to be checked GList *stop_needed; // Containers that need stop actions time_t recheck_by; // Hint to controller to re-run scheduler by this time int ninstances; // Total number of resource instances guint shutdown_lock;// How long (seconds) to lock resources to shutdown node int priority_fencing_delay; // Priority fencing delay }; enum pe_check_parameters { /* Clear fail count if parameters changed for un-expired start or monitor * last_failure. */ pe_check_last_failure, /* Clear fail count if parameters changed for start, monitor, promote, or * migrate_from actions for active resources. */ pe_check_active, }; struct pe_node_shared_s { const char *id; const char *uname; enum node_type type; /* @TODO convert these flags into a bitfield */ gboolean online; gboolean standby; gboolean standby_onfail; gboolean pending; gboolean unclean; gboolean unseen; gboolean shutdown; gboolean expected_up; gboolean is_dc; gboolean maintenance; gboolean rsc_discovery_enabled; gboolean remote_requires_reset; gboolean remote_was_fenced; gboolean remote_maintenance; /* what the remote-rsc is thinking */ gboolean unpacked; int num_resources; pe_resource_t *remote_rsc; GListPtr running_rsc; /* pe_resource_t* */ GListPtr allocated_rsc; /* pe_resource_t* */ GHashTable *attrs; /* char* => char* */ GHashTable *utilization; GHashTable *digest_cache; //!< cache of calculated resource digests int priority; // calculated based on the priority of resources running on the node }; struct pe_node_s { int weight; gboolean fixed; int count; struct pe_node_shared_s *details; int rsc_discover_mode; }; # define pe_rsc_orphan 0x00000001ULL # define pe_rsc_managed 0x00000002ULL # define pe_rsc_block 0x00000004ULL # define pe_rsc_orphan_container_filler 0x00000008ULL # define pe_rsc_notify 0x00000010ULL # define pe_rsc_unique 0x00000020ULL # define pe_rsc_fence_device 0x00000040ULL # define pe_rsc_promotable 0x00000080ULL # define pe_rsc_provisional 0x00000100ULL # define pe_rsc_allocating 0x00000200ULL # define pe_rsc_merging 0x00000400ULL # define pe_rsc_stop 0x00001000ULL # define pe_rsc_reload 0x00002000ULL # define pe_rsc_allow_remote_remotes 0x00004000ULL # define pe_rsc_failed 0x00010000ULL # define pe_rsc_runnable 0x00040000ULL # define pe_rsc_start_pending 0x00080000ULL # define pe_rsc_starting 0x00100000ULL # define pe_rsc_stopping 0x00200000ULL # define pe_rsc_allow_migrate 0x00800000ULL # define pe_rsc_failure_ignored 0x01000000ULL # define pe_rsc_maintenance 0x04000000ULL # define pe_rsc_is_container 0x08000000ULL # define pe_rsc_needs_quorum 0x10000000ULL # define pe_rsc_needs_fencing 0x20000000ULL # define pe_rsc_needs_unfencing 0x40000000ULL enum pe_graph_flags { pe_graph_none = 0x00000, pe_graph_updated_first = 0x00001, pe_graph_updated_then = 0x00002, pe_graph_disable = 0x00004, }; /* *INDENT-OFF* */ enum pe_action_flags { pe_action_pseudo = 0x00001, pe_action_runnable = 0x00002, pe_action_optional = 0x00004, pe_action_print_always = 0x00008, pe_action_have_node_attrs = 0x00010, pe_action_implied_by_stonith = 0x00040, pe_action_migrate_runnable = 0x00080, pe_action_dumped = 0x00100, pe_action_processed = 0x00200, pe_action_clear = 0x00400, pe_action_dangle = 0x00800, /* This action requires one or more of its dependencies to be runnable. * We use this to clear the runnable flag before checking dependencies. */ pe_action_requires_any = 0x01000, pe_action_reschedule = 0x02000, pe_action_tracking = 0x04000, pe_action_dedup = 0x08000, //! Internal state tracking when creating graph pe_action_dc = 0x10000, //! Action may run on DC instead of target }; /* *INDENT-ON* */ struct pe_resource_s { char *id; char *clone_name; xmlNode *xml; xmlNode *orig_xml; xmlNode *ops_xml; pe_working_set_t *cluster; pe_resource_t *parent; enum pe_obj_types variant; void *variant_opaque; resource_object_functions_t *fns; resource_alloc_functions_t *cmds; enum rsc_recovery_type recovery_type; // @TODO only pe_restart_restart is of interest, so merge into flags enum pe_restart restart_type; //!< \deprecated will be removed in future release int priority; int stickiness; int sort_index; int failure_timeout; int migration_threshold; guint remote_reconnect_ms; char *pending_task; unsigned long long flags; // @TODO merge these into flags gboolean is_remote_node; gboolean exclusive_discover; //!@{ //! This field should be treated as internal to Pacemaker - GListPtr rsc_cons_lhs; // List of rsc_colocation_t* - GListPtr rsc_cons; // List of rsc_colocation_t* + GListPtr rsc_cons_lhs; // List of pcmk__colocation_t* + GListPtr rsc_cons; // List of pcmk__colocation_t* GListPtr rsc_location; // List of pe__location_t* GListPtr actions; // List of pe_action_t* GListPtr rsc_tickets; // List of rsc_ticket* //!@} pe_node_t *allocated_to; pe_node_t *partial_migration_target; pe_node_t *partial_migration_source; GListPtr running_on; /* pe_node_t* */ GHashTable *known_on; /* pe_node_t* */ GHashTable *allowed_nodes; /* pe_node_t* */ enum rsc_role_e role; enum rsc_role_e next_role; GHashTable *meta; GHashTable *parameters; GHashTable *utilization; GListPtr children; /* pe_resource_t* */ GListPtr dangling_migrations; /* pe_node_t* */ pe_resource_t *container; GListPtr fillers; pe_node_t *pending_node; // Node on which pending_task is happening pe_node_t *lock_node; // Resource is shutdown-locked to this node time_t lock_time; // When shutdown lock started #if ENABLE_VERSIONED_ATTRS xmlNode *versioned_parameters; #endif }; #if ENABLE_VERSIONED_ATTRS // Used as action->action_details if action->rsc is not NULL typedef struct pe_rsc_action_details_s { xmlNode *versioned_parameters; xmlNode *versioned_meta; } pe_rsc_action_details_t; #endif struct pe_action_s { int id; int priority; pe_resource_t *rsc; pe_node_t *node; xmlNode *op_entry; char *task; char *uuid; char *cancel_task; char *reason; enum pe_action_flags flags; enum rsc_start_requirement needs; enum action_fail_response on_fail; enum rsc_role_e fail_role; GHashTable *meta; GHashTable *extra; /* * These two varables are associated with the constraint logic * that involves first having one or more actions runnable before * then allowing this action to execute. * * These varables are used with features such as 'clone-min' which * requires at minimum X number of cloned instances to be running * before an order dependency can run. Another option that uses * this is 'require-all=false' in ordering constrants. This option * says "only require one instance of a resource to start before * allowing dependencies to start" -- basically, require-all=false is * the same as clone-min=1. */ /* current number of known runnable actions in the before list. */ int runnable_before; /* the number of "before" runnable actions required for this action * to be considered runnable */ int required_runnable_before; GListPtr actions_before; /* pe_action_wrapper_t* */ GListPtr actions_after; /* pe_action_wrapper_t* */ /* Some of the above fields could be moved to the details, * except for API backward compatibility. */ void *action_details; // varies by type of action }; typedef struct pe_ticket_s { char *id; gboolean granted; time_t last_granted; gboolean standby; GHashTable *state; } pe_ticket_t; typedef struct pe_tag_s { char *id; GListPtr refs; } pe_tag_t; //! Internal tracking for transition graph creation enum pe_link_state { pe_link_not_dumped, //! Internal tracking for transition graph creation pe_link_dumped, //! Internal tracking for transition graph creation pe_link_dup, //! \deprecated No longer used by Pacemaker }; enum pe_discover_e { pe_discover_always = 0, pe_discover_never, pe_discover_exclusive, }; /* *INDENT-OFF* */ enum pe_ordering { pe_order_none = 0x0, /* deleted */ pe_order_optional = 0x1, /* pure ordering, nothing implied */ pe_order_apply_first_non_migratable = 0x2, /* Only apply this constraint's ordering if first is not migratable. */ pe_order_implies_first = 0x10, /* If 'then' is required, ensure 'first' is too */ pe_order_implies_then = 0x20, /* If 'first' is required, ensure 'then' is too */ pe_order_implies_first_master = 0x40, /* Imply 'first' is required when 'then' is required and then's rsc holds Master role. */ /* first requires then to be both runnable and migrate runnable. */ pe_order_implies_first_migratable = 0x80, pe_order_runnable_left = 0x100, /* 'then' requires 'first' to be runnable */ pe_order_pseudo_left = 0x200, /* 'then' can only be pseudo if 'first' is runnable */ pe_order_implies_then_on_node = 0x400, /* If 'first' is required on 'nodeX', * ensure instances of 'then' on 'nodeX' are too. * Only really useful if 'then' is a clone and 'first' is not */ pe_order_probe = 0x800, /* If 'first->rsc' is * - running but about to stop, ignore the constraint * - otherwise, behave as runnable_left */ pe_order_restart = 0x1000, /* 'then' is runnable if 'first' is optional or runnable */ pe_order_stonith_stop = 0x2000, /* only applies if the action is non-pseudo */ pe_order_serialize_only = 0x4000, /* serialize */ pe_order_same_node = 0x8000, /* applies only if 'first' and 'then' are on same node */ pe_order_implies_first_printed = 0x10000, /* Like ..implies_first but only ensures 'first' is printed, not mandatory */ pe_order_implies_then_printed = 0x20000, /* Like ..implies_then but only ensures 'then' is printed, not mandatory */ pe_order_asymmetrical = 0x100000, /* Indicates asymmetrical one way ordering constraint. */ pe_order_load = 0x200000, /* Only relevant if... */ pe_order_one_or_more = 0x400000, /* 'then' is runnable only if one or more of its dependencies are too */ pe_order_anti_colocation = 0x800000, pe_order_preserve = 0x1000000, /* Hack for breaking user ordering constraints with container resources */ pe_order_then_cancels_first = 0x2000000, // if 'then' becomes required, 'first' becomes optional pe_order_trace = 0x4000000, /* test marker */ }; /* *INDENT-ON* */ typedef struct pe_action_wrapper_s { enum pe_ordering type; enum pe_link_state state; pe_action_t *action; } pe_action_wrapper_t; #ifndef PCMK__NO_COMPAT /* Everything here is deprecated and kept only for public API backward * compatibility. It will be moved to compatibility.h in a future release. */ //!< \deprecated Use pe_action_t instead typedef struct pe_action_s action_t; //!< \deprecated Use pe_action_wrapper_t instead typedef struct pe_action_wrapper_s action_wrapper_t; //!< \deprecated Use pe_node_t instead typedef struct pe_node_s node_t; //!< \deprecated Use enum pe_quorum_policy instead typedef enum pe_quorum_policy no_quorum_policy_t; //!< \deprecated use pe_resource_t instead typedef struct pe_resource_s resource_t; //!< \deprecated Use pe_tag_t instead typedef struct pe_tag_s tag_t; //!< \deprecated Use pe_ticket_t instead typedef struct pe_ticket_s ticket_t; #endif #ifdef __cplusplus } #endif #endif // PE_TYPES__H diff --git a/include/pcmki/pcmki_sched_allocate.h b/include/pcmki/pcmki_sched_allocate.h index a7f8c113cf..d226c8cf01 100644 --- a/include/pcmki/pcmki_sched_allocate.h +++ b/include/pcmki/pcmki_sched_allocate.h @@ -1,182 +1,180 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef SCHED_ALLOCATE__H # define SCHED_ALLOCATE__H # include # include # include # include # include # include struct resource_alloc_functions_s { GHashTable *(*merge_weights) (pe_resource_t *, const char *, GHashTable *, const char *, float, enum pe_weights); pe_node_t *(*allocate) (pe_resource_t *, pe_node_t *, pe_working_set_t *); void (*create_actions) (pe_resource_t *, pe_working_set_t *); gboolean(*create_probe) (pe_resource_t *, pe_node_t *, pe_action_t *, gboolean, pe_working_set_t *); void (*internal_constraints) (pe_resource_t *, pe_working_set_t *); void (*rsc_colocation_lh) (pe_resource_t *, pe_resource_t *, - rsc_colocation_t *, pe_working_set_t *); + pcmk__colocation_t *, pe_working_set_t *); void (*rsc_colocation_rh) (pe_resource_t *, pe_resource_t *, - rsc_colocation_t *, pe_working_set_t *); + pcmk__colocation_t *, pe_working_set_t *); void (*rsc_location) (pe_resource_t *, pe__location_t *); enum pe_action_flags (*action_flags) (pe_action_t *, pe_node_t *); enum pe_graph_flags (*update_actions) (pe_action_t *, pe_action_t *, pe_node_t *, enum pe_action_flags, enum pe_action_flags, enum pe_ordering, pe_working_set_t *data_set); void (*expand) (pe_resource_t *, pe_working_set_t *); void (*append_meta) (pe_resource_t * rsc, xmlNode * xml); }; GHashTable *pcmk__native_merge_weights(pe_resource_t *rsc, const char *rhs, GHashTable *nodes, const char *attr, float factor, uint32_t flags); GHashTable *pcmk__group_merge_weights(pe_resource_t *rsc, const char *rhs, GHashTable *nodes, const char *attr, float factor, uint32_t flags); pe_node_t *pcmk__native_allocate(pe_resource_t *rsc, pe_node_t *preferred, pe_working_set_t *data_set); extern void native_create_actions(pe_resource_t * rsc, pe_working_set_t * data_set); extern void native_internal_constraints(pe_resource_t * rsc, pe_working_set_t * data_set); void native_rsc_colocation_lh(pe_resource_t *lh_rsc, pe_resource_t *rh_rsc, - rsc_colocation_t *constraint, + pcmk__colocation_t *constraint, pe_working_set_t *data_set); void native_rsc_colocation_rh(pe_resource_t *lh_rsc, pe_resource_t *rh_rsc, - rsc_colocation_t *constraint, + pcmk__colocation_t *constraint, pe_working_set_t *data_set); extern void rsc_ticket_constraint(pe_resource_t * lh_rsc, rsc_ticket_t * rsc_ticket, pe_working_set_t * data_set); extern enum pe_action_flags native_action_flags(pe_action_t * action, pe_node_t * node); void native_rsc_location(pe_resource_t *rsc, pe__location_t *constraint); extern void native_expand(pe_resource_t * rsc, pe_working_set_t * data_set); extern gboolean native_create_probe(pe_resource_t * rsc, pe_node_t * node, pe_action_t * complete, gboolean force, pe_working_set_t * data_set); extern void native_append_meta(pe_resource_t * rsc, xmlNode * xml); pe_node_t *pcmk__group_allocate(pe_resource_t *rsc, pe_node_t *preferred, pe_working_set_t *data_set); extern void group_create_actions(pe_resource_t * rsc, pe_working_set_t * data_set); extern void group_internal_constraints(pe_resource_t * rsc, pe_working_set_t * data_set); void group_rsc_colocation_lh(pe_resource_t *lh_rsc, pe_resource_t *rh_rsc, - rsc_colocation_t *constraint, + pcmk__colocation_t *constraint, pe_working_set_t *data_set); void group_rsc_colocation_rh(pe_resource_t *lh_rsc, pe_resource_t *rh_rsc, - rsc_colocation_t *constraint, + pcmk__colocation_t *constraint, pe_working_set_t *data_set); extern enum pe_action_flags group_action_flags(pe_action_t * action, pe_node_t * node); void group_rsc_location(pe_resource_t *rsc, pe__location_t *constraint); extern void group_expand(pe_resource_t * rsc, pe_working_set_t * data_set); extern void group_append_meta(pe_resource_t * rsc, xmlNode * xml); pe_node_t *pcmk__bundle_allocate(pe_resource_t *rsc, pe_node_t *preferred, pe_working_set_t *data_set); void pcmk__bundle_create_actions(pe_resource_t *rsc, pe_working_set_t *data_set); gboolean pcmk__bundle_create_probe(pe_resource_t *rsc, pe_node_t *node, pe_action_t *complete, gboolean force, pe_working_set_t *data_set); void pcmk__bundle_internal_constraints(pe_resource_t *rsc, pe_working_set_t *data_set); void pcmk__bundle_rsc_colocation_lh(pe_resource_t *lh_rsc, pe_resource_t *rh_rsc, - rsc_colocation_t *constraint, + pcmk__colocation_t *constraint, pe_working_set_t *data_set); void pcmk__bundle_rsc_colocation_rh(pe_resource_t *lh_rsc, pe_resource_t *rh_rsc, - rsc_colocation_t *constraint, + pcmk__colocation_t *constraint, pe_working_set_t *data_set); void pcmk__bundle_rsc_location(pe_resource_t *rsc, pe__location_t *constraint); enum pe_action_flags pcmk__bundle_action_flags(pe_action_t *action, pe_node_t *node); void pcmk__bundle_expand(pe_resource_t *rsc, pe_working_set_t *data_set); void pcmk__bundle_append_meta(pe_resource_t *rsc, xmlNode *xml); pe_node_t *pcmk__clone_allocate(pe_resource_t *rsc, pe_node_t *preferred, pe_working_set_t *data_set); extern void clone_create_actions(pe_resource_t * rsc, pe_working_set_t * data_set); extern void clone_internal_constraints(pe_resource_t * rsc, pe_working_set_t * data_set); void clone_rsc_colocation_lh(pe_resource_t *lh_rsc, pe_resource_t *rh_rsc, - rsc_colocation_t *constraint, + pcmk__colocation_t *constraint, pe_working_set_t *data_set); void clone_rsc_colocation_rh(pe_resource_t *lh_rsc, pe_resource_t *rh_rsc, - rsc_colocation_t *constraint, + pcmk__colocation_t *constraint, pe_working_set_t *data_set); void clone_rsc_location(pe_resource_t *rsc, pe__location_t *constraint); extern enum pe_action_flags clone_action_flags(pe_action_t * action, pe_node_t * node); extern void clone_expand(pe_resource_t * rsc, pe_working_set_t * data_set); extern gboolean clone_create_probe(pe_resource_t * rsc, pe_node_t * node, pe_action_t * complete, gboolean force, pe_working_set_t * data_set); extern void clone_append_meta(pe_resource_t * rsc, xmlNode * xml); void apply_master_prefs(pe_resource_t *rsc); pe_node_t *pcmk__set_instance_roles(pe_resource_t *rsc, pe_working_set_t *data_set); void create_promotable_actions(pe_resource_t *rsc, pe_working_set_t *data_set); void promote_demote_constraints(pe_resource_t *rsc, pe_working_set_t *data_set); void promotable_constraints(pe_resource_t *rsc, pe_working_set_t *data_set); void promotable_colocation_rh(pe_resource_t *lh_rsc, pe_resource_t *rh_rsc, - rsc_colocation_t *constraint, + pcmk__colocation_t *constraint, pe_working_set_t *data_set); /* extern resource_object_functions_t resource_variants[]; */ extern resource_alloc_functions_t resource_class_alloc_functions[]; extern gboolean unpack_rsc_order(xmlNode * xml_obj, pe_working_set_t * data_set); -extern gboolean unpack_rsc_colocation(xmlNode * xml_obj, pe_working_set_t * data_set); - extern gboolean unpack_rsc_ticket(xmlNode * xml_obj, pe_working_set_t * data_set); void LogNodeActions(pe_working_set_t * data_set, gboolean terminal); void LogActions(pe_resource_t * rsc, pe_working_set_t * data_set, gboolean terminal); void pcmk__bundle_log_actions(pe_resource_t *rsc, pe_working_set_t *data_set, gboolean terminal); extern void rsc_stonith_ordering(pe_resource_t * rsc, pe_action_t * stonith_op, pe_working_set_t * data_set); enum pe_graph_flags native_update_actions(pe_action_t *first, pe_action_t *then, pe_node_t *node, enum pe_action_flags flags, enum pe_action_flags filter, enum pe_ordering type, pe_working_set_t *data_set); enum pe_graph_flags group_update_actions(pe_action_t *first, pe_action_t *then, pe_node_t *node, enum pe_action_flags flags, enum pe_action_flags filter, enum pe_ordering type, pe_working_set_t *data_set); enum pe_graph_flags pcmk__multi_update_actions(pe_action_t *first, pe_action_t *then, pe_node_t *node, enum pe_action_flags flags, enum pe_action_flags filter, enum pe_ordering type, pe_working_set_t *data_set); gboolean update_action_flags(pe_action_t * action, enum pe_action_flags flags, const char *source, int line); gboolean update_action(pe_action_t *action, pe_working_set_t *data_set); void complex_set_cmds(pe_resource_t * rsc); void pcmk__log_transition_summary(const char *filename); void clone_create_pseudo_actions( pe_resource_t * rsc, GListPtr children, notify_data_t **start_notify, notify_data_t **stop_notify, pe_working_set_t * data_set); #endif diff --git a/include/pcmki/pcmki_sched_utils.h b/include/pcmki/pcmki_sched_utils.h index f2318c5258..c7ae1b894a 100644 --- a/include/pcmki/pcmki_sched_utils.h +++ b/include/pcmki/pcmki_sched_utils.h @@ -1,102 +1,105 @@ /* - * Copyright 2004-2019 the Pacemaker project contributors + * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef PENGINE_AUTILS__H # define PENGINE_AUTILS__H #include // bool #include // GList, GHashTable, gboolean, guint #include // GListPtr #include // lrmd_event_data_t #include // cib_t #include #include #include /* Constraint helper functions */ -extern rsc_colocation_t *invert_constraint(rsc_colocation_t * constraint); +pcmk__colocation_t *invert_constraint(pcmk__colocation_t *constraint); pe__location_t *copy_constraint(pe__location_t *constraint); pe__location_t *rsc2node_new(const char *id, pe_resource_t *rsc, int weight, const char *discovery_mode, pe_node_t *node, pe_working_set_t *data_set); -extern gboolean rsc_colocation_new(const char *id, const char *node_attr, int score, - pe_resource_t * rsc_lh, pe_resource_t * rsc_rh, - const char *state_lh, const char *state_rh, - pe_working_set_t * data_set); +void pcmk__new_colocation(const char *id, const char *node_attr, int score, + pe_resource_t *rsc_lh, pe_resource_t *rsc_rh, + const char *state_lh, const char *state_rh, + pe_working_set_t *data_set); extern gboolean rsc_ticket_new(const char *id, pe_resource_t * rsc_lh, pe_ticket_t * ticket, const char *state_lh, const char *loss_policy, pe_working_set_t * data_set); GHashTable *pcmk__copy_node_table(GHashTable *nodes); GList *pcmk__copy_node_list(const GList *list, bool reset); GList *sort_nodes_by_weight(GList *nodes, pe_node_t *active_node, pe_working_set_t *data_set); extern gboolean can_run_resources(const pe_node_t * node); extern gboolean native_assign_node(pe_resource_t * rsc, GListPtr candidates, pe_node_t * chosen, gboolean force); void native_deallocate(pe_resource_t * rsc); extern void log_action(unsigned int log_level, const char *pre_text, pe_action_t * action, gboolean details); gboolean can_run_any(GHashTable * nodes); pe_resource_t *find_compatible_child(pe_resource_t *local_child, pe_resource_t *rsc, enum rsc_role_e filter, gboolean current, pe_working_set_t *data_set); pe_resource_t *find_compatible_child_by_node(pe_resource_t * local_child, pe_node_t * local_node, pe_resource_t * rsc, enum rsc_role_e filter, gboolean current); gboolean is_child_compatible(pe_resource_t *child_rsc, pe_node_t * local_node, enum rsc_role_e filter, gboolean current); bool assign_node(pe_resource_t * rsc, pe_node_t * node, gboolean force); enum pe_action_flags summary_action_flags(pe_action_t * action, GListPtr children, pe_node_t * node); enum action_tasks clone_child_action(pe_action_t * action); int copies_per_node(pe_resource_t * rsc); enum filter_colocation_res { influence_nothing = 0, influence_rsc_location, influence_rsc_priority, }; extern enum filter_colocation_res filter_colocation_constraint(pe_resource_t * rsc_lh, pe_resource_t * rsc_rh, - rsc_colocation_t * constraint, gboolean preview); + pcmk__colocation_t *constraint, gboolean preview); +bool pcmk__colocation_applies(pe_resource_t *rsc, + pcmk__colocation_t *colocation, + bool promoted_only); extern int compare_capacity(const pe_node_t * node1, const pe_node_t * node2); extern void calculate_utilization(GHashTable * current_utilization, GHashTable * utilization, gboolean plus); extern void process_utilization(pe_resource_t * rsc, pe_node_t ** prefer, pe_working_set_t * data_set); pe_action_t *create_pseudo_resource_op(pe_resource_t * rsc, const char *task, bool optional, bool runnable, pe_working_set_t *data_set); pe_action_t *pe_cancel_op(pe_resource_t *rsc, const char *name, guint interval_ms, pe_node_t *node, pe_working_set_t *data_set); pe_action_t *sched_shutdown_op(pe_node_t *node, pe_working_set_t *data_set); xmlNode *pcmk__create_history_xml(xmlNode *parent, lrmd_event_data_t *event, const char *caller_version, int target_rc, const char *node, const char *origin, int level); # define LOAD_STOPPED "load_stopped" void modify_configuration( pe_working_set_t * data_set, cib_t *cib, const char *quorum, const char *watchdog, GListPtr node_up, GListPtr node_down, GListPtr node_fail, GListPtr op_inject, GListPtr ticket_grant, GListPtr ticket_revoke, GListPtr ticket_standby, GListPtr ticket_activate); int run_simulation(pe_working_set_t * data_set, cib_t *cib, GListPtr op_fail_list, bool quiet); #endif diff --git a/include/pcmki/pcmki_scheduler.h b/include/pcmki/pcmki_scheduler.h index 342d57a2bf..b24e9944d7 100644 --- a/include/pcmki/pcmki_scheduler.h +++ b/include/pcmki/pcmki_scheduler.h @@ -1,110 +1,109 @@ /* * Copyright 2014-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef PENGINE__H # define PENGINE__H -typedef struct rsc_colocation_s rsc_colocation_t; typedef struct rsc_ticket_s rsc_ticket_t; typedef struct lrm_agent_s lrm_agent_t; # include # include # include # include # include # include # include enum pe_stop_fail { pesf_block, pesf_stonith, pesf_ignore }; enum pe_weights { pe_weights_none = 0x0, pe_weights_init = 0x1, pe_weights_forward = 0x4, pe_weights_positive = 0x8, pe_weights_rollback = 0x10, }; -struct rsc_colocation_s { +typedef struct { const char *id; const char *node_attribute; pe_resource_t *rsc_lh; pe_resource_t *rsc_rh; int role_lh; int role_rh; int score; -}; +} pcmk__colocation_t; enum loss_ticket_policy_e { loss_ticket_stop, loss_ticket_demote, loss_ticket_fence, loss_ticket_freeze }; struct rsc_ticket_s { const char *id; pe_resource_t *rsc_lh; pe_ticket_t *ticket; enum loss_ticket_policy_e loss_policy; int role_lh; }; extern gboolean stage0(pe_working_set_t * data_set); extern gboolean probe_resources(pe_working_set_t * data_set); extern gboolean stage2(pe_working_set_t * data_set); extern gboolean stage3(pe_working_set_t * data_set); extern gboolean stage4(pe_working_set_t * data_set); extern gboolean stage5(pe_working_set_t * data_set); extern gboolean stage6(pe_working_set_t * data_set); extern gboolean stage7(pe_working_set_t * data_set); extern gboolean stage8(pe_working_set_t * data_set); extern gboolean summary(GListPtr resources); extern gboolean unpack_constraints(xmlNode * xml_constraints, pe_working_set_t * data_set); extern gboolean shutdown_constraints(pe_node_t * node, pe_action_t * shutdown_op, pe_working_set_t * data_set); void pcmk__order_vs_fence(pe_action_t *stonith_op, pe_working_set_t *data_set); extern int custom_action_order(pe_resource_t * lh_rsc, char *lh_task, pe_action_t * lh_action, pe_resource_t * rh_rsc, char *rh_task, pe_action_t * rh_action, enum pe_ordering type, pe_working_set_t * data_set); extern int new_rsc_order(pe_resource_t * lh_rsc, const char *lh_task, pe_resource_t * rh_rsc, const char *rh_task, enum pe_ordering type, pe_working_set_t * data_set); # define order_start_start(rsc1,rsc2, type) \ new_rsc_order(rsc1, CRMD_ACTION_START, rsc2, CRMD_ACTION_START, type, data_set) # define order_stop_stop(rsc1, rsc2, type) \ new_rsc_order(rsc1, CRMD_ACTION_STOP, rsc2, CRMD_ACTION_STOP, type, data_set) extern void graph_element_from_action(pe_action_t * action, pe_working_set_t * data_set); extern void add_maintenance_update(pe_working_set_t *data_set); xmlNode *pcmk__schedule_actions(pe_working_set_t *data_set, xmlNode *xml_input, crm_time_t *now); bool pcmk__ordering_is_invalid(pe_action_t *action, pe_action_wrapper_t *input); extern gboolean show_scores; extern gboolean show_utilization; extern const char *transition_idle_timeout; #endif diff --git a/lib/pacemaker/pcmk_output.c b/lib/pacemaker/pcmk_output.c index 0d20a54db9..500afd11ee 100644 --- a/lib/pacemaker/pcmk_output.c +++ b/lib/pacemaker/pcmk_output.c @@ -1,567 +1,567 @@ /* * Copyright 2019-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include pcmk__supported_format_t pcmk__out_formats[] = { PCMK__SUPPORTED_FORMAT_XML, { NULL, NULL, NULL } }; int pcmk__out_prologue(pcmk__output_t **out, xmlNodePtr *xml) { int rc = pcmk_rc_ok; if (*xml != NULL) { xmlFreeNode(*xml); } pcmk__register_formats(NULL, pcmk__out_formats); rc = pcmk__output_new(out, "xml", NULL, NULL); if (rc != pcmk_rc_ok) { return rc; } return rc; } void pcmk__out_epilogue(pcmk__output_t *out, xmlNodePtr *xml, int retval) { if (retval == pcmk_rc_ok) { out->finish(out, 0, FALSE, (void **) xml); } pcmk__output_free(out); } static char * -colocations_header(pe_resource_t *rsc, rsc_colocation_t *cons, +colocations_header(pe_resource_t *rsc, pcmk__colocation_t *cons, gboolean dependents) { char *score = NULL; char *retval = NULL; score = score2char(cons->score); if (cons->role_rh > RSC_ROLE_STARTED) { retval = crm_strdup_printf("%s (score=%s, %s role=%s, id=%s)", rsc->id, score, dependents ? "needs" : "with", role2text(cons->role_rh), cons->id); } else { retval = crm_strdup_printf("%s (score=%s, id=%s)", rsc->id, score, cons->id); } free(score); return retval; } static void colocations_xml_node(pcmk__output_t *out, pe_resource_t *rsc, - rsc_colocation_t *cons) { + pcmk__colocation_t *cons) { char *score = NULL; xmlNodePtr node = NULL; score = score2char(cons->score); node = pcmk__output_create_xml_node(out, XML_CONS_TAG_RSC_DEPEND, "id", cons->id, "rsc", cons->rsc_lh->id, "with-rsc", cons->rsc_rh->id, "score", score, NULL); if (cons->node_attribute) { xmlSetProp(node, (pcmkXmlStr) "node-attribute", (pcmkXmlStr) cons->node_attribute); } if (cons->role_lh != RSC_ROLE_UNKNOWN) { xmlSetProp(node, (pcmkXmlStr) "rsc-role", (pcmkXmlStr) role2text(cons->role_lh)); } if (cons->role_rh != RSC_ROLE_UNKNOWN) { xmlSetProp(node, (pcmkXmlStr) "with-rsc-role", (pcmkXmlStr) role2text(cons->role_rh)); } free(score); } static int do_locations_list_xml(pcmk__output_t *out, pe_resource_t *rsc, bool add_header) { GList *lpc = NULL; GList *list = rsc->rsc_location; int rc = pcmk_rc_no_output; for (lpc = list; lpc != NULL; lpc = lpc->next) { pe__location_t *cons = lpc->data; GList *lpc2 = NULL; for (lpc2 = cons->node_list_rh; lpc2 != NULL; lpc2 = lpc2->next) { pe_node_t *node = (pe_node_t *) lpc2->data; char *score = score2char(node->weight); if (add_header) { PCMK__OUTPUT_LIST_HEADER(out, FALSE, rc, "locations"); } pcmk__output_create_xml_node(out, XML_CONS_TAG_RSC_LOCATION, "node", node->details->uname, "rsc", rsc->id, "id", cons->id, "score", score, NULL); free(score); } } if (add_header) { PCMK__OUTPUT_LIST_FOOTER(out, rc); } return rc; } PCMK__OUTPUT_ARGS("rsc-is-colocated-with-list", "pe_resource_t *", "gboolean") static int rsc_is_colocated_with_list(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); gboolean recursive = va_arg(args, gboolean); int rc = pcmk_rc_no_output; if (pcmk_is_set(rsc->flags, pe_rsc_allocating)) { return rc; } pe__set_resource_flags(rsc, pe_rsc_allocating); for (GList *lpc = rsc->rsc_cons; lpc != NULL; lpc = lpc->next) { - rsc_colocation_t *cons = (rsc_colocation_t *) lpc->data; + pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data; char *hdr = NULL; PCMK__OUTPUT_LIST_HEADER(out, FALSE, rc, "Resources %s is colocated with", rsc->id); if (pcmk_is_set(cons->rsc_rh->flags, pe_rsc_allocating)) { out->list_item(out, NULL, "%s (id=%s - loop)", cons->rsc_rh->id, cons->id); continue; } hdr = colocations_header(cons->rsc_rh, cons, FALSE); out->list_item(out, NULL, "%s", hdr); free(hdr); /* Empty list header just for indentation of information about this resource. */ out->begin_list(out, NULL, NULL, NULL); out->message(out, "locations-list", cons->rsc_rh); if (recursive) { out->message(out, "rsc-is-colocated-with-list", cons->rsc_rh, recursive); } out->end_list(out); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("rsc-is-colocated-with-list", "pe_resource_t *", "gboolean") static int rsc_is_colocated_with_list_xml(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); gboolean recursive = va_arg(args, gboolean); int rc = pcmk_rc_no_output; if (pcmk_is_set(rsc->flags, pe_rsc_allocating)) { return rc; } pe__set_resource_flags(rsc, pe_rsc_allocating); for (GList *lpc = rsc->rsc_cons; lpc != NULL; lpc = lpc->next) { - rsc_colocation_t *cons = (rsc_colocation_t *) lpc->data; + pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data; if (pcmk_is_set(cons->rsc_rh->flags, pe_rsc_allocating)) { colocations_xml_node(out, cons->rsc_rh, cons); continue; } colocations_xml_node(out, cons->rsc_rh, cons); do_locations_list_xml(out, cons->rsc_rh, false); if (recursive) { out->message(out, "rsc-is-colocated-with-list", cons->rsc_rh, recursive); } } return rc; } PCMK__OUTPUT_ARGS("rscs-colocated-with-list", "pe_resource_t *", "gboolean") static int rscs_colocated_with_list(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); gboolean recursive = va_arg(args, gboolean); int rc = pcmk_rc_no_output; if (pcmk_is_set(rsc->flags, pe_rsc_allocating)) { return rc; } pe__set_resource_flags(rsc, pe_rsc_allocating); for (GList *lpc = rsc->rsc_cons_lhs; lpc != NULL; lpc = lpc->next) { - rsc_colocation_t *cons = (rsc_colocation_t *) lpc->data; + pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data; char *hdr = NULL; PCMK__OUTPUT_LIST_HEADER(out, FALSE, rc, "Resources colocated with %s", rsc->id); if (pcmk_is_set(cons->rsc_lh->flags, pe_rsc_allocating)) { out->list_item(out, NULL, "%s (id=%s - loop)", cons->rsc_lh->id, cons->id); continue; } hdr = colocations_header(cons->rsc_lh, cons, TRUE); out->list_item(out, NULL, "%s", hdr); free(hdr); /* Empty list header just for indentation of information about this resource. */ out->begin_list(out, NULL, NULL, NULL); out->message(out, "locations-list", cons->rsc_lh); if (recursive) { out->message(out, "rscs-colocated-with-list", cons->rsc_lh, recursive); } out->end_list(out); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("rscs-colocated-with-list", "pe_resource_t *", "gboolean") static int rscs_colocated_with_list_xml(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); gboolean recursive = va_arg(args, gboolean); int rc = pcmk_rc_no_output; if (pcmk_is_set(rsc->flags, pe_rsc_allocating)) { return rc; } pe__set_resource_flags(rsc, pe_rsc_allocating); for (GList *lpc = rsc->rsc_cons_lhs; lpc != NULL; lpc = lpc->next) { - rsc_colocation_t *cons = (rsc_colocation_t *) lpc->data; + pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data; if (pcmk_is_set(cons->rsc_lh->flags, pe_rsc_allocating)) { colocations_xml_node(out, cons->rsc_lh, cons); continue; } colocations_xml_node(out, cons->rsc_lh, cons); do_locations_list_xml(out, cons->rsc_lh, false); if (recursive) { out->message(out, "rscs-colocated-with-list", cons->rsc_lh, recursive); } } return rc; } PCMK__OUTPUT_ARGS("locations-list", "pe_resource_t *") static int locations_list(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); GList *lpc = NULL; GList *list = rsc->rsc_location; int rc = pcmk_rc_no_output; for (lpc = list; lpc != NULL; lpc = lpc->next) { pe__location_t *cons = lpc->data; GList *lpc2 = NULL; for (lpc2 = cons->node_list_rh; lpc2 != NULL; lpc2 = lpc2->next) { pe_node_t *node = (pe_node_t *) lpc2->data; char *score = score2char(node->weight); PCMK__OUTPUT_LIST_HEADER(out, FALSE, rc, "Locations"); out->list_item(out, NULL, "Node %s (score=%s, id=%s, rsc=%s)", node->details->uname, score, cons->id, rsc->id); free(score); } } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("locations-list", "pe_resource_t *") static int locations_list_xml(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); return do_locations_list_xml(out, rsc, true); } PCMK__OUTPUT_ARGS("stacks-constraints", "pe_resource_t *", "pe_working_set_t *", "gboolean") static int stacks_and_constraints(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); gboolean recursive = va_arg(args, gboolean); xmlNodePtr cib_constraints = get_object_root(XML_CIB_TAG_CONSTRAINTS, data_set->input); unpack_constraints(cib_constraints, data_set); // Constraints apply to group/clone, not member/instance rsc = uber_parent(rsc); out->message(out, "locations-list", rsc); pe__clear_resource_flags_on_all(data_set, pe_rsc_allocating); out->message(out, "rscs-colocated-with-list", rsc, recursive); pe__clear_resource_flags_on_all(data_set, pe_rsc_allocating); out->message(out, "rsc-is-colocated-with-list", rsc, recursive); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("stacks-constraints", "pe_resource_t *", "pe_working_set_t *", "gboolean") static int stacks_and_constraints_xml(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); gboolean recursive = va_arg(args, gboolean); xmlNodePtr cib_constraints = get_object_root(XML_CIB_TAG_CONSTRAINTS, data_set->input); unpack_constraints(cib_constraints, data_set); // Constraints apply to group/clone, not member/instance rsc = uber_parent(rsc); pcmk__output_xml_create_parent(out, "constraints", NULL); do_locations_list_xml(out, rsc, false); pe__clear_resource_flags_on_all(data_set, pe_rsc_allocating); out->message(out, "rscs-colocated-with-list", rsc, recursive); pe__clear_resource_flags_on_all(data_set, pe_rsc_allocating); out->message(out, "rsc-is-colocated-with-list", rsc, recursive); pcmk__output_xml_pop_parent(out); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("health", "const char *", "const char *", "const char *", "const char *") static int health_text(pcmk__output_t *out, va_list args) { const char *sys_from = va_arg(args, const char *); const char *host_from = va_arg(args, const char *); const char *fsa_state = va_arg(args, const char *); const char *result = va_arg(args, const char *); if (!out->is_quiet(out)) { out->info(out, "Status of %s@%s: %s (%s)", crm_str(sys_from), crm_str(host_from), crm_str(fsa_state), crm_str(result)); } else if (fsa_state != NULL) { out->info(out, "%s", fsa_state); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("health", "const char *", "const char *", "const char *", "const char *") static int health_xml(pcmk__output_t *out, va_list args) { const char *sys_from = va_arg(args, const char *); const char *host_from = va_arg(args, const char *); const char *fsa_state = va_arg(args, const char *); const char *result = va_arg(args, const char *); pcmk__output_create_xml_node(out, crm_str(sys_from), "node_name", crm_str(host_from), "state", crm_str(fsa_state), "result", crm_str(result), NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("pacemakerd-health", "const char *", "const char *", "const char *") static int pacemakerd_health_text(pcmk__output_t *out, va_list args) { const char *sys_from = va_arg(args, const char *); const char *state = va_arg(args, const char *); const char *last_updated = va_arg(args, const char *); if (!out->is_quiet(out)) { out->info(out, "Status of %s: '%s' %s %s", crm_str(sys_from), crm_str(state), (!pcmk__str_empty(last_updated))? "last updated":"", crm_str(last_updated)); } else { out->info(out, "%s", crm_str(state)); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("pacemakerd-health", "const char *", "const char *", "const char *") static int pacemakerd_health_xml(pcmk__output_t *out, va_list args) { const char *sys_from = va_arg(args, const char *); const char *state = va_arg(args, const char *); const char *last_updated = va_arg(args, const char *); pcmk__output_create_xml_node(out, crm_str(sys_from), "state", crm_str(state), "last_updated", crm_str(last_updated), NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("dc", "const char *") static int dc_text(pcmk__output_t *out, va_list args) { const char *dc = va_arg(args, const char *); if (!out->is_quiet(out)) { out->info(out, "Designated Controller is: %s", crm_str(dc)); } else if (dc != NULL) { out->info(out, "%s", dc); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("dc", "const char *") static int dc_xml(pcmk__output_t *out, va_list args) { const char *dc = va_arg(args, const char *); pcmk__output_create_xml_node(out, "dc", "node_name", crm_str(dc), NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("crmadmin-node-list", "xmlNodePtr", "gboolean") static int crmadmin_node_list(pcmk__output_t *out, va_list args) { xmlNodePtr xml_node = va_arg(args, xmlNodePtr); gboolean BASH_EXPORT = va_arg(args, gboolean); int found = 0; xmlNode *node = NULL; xmlNode *nodes = get_object_root(XML_CIB_TAG_NODES, xml_node); out->begin_list(out, NULL, NULL, "nodes"); for (node = first_named_child(nodes, XML_CIB_TAG_NODE); node != NULL; node = crm_next_same_xml(node)) { const char *node_type = BASH_EXPORT ? NULL : crm_element_value(node, XML_ATTR_TYPE); out->message(out, "crmadmin-node", node_type, crm_str(crm_element_value(node, XML_ATTR_UNAME)), crm_str(crm_element_value(node, XML_ATTR_ID)), BASH_EXPORT); found++; } // @TODO List Pacemaker Remote nodes that don't have a entry out->end_list(out); if (found == 0) { out->info(out, "No nodes configured"); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("crmadmin-node", "const char *", "const char *", "const char *", "gboolean") static int crmadmin_node_text(pcmk__output_t *out, va_list args) { const char *type = va_arg(args, const char *); const char *name = va_arg(args, const char *); const char *id = va_arg(args, const char *); gboolean BASH_EXPORT = va_arg(args, gboolean); if (BASH_EXPORT) { out->info(out, "export %s=%s", crm_str(name), crm_str(id)); } else { out->info(out, "%s node: %s (%s)", type ? type : "member", crm_str(name), crm_str(id)); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("crmadmin-node", "const char *", "const char *", "const char *", "gboolean") static int crmadmin_node_xml(pcmk__output_t *out, va_list args) { const char *type = va_arg(args, const char *); const char *name = va_arg(args, const char *); const char *id = va_arg(args, const char *); pcmk__output_create_xml_node(out, "node", "type", type ? type : "member", "name", crm_str(name), "id", crm_str(id), NULL); return pcmk_rc_ok; } static pcmk__message_entry_t fmt_functions[] = { { "rsc-is-colocated-with-list", "default", rsc_is_colocated_with_list }, { "rsc-is-colocated-with-list", "xml", rsc_is_colocated_with_list_xml }, { "rscs-colocated-with-list", "default", rscs_colocated_with_list }, { "rscs-colocated-with-list", "xml", rscs_colocated_with_list_xml }, { "locations-list", "default", locations_list }, { "locations-list", "xml", locations_list_xml }, { "stacks-constraints", "default", stacks_and_constraints }, { "stacks-constraints", "xml", stacks_and_constraints_xml }, { "health", "default", health_text }, { "health", "xml", health_xml }, { "pacemakerd-health", "default", pacemakerd_health_text }, { "pacemakerd-health", "xml", pacemakerd_health_xml }, { "dc", "default", dc_text }, { "dc", "xml", dc_xml }, { "crmadmin-node-list", "default", crmadmin_node_list }, { "crmadmin-node", "default", crmadmin_node_text }, { "crmadmin-node", "xml", crmadmin_node_xml }, { NULL, NULL, NULL } }; void pcmk__register_lib_messages(pcmk__output_t *out) { pcmk__register_messages(out, fmt_functions); } diff --git a/lib/pacemaker/pcmk_sched_bundle.c b/lib/pacemaker/pcmk_sched_bundle.c index e9b8a74811..ac9219c7bc 100644 --- a/lib/pacemaker/pcmk_sched_bundle.c +++ b/lib/pacemaker/pcmk_sched_bundle.c @@ -1,1084 +1,1084 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #define PE__VARIANT_BUNDLE 1 #include static bool is_bundle_node(pe__bundle_variant_data_t *data, pe_node_t *node) { for (GList *gIter = data->replicas; gIter != NULL; gIter = gIter->next) { pe__bundle_replica_t *replica = gIter->data; if (node->details == replica->node->details) { return TRUE; } } return FALSE; } gint sort_clone_instance(gconstpointer a, gconstpointer b, gpointer data_set); void distribute_children(pe_resource_t *rsc, GListPtr children, GListPtr nodes, int max, int per_host_max, pe_working_set_t * data_set); static GList * get_container_list(pe_resource_t *rsc) { GList *containers = NULL; if (rsc->variant == pe_container) { pe__bundle_variant_data_t *data = NULL; get_bundle_variant_data(data, rsc); for (GList *gIter = data->replicas; gIter != NULL; gIter = gIter->next) { pe__bundle_replica_t *replica = gIter->data; containers = g_list_append(containers, replica->container); } } return containers; } static inline GList * get_containers_or_children(pe_resource_t *rsc) { return (rsc->variant == pe_container)? get_container_list(rsc) : rsc->children; } static bool migration_threshold_reached(pe_resource_t *rsc, pe_node_t *node, pe_working_set_t *data_set) { int fail_count, countdown; /* Migration threshold of 0 means never force away */ if (rsc->migration_threshold == 0) { return FALSE; } // If we're ignoring failures, also ignore the migration threshold if (pcmk_is_set(rsc->flags, pe_rsc_failure_ignored)) { return FALSE; } /* If there are no failures, there's no need to force away */ fail_count = pe_get_failcount(node, rsc, NULL, pe_fc_effective|pe_fc_fillers, NULL, data_set); if (fail_count <= 0) { return FALSE; } /* How many more times recovery will be tried on this node */ countdown = QB_MAX(rsc->migration_threshold - fail_count, 0); if (countdown == 0) { crm_warn("Forcing %s away from %s after %d failures (max=%d)", rsc->id, node->details->uname, fail_count, rsc->migration_threshold); return TRUE; } crm_info("%s can fail %d more times on %s before being forced off", rsc->id, countdown, node->details->uname); return FALSE; } pe_node_t * pcmk__bundle_allocate(pe_resource_t *rsc, pe_node_t *prefer, pe_working_set_t *data_set) { GListPtr containers = NULL; GListPtr nodes = NULL; pe__bundle_variant_data_t *bundle_data = NULL; CRM_CHECK(rsc != NULL, return NULL); get_bundle_variant_data(bundle_data, rsc); pe__set_resource_flags(rsc, pe_rsc_allocating); containers = get_container_list(rsc); pe__show_node_weights(!show_scores, rsc, __func__, rsc->allowed_nodes); nodes = g_hash_table_get_values(rsc->allowed_nodes); nodes = sort_nodes_by_weight(nodes, NULL, data_set); containers = g_list_sort_with_data(containers, sort_clone_instance, data_set); distribute_children(rsc, containers, nodes, bundle_data->nreplicas, bundle_data->nreplicas_per_host, data_set); g_list_free(nodes); g_list_free(containers); for (GList *gIter = bundle_data->replicas; gIter != NULL; gIter = gIter->next) { pe__bundle_replica_t *replica = gIter->data; pe_node_t *container_host = NULL; CRM_ASSERT(replica); if (replica->ip) { pe_rsc_trace(rsc, "Allocating bundle %s IP %s", rsc->id, replica->ip->id); replica->ip->cmds->allocate(replica->ip, prefer, data_set); } container_host = replica->container->allocated_to; if (replica->remote && pe__is_guest_or_remote_node(container_host)) { /* We need 'nested' connection resources to be on the same * host because pacemaker-remoted only supports a single * active connection */ - rsc_colocation_new("child-remote-with-docker-remote", NULL, - INFINITY, replica->remote, - container_host->details->remote_rsc, NULL, NULL, - data_set); + pcmk__new_colocation("child-remote-with-docker-remote", NULL, + INFINITY, replica->remote, + container_host->details->remote_rsc, NULL, + NULL, data_set); } if (replica->remote) { pe_rsc_trace(rsc, "Allocating bundle %s connection %s", rsc->id, replica->remote->id); replica->remote->cmds->allocate(replica->remote, prefer, data_set); } // Explicitly allocate replicas' children before bundle child if (replica->child) { pe_node_t *node = NULL; GHashTableIter iter; g_hash_table_iter_init(&iter, replica->child->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) & node)) { if (node->details != replica->node->details) { node->weight = -INFINITY; } else if (!migration_threshold_reached(replica->child, node, data_set)) { node->weight = INFINITY; } } pe__set_resource_flags(replica->child->parent, pe_rsc_allocating); pe_rsc_trace(rsc, "Allocating bundle %s replica child %s", rsc->id, replica->child->id); replica->child->cmds->allocate(replica->child, replica->node, data_set); pe__clear_resource_flags(replica->child->parent, pe_rsc_allocating); } } if (bundle_data->child) { pe_node_t *node = NULL; GHashTableIter iter; g_hash_table_iter_init(&iter, bundle_data->child->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) & node)) { if (is_bundle_node(bundle_data, node)) { node->weight = 0; } else { node->weight = -INFINITY; } } pe_rsc_trace(rsc, "Allocating bundle %s child %s", rsc->id, bundle_data->child->id); bundle_data->child->cmds->allocate(bundle_data->child, prefer, data_set); } pe__clear_resource_flags(rsc, pe_rsc_allocating|pe_rsc_provisional); return NULL; } void pcmk__bundle_create_actions(pe_resource_t *rsc, pe_working_set_t *data_set) { pe_action_t *action = NULL; GListPtr containers = NULL; pe__bundle_variant_data_t *bundle_data = NULL; CRM_CHECK(rsc != NULL, return); containers = get_container_list(rsc); get_bundle_variant_data(bundle_data, rsc); for (GList *gIter = bundle_data->replicas; gIter != NULL; gIter = gIter->next) { pe__bundle_replica_t *replica = gIter->data; CRM_ASSERT(replica); if (replica->ip) { replica->ip->cmds->create_actions(replica->ip, data_set); } if (replica->container) { replica->container->cmds->create_actions(replica->container, data_set); } if (replica->remote) { replica->remote->cmds->create_actions(replica->remote, data_set); } } clone_create_pseudo_actions(rsc, containers, NULL, NULL, data_set); if (bundle_data->child) { bundle_data->child->cmds->create_actions(bundle_data->child, data_set); if (pcmk_is_set(bundle_data->child->flags, pe_rsc_promotable)) { /* promote */ create_pseudo_resource_op(rsc, RSC_PROMOTE, TRUE, TRUE, data_set); action = create_pseudo_resource_op(rsc, RSC_PROMOTED, TRUE, TRUE, data_set); action->priority = INFINITY; /* demote */ create_pseudo_resource_op(rsc, RSC_DEMOTE, TRUE, TRUE, data_set); action = create_pseudo_resource_op(rsc, RSC_DEMOTED, TRUE, TRUE, data_set); action->priority = INFINITY; } } g_list_free(containers); } void pcmk__bundle_internal_constraints(pe_resource_t *rsc, pe_working_set_t *data_set) { pe__bundle_variant_data_t *bundle_data = NULL; CRM_CHECK(rsc != NULL, return); get_bundle_variant_data(bundle_data, rsc); if (bundle_data->child) { new_rsc_order(rsc, RSC_START, bundle_data->child, RSC_START, pe_order_implies_first_printed, data_set); new_rsc_order(rsc, RSC_STOP, bundle_data->child, RSC_STOP, pe_order_implies_first_printed, data_set); if (bundle_data->child->children) { new_rsc_order(bundle_data->child, RSC_STARTED, rsc, RSC_STARTED, pe_order_implies_then_printed, data_set); new_rsc_order(bundle_data->child, RSC_STOPPED, rsc, RSC_STOPPED, pe_order_implies_then_printed, data_set); } else { new_rsc_order(bundle_data->child, RSC_START, rsc, RSC_STARTED, pe_order_implies_then_printed, data_set); new_rsc_order(bundle_data->child, RSC_STOP, rsc, RSC_STOPPED, pe_order_implies_then_printed, data_set); } } for (GList *gIter = bundle_data->replicas; gIter != NULL; gIter = gIter->next) { pe__bundle_replica_t *replica = gIter->data; CRM_ASSERT(replica); CRM_ASSERT(replica->container); replica->container->cmds->internal_constraints(replica->container, data_set); order_start_start(rsc, replica->container, pe_order_runnable_left|pe_order_implies_first_printed); if (replica->child) { order_stop_stop(rsc, replica->child, pe_order_implies_first_printed); } order_stop_stop(rsc, replica->container, pe_order_implies_first_printed); new_rsc_order(replica->container, RSC_START, rsc, RSC_STARTED, pe_order_implies_then_printed, data_set); new_rsc_order(replica->container, RSC_STOP, rsc, RSC_STOPPED, pe_order_implies_then_printed, data_set); if (replica->ip) { replica->ip->cmds->internal_constraints(replica->ip, data_set); // Start ip then container new_rsc_order(replica->ip, RSC_START, replica->container, RSC_START, pe_order_runnable_left|pe_order_preserve, data_set); new_rsc_order(replica->container, RSC_STOP, replica->ip, RSC_STOP, pe_order_implies_first|pe_order_preserve, data_set); - rsc_colocation_new("ip-with-docker", NULL, INFINITY, replica->ip, - replica->container, NULL, NULL, data_set); + pcmk__new_colocation("ip-with-docker", NULL, INFINITY, replica->ip, + replica->container, NULL, NULL, data_set); } if (replica->remote) { /* This handles ordering and colocating remote relative to container * (via "resource-with-container"). Since IP is also ordered and * colocated relative to the container, we don't need to do anything * explicit here with IP. */ replica->remote->cmds->internal_constraints(replica->remote, data_set); } if (replica->child) { CRM_ASSERT(replica->remote); // "Start remote then child" is implicit in scheduler's remote logic } } if (bundle_data->child) { bundle_data->child->cmds->internal_constraints(bundle_data->child, data_set); if (pcmk_is_set(bundle_data->child->flags, pe_rsc_promotable)) { promote_demote_constraints(rsc, data_set); /* child demoted before global demoted */ new_rsc_order(bundle_data->child, RSC_DEMOTED, rsc, RSC_DEMOTED, pe_order_implies_then_printed, data_set); /* global demote before child demote */ new_rsc_order(rsc, RSC_DEMOTE, bundle_data->child, RSC_DEMOTE, pe_order_implies_first_printed, data_set); /* child promoted before global promoted */ new_rsc_order(bundle_data->child, RSC_PROMOTED, rsc, RSC_PROMOTED, pe_order_implies_then_printed, data_set); /* global promote before child promote */ new_rsc_order(rsc, RSC_PROMOTE, bundle_data->child, RSC_PROMOTE, pe_order_implies_first_printed, data_set); } } else { // int type = pe_order_optional | pe_order_implies_then | pe_order_restart; // custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_STOP, 0), NULL, // rsc, pcmk__op_key(rsc->id, RSC_START, 0), NULL, pe_order_optional, data_set); } } static pe_resource_t * compatible_replica_for_node(pe_resource_t *rsc_lh, pe_node_t *candidate, pe_resource_t *rsc, enum rsc_role_e filter, gboolean current) { pe__bundle_variant_data_t *bundle_data = NULL; CRM_CHECK(candidate != NULL, return NULL); get_bundle_variant_data(bundle_data, rsc); crm_trace("Looking for compatible child from %s for %s on %s", rsc_lh->id, rsc->id, candidate->details->uname); for (GList *gIter = bundle_data->replicas; gIter != NULL; gIter = gIter->next) { pe__bundle_replica_t *replica = gIter->data; if (is_child_compatible(replica->container, candidate, filter, current)) { crm_trace("Pairing %s with %s on %s", rsc_lh->id, replica->container->id, candidate->details->uname); return replica->container; } } crm_trace("Can't pair %s with %s", rsc_lh->id, rsc->id); return NULL; } static pe_resource_t * compatible_replica(pe_resource_t *rsc_lh, pe_resource_t *rsc, enum rsc_role_e filter, gboolean current, pe_working_set_t *data_set) { GListPtr scratch = NULL; pe_resource_t *pair = NULL; pe_node_t *active_node_lh = NULL; active_node_lh = rsc_lh->fns->location(rsc_lh, NULL, current); if (active_node_lh) { return compatible_replica_for_node(rsc_lh, active_node_lh, rsc, filter, current); } scratch = g_hash_table_get_values(rsc_lh->allowed_nodes); scratch = sort_nodes_by_weight(scratch, NULL, data_set); for (GListPtr gIter = scratch; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; pair = compatible_replica_for_node(rsc_lh, node, rsc, filter, current); if (pair) { goto done; } } pe_rsc_debug(rsc, "Can't pair %s with %s", rsc_lh->id, (rsc? rsc->id : "none")); done: g_list_free(scratch); return pair; } void pcmk__bundle_rsc_colocation_lh(pe_resource_t *rsc, pe_resource_t *rsc_rh, - rsc_colocation_t *constraint, + pcmk__colocation_t *constraint, pe_working_set_t *data_set) { /* -- Never called -- * * Instead we add the colocation constraints to the child and call from there */ CRM_ASSERT(FALSE); } int copies_per_node(pe_resource_t * rsc) { /* Strictly speaking, there should be a 'copies_per_node' addition * to the resource function table and each case would be a * function. However that would be serious overkill to return an * int. In fact, it seems to me that both function tables * could/should be replaced by resources.{c,h} full of * rsc_{some_operation} functions containing a switch as below * which calls out to functions named {variant}_{some_operation} * as needed. */ switch(rsc->variant) { case pe_unknown: return 0; case pe_native: case pe_group: return 1; case pe_clone: { const char *max_clones_node = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_INCARNATION_NODEMAX); return crm_parse_int(max_clones_node, "1"); } case pe_container: { pe__bundle_variant_data_t *data = NULL; get_bundle_variant_data(data, rsc); return data->nreplicas_per_host; } } return 0; } void pcmk__bundle_rsc_colocation_rh(pe_resource_t *rsc_lh, pe_resource_t *rsc, - rsc_colocation_t *constraint, + pcmk__colocation_t *constraint, pe_working_set_t *data_set) { GListPtr allocated_rhs = NULL; pe__bundle_variant_data_t *bundle_data = NULL; CRM_CHECK(constraint != NULL, return); CRM_CHECK(rsc_lh != NULL, pe_err("rsc_lh was NULL for %s", constraint->id); return); CRM_CHECK(rsc != NULL, pe_err("rsc was NULL for %s", constraint->id); return); CRM_ASSERT(rsc_lh->variant == pe_native); if (constraint->score == 0) { return; } if (pcmk_is_set(rsc->flags, pe_rsc_provisional)) { pe_rsc_trace(rsc, "%s is still provisional", rsc->id); return; } else if(constraint->rsc_lh->variant > pe_group) { pe_resource_t *rh_child = compatible_replica(rsc_lh, rsc, RSC_ROLE_UNKNOWN, FALSE, data_set); if (rh_child) { pe_rsc_debug(rsc, "Pairing %s with %s", rsc_lh->id, rh_child->id); rsc_lh->cmds->rsc_colocation_lh(rsc_lh, rh_child, constraint, data_set); } else if (constraint->score >= INFINITY) { crm_notice("Cannot pair %s with instance of %s", rsc_lh->id, rsc->id); assign_node(rsc_lh, NULL, TRUE); } else { pe_rsc_debug(rsc, "Cannot pair %s with instance of %s", rsc_lh->id, rsc->id); } return; } get_bundle_variant_data(bundle_data, rsc); pe_rsc_trace(rsc, "Processing constraint %s: %s -> %s %d", constraint->id, rsc_lh->id, rsc->id, constraint->score); for (GList *gIter = bundle_data->replicas; gIter != NULL; gIter = gIter->next) { pe__bundle_replica_t *replica = gIter->data; if (constraint->score < INFINITY) { replica->container->cmds->rsc_colocation_rh(rsc_lh, replica->container, constraint, data_set); } else { pe_node_t *chosen = replica->container->fns->location(replica->container, NULL, FALSE); if ((chosen == NULL) || is_set_recursive(replica->container, pe_rsc_block, TRUE)) { continue; } if ((constraint->role_rh >= RSC_ROLE_MASTER) && (replica->child == NULL)) { continue; } if ((constraint->role_rh >= RSC_ROLE_MASTER) && (replica->child->next_role < RSC_ROLE_MASTER)) { continue; } pe_rsc_trace(rsc, "Allowing %s: %s %d", constraint->id, chosen->details->uname, chosen->weight); allocated_rhs = g_list_prepend(allocated_rhs, chosen); } } if (constraint->score >= INFINITY) { node_list_exclude(rsc_lh->allowed_nodes, allocated_rhs, FALSE); } g_list_free(allocated_rhs); } enum pe_action_flags pcmk__bundle_action_flags(pe_action_t *action, pe_node_t *node) { GListPtr containers = NULL; enum pe_action_flags flags = 0; pe__bundle_variant_data_t *data = NULL; get_bundle_variant_data(data, action->rsc); if(data->child) { enum action_tasks task = get_complex_task(data->child, action->task, TRUE); switch(task) { case no_action: case action_notify: case action_notified: case action_promote: case action_promoted: case action_demote: case action_demoted: return summary_action_flags(action, data->child->children, node); default: break; } } containers = get_container_list(action->rsc); flags = summary_action_flags(action, containers, node); g_list_free(containers); return flags; } pe_resource_t * find_compatible_child_by_node(pe_resource_t * local_child, pe_node_t * local_node, pe_resource_t * rsc, enum rsc_role_e filter, gboolean current) { GListPtr gIter = NULL; GListPtr children = NULL; if (local_node == NULL) { crm_err("Can't colocate unrunnable child %s with %s", local_child->id, rsc->id); return NULL; } crm_trace("Looking for compatible child from %s for %s on %s", local_child->id, rsc->id, local_node->details->uname); children = get_containers_or_children(rsc); for (gIter = children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; if(is_child_compatible(child_rsc, local_node, filter, current)) { crm_trace("Pairing %s with %s on %s", local_child->id, child_rsc->id, local_node->details->uname); return child_rsc; } } crm_trace("Can't pair %s with %s", local_child->id, rsc->id); if(children != rsc->children) { g_list_free(children); } return NULL; } static pe__bundle_replica_t * replica_for_container(pe_resource_t *rsc, pe_resource_t *container, pe_node_t *node) { if (rsc->variant == pe_container) { pe__bundle_variant_data_t *data = NULL; get_bundle_variant_data(data, rsc); for (GList *gIter = data->replicas; gIter != NULL; gIter = gIter->next) { pe__bundle_replica_t *replica = gIter->data; if (replica->child && (container == replica->container) && (node->details == replica->node->details)) { return replica; } } } return NULL; } static enum pe_graph_flags multi_update_interleave_actions(pe_action_t *first, pe_action_t *then, pe_node_t *node, enum pe_action_flags flags, enum pe_action_flags filter, enum pe_ordering type, pe_working_set_t *data_set) { GListPtr gIter = NULL; GListPtr children = NULL; gboolean current = FALSE; enum pe_graph_flags changed = pe_graph_none; /* Fix this - lazy */ if (pcmk__ends_with(first->uuid, "_stopped_0") || pcmk__ends_with(first->uuid, "_demoted_0")) { current = TRUE; } children = get_containers_or_children(then->rsc); for (gIter = children; gIter != NULL; gIter = gIter->next) { pe_resource_t *then_child = gIter->data; pe_resource_t *first_child = find_compatible_child(then_child, first->rsc, RSC_ROLE_UNKNOWN, current, data_set); if (first_child == NULL && current) { crm_trace("Ignore"); } else if (first_child == NULL) { crm_debug("No match found for %s (%d / %s / %s)", then_child->id, current, first->uuid, then->uuid); /* Me no like this hack - but what else can we do? * * If there is no-one active or about to be active * on the same node as then_child, then they must * not be allowed to start */ if (type & (pe_order_runnable_left | pe_order_implies_then) /* Mandatory */ ) { pe_rsc_info(then->rsc, "Inhibiting %s from being active", then_child->id); if(assign_node(then_child, NULL, TRUE)) { pe__set_graph_flags(changed, first, pe_graph_updated_then); } } } else { pe_action_t *first_action = NULL; pe_action_t *then_action = NULL; enum action_tasks task = clone_child_action(first); const char *first_task = task2text(task); pe__bundle_replica_t *first_replica = NULL; pe__bundle_replica_t *then_replica = NULL; first_replica = replica_for_container(first->rsc, first_child, node); if (strstr(first->task, "stop") && first_replica && first_replica->child) { /* Except for 'stopped' we should be looking at the * in-container resource, actions for the child will * happen later and are therefor more likely to align * with the user's intent. */ first_action = find_first_action(first_replica->child->actions, NULL, task2text(task), node); } else { first_action = find_first_action(first_child->actions, NULL, task2text(task), node); } then_replica = replica_for_container(then->rsc, then_child, node); if (strstr(then->task, "mote") && then_replica && then_replica->child) { /* Promote/demote actions will never be found for the * container resource, look in the child instead * * Alternatively treat: * 'XXXX then promote YYYY' as 'XXXX then start container for YYYY', and * 'demote XXXX then stop YYYY' as 'stop container for XXXX then stop YYYY' */ then_action = find_first_action(then_replica->child->actions, NULL, then->task, node); } else { then_action = find_first_action(then_child->actions, NULL, then->task, node); } if (first_action == NULL) { if (!pcmk_is_set(first_child->flags, pe_rsc_orphan) && !pcmk__str_any_of(first_task, RSC_STOP, RSC_DEMOTE, NULL)) { crm_err("Internal error: No action found for %s in %s (first)", first_task, first_child->id); } else { crm_trace("No action found for %s in %s%s (first)", first_task, first_child->id, pcmk_is_set(first_child->flags, pe_rsc_orphan)? " (ORPHAN)" : ""); } continue; } /* We're only interested if 'then' is neither stopping nor being demoted */ if (then_action == NULL) { if (!pcmk_is_set(then_child->flags, pe_rsc_orphan) && !pcmk__str_any_of(then->task, RSC_STOP, RSC_DEMOTE, NULL)) { crm_err("Internal error: No action found for %s in %s (then)", then->task, then_child->id); } else { crm_trace("No action found for %s in %s%s (then)", then->task, then_child->id, pcmk_is_set(then_child->flags, pe_rsc_orphan)? " (ORPHAN)" : ""); } continue; } if (order_actions(first_action, then_action, type)) { crm_debug("Created constraint for %s (%d) -> %s (%d) %.6x", first_action->uuid, pcmk_is_set(first_action->flags, pe_action_optional), then_action->uuid, pcmk_is_set(then_action->flags, pe_action_optional), type); pe__set_graph_flags(changed, first, pe_graph_updated_first|pe_graph_updated_then); } if(first_action && then_action) { changed |= then_child->cmds->update_actions(first_action, then_action, node, first_child->cmds->action_flags(first_action, node), filter, type, data_set); } else { crm_err("Nothing found either for %s (%p) or %s (%p) %s", first_child->id, first_action, then_child->id, then_action, task2text(task)); } } } if(children != then->rsc->children) { g_list_free(children); } return changed; } static bool can_interleave_actions(pe_action_t *first, pe_action_t *then) { bool interleave = FALSE; pe_resource_t *rsc = NULL; const char *interleave_s = NULL; if(first->rsc == NULL || then->rsc == NULL) { crm_trace("Not interleaving %s with %s (both must be resources)", first->uuid, then->uuid); return FALSE; } else if(first->rsc == then->rsc) { crm_trace("Not interleaving %s with %s (must belong to different resources)", first->uuid, then->uuid); return FALSE; } else if(first->rsc->variant < pe_clone || then->rsc->variant < pe_clone) { crm_trace("Not interleaving %s with %s (both sides must be clones or bundles)", first->uuid, then->uuid); return FALSE; } if (pcmk__ends_with(then->uuid, "_stop_0") || pcmk__ends_with(then->uuid, "_demote_0")) { rsc = first->rsc; } else { rsc = then->rsc; } interleave_s = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_INTERLEAVE); interleave = crm_is_true(interleave_s); crm_trace("Interleave %s -> %s: %s (based on %s)", first->uuid, then->uuid, interleave ? "yes" : "no", rsc->id); return interleave; } enum pe_graph_flags pcmk__multi_update_actions(pe_action_t *first, pe_action_t *then, pe_node_t *node, enum pe_action_flags flags, enum pe_action_flags filter, enum pe_ordering type, pe_working_set_t *data_set) { enum pe_graph_flags changed = pe_graph_none; crm_trace("%s -> %s", first->uuid, then->uuid); if(can_interleave_actions(first, then)) { changed = multi_update_interleave_actions(first, then, node, flags, filter, type, data_set); } else if(then->rsc) { GListPtr gIter = NULL; GListPtr children = NULL; // Handle the 'primitive' ordering case changed |= native_update_actions(first, then, node, flags, filter, type, data_set); // Now any children (or containers in the case of a bundle) children = get_containers_or_children(then->rsc); for (gIter = children; gIter != NULL; gIter = gIter->next) { pe_resource_t *then_child = (pe_resource_t *) gIter->data; enum pe_graph_flags then_child_changed = pe_graph_none; pe_action_t *then_child_action = find_first_action(then_child->actions, NULL, then->task, node); if (then_child_action) { enum pe_action_flags then_child_flags = then_child->cmds->action_flags(then_child_action, node); if (pcmk_is_set(then_child_flags, pe_action_runnable)) { then_child_changed |= then_child->cmds->update_actions(first, then_child_action, node, flags, filter, type, data_set); } changed |= then_child_changed; if (then_child_changed & pe_graph_updated_then) { for (GListPtr lpc = then_child_action->actions_after; lpc != NULL; lpc = lpc->next) { pe_action_wrapper_t *next = (pe_action_wrapper_t *) lpc->data; update_action(next->action, data_set); } } } } if(children != then->rsc->children) { g_list_free(children); } } return changed; } void pcmk__bundle_rsc_location(pe_resource_t *rsc, pe__location_t *constraint) { pe__bundle_variant_data_t *bundle_data = NULL; get_bundle_variant_data(bundle_data, rsc); native_rsc_location(rsc, constraint); for (GList *gIter = bundle_data->replicas; gIter != NULL; gIter = gIter->next) { pe__bundle_replica_t *replica = gIter->data; if (replica->container) { replica->container->cmds->rsc_location(replica->container, constraint); } if (replica->ip) { replica->ip->cmds->rsc_location(replica->ip, constraint); } } if (bundle_data->child && ((constraint->role_filter == RSC_ROLE_SLAVE) || (constraint->role_filter == RSC_ROLE_MASTER))) { bundle_data->child->cmds->rsc_location(bundle_data->child, constraint); bundle_data->child->rsc_location = g_list_prepend(bundle_data->child->rsc_location, constraint); } } void pcmk__bundle_expand(pe_resource_t *rsc, pe_working_set_t * data_set) { pe__bundle_variant_data_t *bundle_data = NULL; CRM_CHECK(rsc != NULL, return); get_bundle_variant_data(bundle_data, rsc); if (bundle_data->child) { bundle_data->child->cmds->expand(bundle_data->child, data_set); } for (GList *gIter = bundle_data->replicas; gIter != NULL; gIter = gIter->next) { pe__bundle_replica_t *replica = gIter->data; CRM_ASSERT(replica); if (replica->remote && replica->container && pe__bundle_needs_remote_name(replica->remote)) { /* REMOTE_CONTAINER_HACK: Allow remote nodes to run containers that * run pacemaker-remoted inside, without needing a separate IP for * the container. This is done by configuring the inner remote's * connection host as the magic string "#uname", then * replacing it with the underlying host when needed. */ xmlNode *nvpair = get_xpath_object("//nvpair[@name='" XML_RSC_ATTR_REMOTE_RA_ADDR "']", replica->remote->xml, LOG_ERR); const char *calculated_addr = NULL; calculated_addr = pe__add_bundle_remote_name(replica->remote, nvpair, "value"); if (calculated_addr) { crm_trace("Set address for bundle connection %s to bundle host %s", replica->remote->id, calculated_addr); g_hash_table_replace(replica->remote->parameters, strdup(XML_RSC_ATTR_REMOTE_RA_ADDR), strdup(calculated_addr)); } else { /* The only way to get here is if the remote connection is * neither currently running nor scheduled to run. That means we * won't be doing any operations that require addr (only start * requires it; we additionally use it to compare digests when * unpacking status, promote, and migrate_from history, but * that's already happened by this point). */ crm_info("Unable to determine address for bundle %s remote connection", rsc->id); } } if (replica->ip) { replica->ip->cmds->expand(replica->ip, data_set); } if (replica->container) { replica->container->cmds->expand(replica->container, data_set); } if (replica->remote) { replica->remote->cmds->expand(replica->remote, data_set); } } } gboolean pcmk__bundle_create_probe(pe_resource_t *rsc, pe_node_t *node, pe_action_t *complete, gboolean force, pe_working_set_t * data_set) { bool any_created = FALSE; pe__bundle_variant_data_t *bundle_data = NULL; CRM_CHECK(rsc != NULL, return FALSE); get_bundle_variant_data(bundle_data, rsc); for (GList *gIter = bundle_data->replicas; gIter != NULL; gIter = gIter->next) { pe__bundle_replica_t *replica = gIter->data; CRM_ASSERT(replica); if (replica->ip) { any_created |= replica->ip->cmds->create_probe(replica->ip, node, complete, force, data_set); } if (replica->child && (node->details == replica->node->details)) { any_created |= replica->child->cmds->create_probe(replica->child, node, complete, force, data_set); } if (replica->container) { bool created = replica->container->cmds->create_probe(replica->container, node, complete, force, data_set); if(created) { any_created = TRUE; /* If we're limited to one replica per host (due to * the lack of an IP range probably), then we don't * want any of our peer containers starting until * we've established that no other copies are already * running. * * Partly this is to ensure that nreplicas_per_host is * observed, but also to ensure that the containers * don't fail to start because the necessary port * mappings (which won't include an IP for uniqueness) * are already taken */ for (GList *tIter = bundle_data->replicas; tIter && (bundle_data->nreplicas_per_host == 1); tIter = tIter->next) { pe__bundle_replica_t *other = tIter->data; if ((other != replica) && (other != NULL) && (other->container != NULL)) { custom_action_order(replica->container, pcmk__op_key(replica->container->id, RSC_STATUS, 0), NULL, other->container, pcmk__op_key(other->container->id, RSC_START, 0), NULL, pe_order_optional|pe_order_same_node, data_set); } } } } if (replica->container && replica->remote && replica->remote->cmds->create_probe(replica->remote, node, complete, force, data_set)) { /* Do not probe the remote resource until we know where the * container is running. This is required for REMOTE_CONTAINER_HACK * to correctly probe remote resources. */ char *probe_uuid = pcmk__op_key(replica->remote->id, RSC_STATUS, 0); pe_action_t *probe = find_first_action(replica->remote->actions, probe_uuid, NULL, node); free(probe_uuid); if (probe) { any_created = TRUE; crm_trace("Ordering %s probe on %s", replica->remote->id, node->details->uname); custom_action_order(replica->container, pcmk__op_key(replica->container->id, RSC_START, 0), NULL, replica->remote, NULL, probe, pe_order_probe, data_set); } } } return any_created; } void pcmk__bundle_append_meta(pe_resource_t *rsc, xmlNode *xml) { } void pcmk__bundle_log_actions(pe_resource_t *rsc, pe_working_set_t *data_set, gboolean terminal) { pe__bundle_variant_data_t *bundle_data = NULL; CRM_CHECK(rsc != NULL, return); get_bundle_variant_data(bundle_data, rsc); for (GList *gIter = bundle_data->replicas; gIter != NULL; gIter = gIter->next) { pe__bundle_replica_t *replica = gIter->data; CRM_ASSERT(replica); if (replica->ip) { LogActions(replica->ip, data_set, terminal); } if (replica->container) { LogActions(replica->container, data_set, terminal); } if (replica->remote) { LogActions(replica->remote, data_set, terminal); } if (replica->child) { LogActions(replica->child, data_set, terminal); } } } diff --git a/lib/pacemaker/pcmk_sched_clone.c b/lib/pacemaker/pcmk_sched_clone.c index e419a246c2..5a061510e2 100644 --- a/lib/pacemaker/pcmk_sched_clone.c +++ b/lib/pacemaker/pcmk_sched_clone.c @@ -1,1517 +1,1515 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #define VARIANT_CLONE 1 #include gint sort_clone_instance(gconstpointer a, gconstpointer b, gpointer data_set); static void append_parent_colocation(pe_resource_t * rsc, pe_resource_t * child, gboolean all); static gint sort_rsc_id(gconstpointer a, gconstpointer b) { const pe_resource_t *resource1 = (const pe_resource_t *)a; const pe_resource_t *resource2 = (const pe_resource_t *)b; long num1, num2; CRM_ASSERT(resource1 != NULL); CRM_ASSERT(resource2 != NULL); /* * Sort clone instances numerically by instance number, so instance :10 * comes after :9. */ num1 = strtol(strrchr(resource1->id, ':') + 1, NULL, 10); num2 = strtol(strrchr(resource2->id, ':') + 1, NULL, 10); if (num1 < num2) { return -1; } else if (num1 > num2) { return 1; } return 0; } static pe_node_t * parent_node_instance(const pe_resource_t * rsc, pe_node_t * node) { pe_node_t *ret = NULL; if (node != NULL && rsc->parent) { ret = pe_hash_table_lookup(rsc->parent->allowed_nodes, node->details->id); } else if(node != NULL) { ret = pe_hash_table_lookup(rsc->allowed_nodes, node->details->id); } return ret; } static gboolean did_fail(const pe_resource_t * rsc) { GListPtr gIter = rsc->children; if (pcmk_is_set(rsc->flags, pe_rsc_failed)) { return TRUE; } for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; if (did_fail(child_rsc)) { return TRUE; } } return FALSE; } gint sort_clone_instance(gconstpointer a, gconstpointer b, gpointer data_set) { int rc = 0; pe_node_t *node1 = NULL; pe_node_t *node2 = NULL; pe_node_t *current_node1 = NULL; pe_node_t *current_node2 = NULL; unsigned int nnodes1 = 0; unsigned int nnodes2 = 0; gboolean can1 = TRUE; gboolean can2 = TRUE; const pe_resource_t *resource1 = (const pe_resource_t *)a; const pe_resource_t *resource2 = (const pe_resource_t *)b; CRM_ASSERT(resource1 != NULL); CRM_ASSERT(resource2 != NULL); /* allocation order: * - active instances * - instances running on nodes with the least copies * - active instances on nodes that can't support them or are to be fenced * - failed instances * - inactive instances */ current_node1 = pe__find_active_on(resource1, &nnodes1, NULL); current_node2 = pe__find_active_on(resource2, &nnodes2, NULL); if (nnodes1 && nnodes2) { if (nnodes1 < nnodes2) { crm_trace("%s < %s: running_on", resource1->id, resource2->id); return -1; } else if (nnodes1 > nnodes2) { crm_trace("%s > %s: running_on", resource1->id, resource2->id); return 1; } } node1 = current_node1; node2 = current_node2; if (node1) { pe_node_t *match = pe_hash_table_lookup(resource1->allowed_nodes, node1->details->id); if (match == NULL || match->weight < 0) { crm_trace("%s: current location is unavailable", resource1->id); node1 = NULL; can1 = FALSE; } } if (node2) { pe_node_t *match = pe_hash_table_lookup(resource2->allowed_nodes, node2->details->id); if (match == NULL || match->weight < 0) { crm_trace("%s: current location is unavailable", resource2->id); node2 = NULL; can2 = FALSE; } } if (can1 != can2) { if (can1) { crm_trace("%s < %s: availability of current location", resource1->id, resource2->id); return -1; } crm_trace("%s > %s: availability of current location", resource1->id, resource2->id); return 1; } if (resource1->priority < resource2->priority) { crm_trace("%s < %s: priority", resource1->id, resource2->id); return 1; } else if (resource1->priority > resource2->priority) { crm_trace("%s > %s: priority", resource1->id, resource2->id); return -1; } if (node1 == NULL && node2 == NULL) { crm_trace("%s == %s: not active", resource1->id, resource2->id); return 0; } if (node1 != node2) { if (node1 == NULL) { crm_trace("%s > %s: active", resource1->id, resource2->id); return 1; } else if (node2 == NULL) { crm_trace("%s < %s: active", resource1->id, resource2->id); return -1; } } can1 = can_run_resources(node1); can2 = can_run_resources(node2); if (can1 != can2) { if (can1) { crm_trace("%s < %s: can", resource1->id, resource2->id); return -1; } crm_trace("%s > %s: can", resource1->id, resource2->id); return 1; } node1 = parent_node_instance(resource1, node1); node2 = parent_node_instance(resource2, node2); if (node1 != NULL && node2 == NULL) { crm_trace("%s < %s: not allowed", resource1->id, resource2->id); return -1; } else if (node1 == NULL && node2 != NULL) { crm_trace("%s > %s: not allowed", resource1->id, resource2->id); return 1; } if (node1 == NULL || node2 == NULL) { crm_trace("%s == %s: not allowed", resource1->id, resource2->id); return 0; } if (node1->count < node2->count) { crm_trace("%s < %s: count", resource1->id, resource2->id); return -1; } else if (node1->count > node2->count) { crm_trace("%s > %s: count", resource1->id, resource2->id); return 1; } can1 = did_fail(resource1); can2 = did_fail(resource2); if (can1 != can2) { if (can1) { crm_trace("%s > %s: failed", resource1->id, resource2->id); return 1; } crm_trace("%s < %s: failed", resource1->id, resource2->id); return -1; } if (node1 && node2) { int lpc = 0; int max = 0; pe_node_t *n = NULL; GListPtr gIter = NULL; GListPtr list1 = NULL; GListPtr list2 = NULL; GHashTable *hash1 = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, free); GHashTable *hash2 = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, free); n = pe__copy_node(current_node1); g_hash_table_insert(hash1, (gpointer) n->details->id, n); n = pe__copy_node(current_node2); g_hash_table_insert(hash2, (gpointer) n->details->id, n); if(resource1->parent) { for (gIter = resource1->parent->rsc_cons; gIter; gIter = gIter->next) { - rsc_colocation_t *constraint = (rsc_colocation_t *) gIter->data; + pcmk__colocation_t *constraint = (pcmk__colocation_t *) gIter->data; if (constraint->score == 0) { continue; } crm_trace("Applying %s to %s", constraint->id, resource1->id); hash1 = pcmk__native_merge_weights(constraint->rsc_rh, resource1->id, hash1, constraint->node_attribute, constraint->score / (float) INFINITY, 0); } for (gIter = resource1->parent->rsc_cons_lhs; gIter; gIter = gIter->next) { - rsc_colocation_t *constraint = (rsc_colocation_t *) gIter->data; + pcmk__colocation_t *constraint = (pcmk__colocation_t *) gIter->data; if (constraint->score == 0) { continue; } crm_trace("Applying %s to %s", constraint->id, resource1->id); hash1 = pcmk__native_merge_weights(constraint->rsc_lh, resource1->id, hash1, constraint->node_attribute, constraint->score / (float) INFINITY, pe_weights_positive); } } if(resource2->parent) { for (gIter = resource2->parent->rsc_cons; gIter; gIter = gIter->next) { - rsc_colocation_t *constraint = (rsc_colocation_t *) gIter->data; + pcmk__colocation_t *constraint = (pcmk__colocation_t *) gIter->data; crm_trace("Applying %s to %s", constraint->id, resource2->id); hash2 = pcmk__native_merge_weights(constraint->rsc_rh, resource2->id, hash2, constraint->node_attribute, constraint->score / (float) INFINITY, 0); } for (gIter = resource2->parent->rsc_cons_lhs; gIter; gIter = gIter->next) { - rsc_colocation_t *constraint = (rsc_colocation_t *) gIter->data; + pcmk__colocation_t *constraint = (pcmk__colocation_t *) gIter->data; crm_trace("Applying %s to %s", constraint->id, resource2->id); hash2 = pcmk__native_merge_weights(constraint->rsc_lh, resource2->id, hash2, constraint->node_attribute, constraint->score / (float) INFINITY, pe_weights_positive); } } /* Current location score */ node1 = g_hash_table_lookup(hash1, current_node1->details->id); node2 = g_hash_table_lookup(hash2, current_node2->details->id); if (node1->weight < node2->weight) { if (node1->weight < 0) { crm_trace("%s > %s: current score: %d %d", resource1->id, resource2->id, node1->weight, node2->weight); rc = -1; goto out; } else { crm_trace("%s < %s: current score: %d %d", resource1->id, resource2->id, node1->weight, node2->weight); rc = 1; goto out; } } else if (node1->weight > node2->weight) { crm_trace("%s > %s: current score: %d %d", resource1->id, resource2->id, node1->weight, node2->weight); rc = -1; goto out; } /* All location scores */ list1 = g_hash_table_get_values(hash1); list2 = g_hash_table_get_values(hash2); list1 = sort_nodes_by_weight(list1, current_node1, data_set); list2 = sort_nodes_by_weight(list2, current_node2, data_set); max = g_list_length(list1); if (max < g_list_length(list2)) { max = g_list_length(list2); } for (; lpc < max; lpc++) { node1 = g_list_nth_data(list1, lpc); node2 = g_list_nth_data(list2, lpc); if (node1 == NULL) { crm_trace("%s < %s: colocated score NULL", resource1->id, resource2->id); rc = 1; break; } else if (node2 == NULL) { crm_trace("%s > %s: colocated score NULL", resource1->id, resource2->id); rc = -1; break; } if (node1->weight < node2->weight) { crm_trace("%s < %s: colocated score", resource1->id, resource2->id); rc = 1; break; } else if (node1->weight > node2->weight) { crm_trace("%s > %s: colocated score", resource1->id, resource2->id); rc = -1; break; } } /* Order by reverse uname - same as sort_node_weight() does? */ out: g_hash_table_destroy(hash1); /* Free mem */ g_hash_table_destroy(hash2); /* Free mem */ g_list_free(list1); g_list_free(list2); if (rc != 0) { return rc; } } rc = strcmp(resource1->id, resource2->id); crm_trace("%s %c %s: default", resource1->id, rc < 0 ? '<' : '>', resource2->id); return rc; } static pe_node_t * can_run_instance(pe_resource_t * rsc, pe_node_t * node, int limit) { pe_node_t *local_node = NULL; if (node == NULL && rsc->allowed_nodes) { GHashTableIter iter; g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&local_node)) { can_run_instance(rsc, local_node, limit); } return NULL; } if (!node) { /* make clang analyzer happy */ goto bail; } else if (can_run_resources(node) == FALSE) { goto bail; } else if (pcmk_is_set(rsc->flags, pe_rsc_orphan)) { goto bail; } local_node = parent_node_instance(rsc, node); if (local_node == NULL) { crm_warn("%s cannot run on %s: node not allowed", rsc->id, node->details->uname); goto bail; } else if (local_node->weight < 0) { common_update_score(rsc, node->details->id, local_node->weight); pe_rsc_trace(rsc, "%s cannot run on %s: Parent node weight doesn't allow it.", rsc->id, node->details->uname); } else if (local_node->count < limit) { pe_rsc_trace(rsc, "%s can run on %s (already running %d)", rsc->id, node->details->uname, local_node->count); return local_node; } else { pe_rsc_trace(rsc, "%s cannot run on %s: node full (%d >= %d)", rsc->id, node->details->uname, local_node->count, limit); } bail: if (node) { common_update_score(rsc, node->details->id, -INFINITY); } return NULL; } static pe_node_t * allocate_instance(pe_resource_t *rsc, pe_node_t *prefer, gboolean all_coloc, int limit, pe_working_set_t *data_set) { pe_node_t *chosen = NULL; GHashTable *backup = NULL; CRM_ASSERT(rsc); pe_rsc_trace(rsc, "Checking allocation of %s (preferring %s, using %s parent colocations)", rsc->id, (prefer? prefer->details->uname: "none"), (all_coloc? "all" : "some")); if (!pcmk_is_set(rsc->flags, pe_rsc_provisional)) { return rsc->fns->location(rsc, NULL, FALSE); } else if (pcmk_is_set(rsc->flags, pe_rsc_allocating)) { pe_rsc_debug(rsc, "Dependency loop detected involving %s", rsc->id); return NULL; } /* Only include positive colocation preferences of dependent resources * if not every node will get a copy of the clone */ append_parent_colocation(rsc->parent, rsc, all_coloc); if (prefer) { pe_node_t *local_prefer = g_hash_table_lookup(rsc->allowed_nodes, prefer->details->id); if (local_prefer == NULL || local_prefer->weight < 0) { pe_rsc_trace(rsc, "Not pre-allocating %s to %s - unavailable", rsc->id, prefer->details->uname); return NULL; } } can_run_instance(rsc, NULL, limit); backup = pcmk__copy_node_table(rsc->allowed_nodes); pe_rsc_trace(rsc, "Allocating instance %s", rsc->id); chosen = rsc->cmds->allocate(rsc, prefer, data_set); if (chosen && prefer && (chosen->details != prefer->details)) { crm_info("Not pre-allocating %s to %s because %s is better", rsc->id, prefer->details->uname, chosen->details->uname); g_hash_table_destroy(rsc->allowed_nodes); rsc->allowed_nodes = backup; native_deallocate(rsc); chosen = NULL; backup = NULL; } if (chosen) { pe_node_t *local_node = parent_node_instance(rsc, chosen); if (local_node) { local_node->count++; } else if (pcmk_is_set(rsc->flags, pe_rsc_managed)) { /* what to do? we can't enforce per-node limits in this case */ pcmk__config_err("%s not found in %s (list of %d)", chosen->details->id, rsc->parent->id, g_hash_table_size(rsc->parent->allowed_nodes)); } } if(backup) { g_hash_table_destroy(backup); } return chosen; } static void append_parent_colocation(pe_resource_t * rsc, pe_resource_t * child, gboolean all) { GListPtr gIter = NULL; gIter = rsc->rsc_cons; for (; gIter != NULL; gIter = gIter->next) { - rsc_colocation_t *cons = (rsc_colocation_t *) gIter->data; + pcmk__colocation_t *cons = (pcmk__colocation_t *) gIter->data; if (cons->score == 0) { continue; } if (all || cons->score < 0 || cons->score == INFINITY) { child->rsc_cons = g_list_prepend(child->rsc_cons, cons); } } gIter = rsc->rsc_cons_lhs; for (; gIter != NULL; gIter = gIter->next) { - rsc_colocation_t *cons = (rsc_colocation_t *) gIter->data; + pcmk__colocation_t *cons = (pcmk__colocation_t *) gIter->data; if (cons->score == 0) { continue; } if (all || cons->score < 0) { child->rsc_cons_lhs = g_list_prepend(child->rsc_cons_lhs, cons); } } } void distribute_children(pe_resource_t *rsc, GListPtr children, GListPtr nodes, int max, int per_host_max, pe_working_set_t * data_set); void distribute_children(pe_resource_t *rsc, GListPtr children, GListPtr nodes, int max, int per_host_max, pe_working_set_t * data_set) { int loop_max = 0; int allocated = 0; int available_nodes = 0; /* count now tracks the number of clones currently allocated */ for(GListPtr nIter = nodes; nIter != NULL; nIter = nIter->next) { pe_node_t *node = nIter->data; node->count = 0; if (can_run_resources(node)) { available_nodes++; } } if(available_nodes) { loop_max = max / available_nodes; } if (loop_max < 1) { loop_max = 1; } pe_rsc_debug(rsc, "Allocating up to %d %s instances to a possible %d nodes (at most %d per host, %d optimal)", max, rsc->id, available_nodes, per_host_max, loop_max); /* Pre-allocate as many instances as we can to their current location */ for (GListPtr gIter = children; gIter != NULL && allocated < max; gIter = gIter->next) { pe_resource_t *child = (pe_resource_t *) gIter->data; if (child->running_on && pcmk_is_set(child->flags, pe_rsc_provisional) && !pcmk_is_set(child->flags, pe_rsc_failed)) { pe_node_t *child_node = pe__current_node(child); pe_node_t *local_node = parent_node_instance(child, child_node); pe_rsc_trace(rsc, "Checking pre-allocation of %s to %s (%d remaining of %d)", child->id, child_node->details->uname, max - allocated, max); if (can_run_resources(child_node) == FALSE || child_node->weight < 0) { pe_rsc_trace(rsc, "Not pre-allocating because %s can not run %s", child_node->details->uname, child->id); } else if(local_node && local_node->count >= loop_max) { pe_rsc_trace(rsc, "Not pre-allocating because %s already allocated optimal instances", child_node->details->uname); } else if (allocate_instance(child, child_node, max < available_nodes, per_host_max, data_set)) { pe_rsc_trace(rsc, "Pre-allocated %s to %s", child->id, child_node->details->uname); allocated++; } } } pe_rsc_trace(rsc, "Done pre-allocating (%d of %d)", allocated, max); for (GListPtr gIter = children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child = (pe_resource_t *) gIter->data; if (child->running_on != NULL) { pe_node_t *child_node = pe__current_node(child); pe_node_t *local_node = parent_node_instance(child, child_node); if (local_node == NULL) { crm_err("%s is running on %s which isn't allowed", child->id, child_node->details->uname); } } if (!pcmk_is_set(child->flags, pe_rsc_provisional)) { } else if (allocated >= max) { pe_rsc_debug(rsc, "Child %s not allocated - limit reached %d %d", child->id, allocated, max); resource_location(child, NULL, -INFINITY, "clone:limit_reached", data_set); } else { if (allocate_instance(child, NULL, max < available_nodes, per_host_max, data_set)) { allocated++; } } } pe_rsc_debug(rsc, "Allocated %d %s instances of a possible %d", allocated, rsc->id, max); } pe_node_t * pcmk__clone_allocate(pe_resource_t *rsc, pe_node_t *prefer, pe_working_set_t *data_set) { GListPtr nodes = NULL; clone_variant_data_t *clone_data = NULL; get_clone_variant_data(clone_data, rsc); if (!pcmk_is_set(rsc->flags, pe_rsc_provisional)) { return NULL; } else if (pcmk_is_set(rsc->flags, pe_rsc_allocating)) { pe_rsc_debug(rsc, "Dependency loop detected involving %s", rsc->id); return NULL; } if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) { apply_master_prefs(rsc); } pe__set_resource_flags(rsc, pe_rsc_allocating); /* this information is used by sort_clone_instance() when deciding in which * order to allocate clone instances */ for (GListPtr gIter = rsc->rsc_cons; gIter != NULL; gIter = gIter->next) { - rsc_colocation_t *constraint = (rsc_colocation_t *) gIter->data; + pcmk__colocation_t *constraint = (pcmk__colocation_t *) gIter->data; if (constraint->score == 0) { continue; } pe_rsc_trace(rsc, "%s: Allocating %s first", rsc->id, constraint->rsc_rh->id); constraint->rsc_rh->cmds->allocate(constraint->rsc_rh, prefer, data_set); } for (GListPtr gIter = rsc->rsc_cons_lhs; gIter != NULL; gIter = gIter->next) { - rsc_colocation_t *constraint = (rsc_colocation_t *) gIter->data; + pcmk__colocation_t *constraint = (pcmk__colocation_t *) gIter->data; - if (constraint->score == 0) { - continue; + if (pcmk__colocation_applies(rsc, constraint, false)) { + rsc->allowed_nodes = constraint->rsc_lh->cmds->merge_weights(constraint->rsc_lh, + rsc->id, rsc->allowed_nodes, constraint->node_attribute, + constraint->score / (float) INFINITY, + pe_weights_rollback|pe_weights_positive); } - rsc->allowed_nodes = - constraint->rsc_lh->cmds->merge_weights(constraint->rsc_lh, rsc->id, rsc->allowed_nodes, - constraint->node_attribute, - (float)constraint->score / INFINITY, - (pe_weights_rollback | pe_weights_positive)); } pe__show_node_weights(!show_scores, rsc, __func__, rsc->allowed_nodes); nodes = g_hash_table_get_values(rsc->allowed_nodes); nodes = sort_nodes_by_weight(nodes, NULL, data_set); rsc->children = g_list_sort_with_data(rsc->children, sort_clone_instance, data_set); distribute_children(rsc, rsc->children, nodes, clone_data->clone_max, clone_data->clone_node_max, data_set); g_list_free(nodes); if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) { pcmk__set_instance_roles(rsc, data_set); } pe__clear_resource_flags(rsc, pe_rsc_provisional|pe_rsc_allocating); pe_rsc_trace(rsc, "Done allocating %s", rsc->id); return NULL; } static void clone_update_pseudo_status(pe_resource_t * rsc, gboolean * stopping, gboolean * starting, gboolean * active) { GListPtr gIter = NULL; if (rsc->children) { gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child = (pe_resource_t *) gIter->data; clone_update_pseudo_status(child, stopping, starting, active); } return; } CRM_ASSERT(active != NULL); CRM_ASSERT(starting != NULL); CRM_ASSERT(stopping != NULL); if (rsc->running_on) { *active = TRUE; } gIter = rsc->actions; for (; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; if (*starting && *stopping) { return; } else if (pcmk_is_set(action->flags, pe_action_optional)) { pe_rsc_trace(rsc, "Skipping optional: %s", action->uuid); continue; } else if (!pcmk_any_flags_set(action->flags, pe_action_pseudo|pe_action_runnable)) { pe_rsc_trace(rsc, "Skipping unrunnable: %s", action->uuid); continue; } else if (pcmk__str_eq(RSC_STOP, action->task, pcmk__str_casei)) { pe_rsc_trace(rsc, "Stopping due to: %s", action->uuid); *stopping = TRUE; } else if (pcmk__str_eq(RSC_START, action->task, pcmk__str_casei)) { if (!pcmk_is_set(action->flags, pe_action_runnable)) { pe_rsc_trace(rsc, "Skipping pseudo-op: %s run=%d, pseudo=%d", action->uuid, pcmk_is_set(action->flags, pe_action_runnable), pcmk_is_set(action->flags, pe_action_pseudo)); } else { pe_rsc_trace(rsc, "Starting due to: %s", action->uuid); pe_rsc_trace(rsc, "%s run=%d, pseudo=%d", action->uuid, pcmk_is_set(action->flags, pe_action_runnable), pcmk_is_set(action->flags, pe_action_pseudo)); *starting = TRUE; } } } } static pe_action_t * find_rsc_action(pe_resource_t *rsc, const char *task, gboolean active_only, GList **list) { pe_action_t *match = NULL; GListPtr possible = NULL; GListPtr active = NULL; possible = pe__resource_actions(rsc, NULL, task, FALSE); if (active_only) { GListPtr gIter = possible; for (; gIter != NULL; gIter = gIter->next) { pe_action_t *op = (pe_action_t *) gIter->data; if (!pcmk_is_set(op->flags, pe_action_optional)) { active = g_list_prepend(active, op); } } if (active && pcmk__list_of_1(active)) { match = g_list_nth_data(active, 0); } if (list) { *list = active; active = NULL; } } else if (possible && pcmk__list_of_1(possible)) { match = g_list_nth_data(possible, 0); } if (list) { *list = possible; possible = NULL; } if (possible) { g_list_free(possible); } if (active) { g_list_free(active); } return match; } static void child_ordering_constraints(pe_resource_t * rsc, pe_working_set_t * data_set) { pe_action_t *stop = NULL; pe_action_t *start = NULL; pe_action_t *last_stop = NULL; pe_action_t *last_start = NULL; GListPtr gIter = NULL; gboolean active_only = TRUE; /* change to false to get the old behavior */ clone_variant_data_t *clone_data = NULL; get_clone_variant_data(clone_data, rsc); if (clone_data->ordered == FALSE) { return; } /* we have to maintain a consistent sorted child list when building order constraints */ rsc->children = g_list_sort(rsc->children, sort_rsc_id); for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child = (pe_resource_t *) gIter->data; stop = find_rsc_action(child, RSC_STOP, active_only, NULL); if (stop) { if (last_stop) { /* child/child relative stop */ order_actions(stop, last_stop, pe_order_optional); } last_stop = stop; } start = find_rsc_action(child, RSC_START, active_only, NULL); if (start) { if (last_start) { /* child/child relative start */ order_actions(last_start, start, pe_order_optional); } last_start = start; } } } void clone_create_actions(pe_resource_t *rsc, pe_working_set_t *data_set) { clone_variant_data_t *clone_data = NULL; get_clone_variant_data(clone_data, rsc); clone_create_pseudo_actions(rsc, rsc->children, &clone_data->start_notify, &clone_data->stop_notify,data_set); child_ordering_constraints(rsc, data_set); if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) { create_promotable_actions(rsc, data_set); } } void clone_create_pseudo_actions( pe_resource_t * rsc, GListPtr children, notify_data_t **start_notify, notify_data_t **stop_notify, pe_working_set_t * data_set) { gboolean child_active = FALSE; gboolean child_starting = FALSE; gboolean child_stopping = FALSE; gboolean allow_dependent_migrations = TRUE; pe_action_t *stop = NULL; pe_action_t *stopped = NULL; pe_action_t *start = NULL; pe_action_t *started = NULL; pe_rsc_trace(rsc, "Creating actions for %s", rsc->id); for (GListPtr gIter = children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; gboolean starting = FALSE; gboolean stopping = FALSE; child_rsc->cmds->create_actions(child_rsc, data_set); clone_update_pseudo_status(child_rsc, &stopping, &starting, &child_active); if (stopping && starting) { allow_dependent_migrations = FALSE; } child_stopping |= stopping; child_starting |= starting; } /* start */ start = create_pseudo_resource_op(rsc, RSC_START, !child_starting, TRUE, data_set); started = create_pseudo_resource_op(rsc, RSC_STARTED, !child_starting, FALSE, data_set); started->priority = INFINITY; if (child_active || child_starting) { update_action_flags(started, pe_action_runnable, __func__, __LINE__); } if (start_notify != NULL && *start_notify == NULL) { *start_notify = create_notification_boundaries(rsc, RSC_START, start, started, data_set); } /* stop */ stop = create_pseudo_resource_op(rsc, RSC_STOP, !child_stopping, TRUE, data_set); stopped = create_pseudo_resource_op(rsc, RSC_STOPPED, !child_stopping, TRUE, data_set); stopped->priority = INFINITY; if (allow_dependent_migrations) { update_action_flags(stop, pe_action_migrate_runnable, __func__, __LINE__); } if (stop_notify != NULL && *stop_notify == NULL) { *stop_notify = create_notification_boundaries(rsc, RSC_STOP, stop, stopped, data_set); if (start_notify && *start_notify && *stop_notify) { order_actions((*stop_notify)->post_done, (*start_notify)->pre, pe_order_optional); } } } void clone_internal_constraints(pe_resource_t *rsc, pe_working_set_t *data_set) { pe_resource_t *last_rsc = NULL; GListPtr gIter; clone_variant_data_t *clone_data = NULL; get_clone_variant_data(clone_data, rsc); pe_rsc_trace(rsc, "Internal constraints for %s", rsc->id); new_rsc_order(rsc, RSC_STOPPED, rsc, RSC_START, pe_order_optional, data_set); new_rsc_order(rsc, RSC_START, rsc, RSC_STARTED, pe_order_runnable_left, data_set); new_rsc_order(rsc, RSC_STOP, rsc, RSC_STOPPED, pe_order_runnable_left, data_set); if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) { new_rsc_order(rsc, RSC_DEMOTED, rsc, RSC_STOP, pe_order_optional, data_set); new_rsc_order(rsc, RSC_STARTED, rsc, RSC_PROMOTE, pe_order_runnable_left, data_set); } if (clone_data->ordered) { /* we have to maintain a consistent sorted child list when building order constraints */ rsc->children = g_list_sort(rsc->children, sort_rsc_id); } for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; child_rsc->cmds->internal_constraints(child_rsc, data_set); order_start_start(rsc, child_rsc, pe_order_runnable_left | pe_order_implies_first_printed); new_rsc_order(child_rsc, RSC_START, rsc, RSC_STARTED, pe_order_implies_then_printed, data_set); if (clone_data->ordered && last_rsc) { order_start_start(last_rsc, child_rsc, pe_order_optional); } order_stop_stop(rsc, child_rsc, pe_order_implies_first_printed); new_rsc_order(child_rsc, RSC_STOP, rsc, RSC_STOPPED, pe_order_implies_then_printed, data_set); if (clone_data->ordered && last_rsc) { order_stop_stop(child_rsc, last_rsc, pe_order_optional); } last_rsc = child_rsc; } if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) { promotable_constraints(rsc, data_set); } } bool assign_node(pe_resource_t * rsc, pe_node_t * node, gboolean force) { bool changed = FALSE; if (rsc->children) { for (GListPtr gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; changed |= assign_node(child_rsc, node, force); } return changed; } if (rsc->allocated_to != NULL) { changed = true; } native_assign_node(rsc, NULL, node, force); return changed; } gboolean is_child_compatible(pe_resource_t *child_rsc, pe_node_t * local_node, enum rsc_role_e filter, gboolean current) { pe_node_t *node = NULL; enum rsc_role_e next_role = child_rsc->fns->state(child_rsc, current); CRM_CHECK(child_rsc && local_node, return FALSE); if (is_set_recursive(child_rsc, pe_rsc_block, TRUE) == FALSE) { /* We only want instances that haven't failed */ node = child_rsc->fns->location(child_rsc, NULL, current); } if (filter != RSC_ROLE_UNKNOWN && next_role != filter) { crm_trace("Filtered %s", child_rsc->id); return FALSE; } if (node && (node->details == local_node->details)) { return TRUE; } else if (node) { crm_trace("%s - %s vs %s", child_rsc->id, node->details->uname, local_node->details->uname); } else { crm_trace("%s - not allocated %d", child_rsc->id, current); } return FALSE; } pe_resource_t * find_compatible_child(pe_resource_t *local_child, pe_resource_t *rsc, enum rsc_role_e filter, gboolean current, pe_working_set_t *data_set) { pe_resource_t *pair = NULL; GListPtr gIter = NULL; GListPtr scratch = NULL; pe_node_t *local_node = NULL; local_node = local_child->fns->location(local_child, NULL, current); if (local_node) { return find_compatible_child_by_node(local_child, local_node, rsc, filter, current); } scratch = g_hash_table_get_values(local_child->allowed_nodes); scratch = sort_nodes_by_weight(scratch, NULL, data_set); gIter = scratch; for (; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; pair = find_compatible_child_by_node(local_child, node, rsc, filter, current); if (pair) { goto done; } } pe_rsc_debug(rsc, "Can't pair %s with %s", local_child->id, rsc->id); done: g_list_free(scratch); return pair; } void clone_rsc_colocation_lh(pe_resource_t *rsc_lh, pe_resource_t *rsc_rh, - rsc_colocation_t *constraint, + pcmk__colocation_t *constraint, pe_working_set_t *data_set) { /* -- Never called -- * * Instead we add the colocation constraints to the child and call from there */ CRM_ASSERT(FALSE); } void clone_rsc_colocation_rh(pe_resource_t *rsc_lh, pe_resource_t *rsc_rh, - rsc_colocation_t *constraint, + pcmk__colocation_t *constraint, pe_working_set_t *data_set) { GListPtr gIter = NULL; gboolean do_interleave = FALSE; const char *interleave_s = NULL; CRM_CHECK(constraint != NULL, return); CRM_CHECK(rsc_lh != NULL, pe_err("rsc_lh was NULL for %s", constraint->id); return); CRM_CHECK(rsc_rh != NULL, pe_err("rsc_rh was NULL for %s", constraint->id); return); CRM_CHECK(rsc_lh->variant == pe_native, return); if (constraint->score == 0) { return; } pe_rsc_trace(rsc_rh, "Processing constraint %s: %s -> %s %d", constraint->id, rsc_lh->id, rsc_rh->id, constraint->score); if (pcmk_is_set(rsc_rh->flags, pe_rsc_promotable)) { if (pcmk_is_set(rsc_rh->flags, pe_rsc_provisional)) { pe_rsc_trace(rsc_rh, "%s is still provisional", rsc_rh->id); return; } else if (constraint->role_rh == RSC_ROLE_UNKNOWN) { pe_rsc_trace(rsc_rh, "Handling %s as a clone colocation", constraint->id); } else { promotable_colocation_rh(rsc_lh, rsc_rh, constraint, data_set); return; } } /* only the LHS side needs to be labeled as interleave */ interleave_s = g_hash_table_lookup(constraint->rsc_lh->meta, XML_RSC_ATTR_INTERLEAVE); if(crm_is_true(interleave_s) && constraint->rsc_lh->variant > pe_group) { // TODO: Do we actually care about multiple RH copies sharing a LH copy anymore? if (copies_per_node(constraint->rsc_lh) != copies_per_node(constraint->rsc_rh)) { pcmk__config_err("Cannot interleave %s and %s because they do not " "support the same number of instances per node", constraint->rsc_lh->id, constraint->rsc_rh->id); } else { do_interleave = TRUE; } } if (pcmk_is_set(rsc_rh->flags, pe_rsc_provisional)) { pe_rsc_trace(rsc_rh, "%s is still provisional", rsc_rh->id); return; } else if (do_interleave) { pe_resource_t *rh_child = NULL; rh_child = find_compatible_child(rsc_lh, rsc_rh, RSC_ROLE_UNKNOWN, FALSE, data_set); if (rh_child) { pe_rsc_debug(rsc_rh, "Pairing %s with %s", rsc_lh->id, rh_child->id); rsc_lh->cmds->rsc_colocation_lh(rsc_lh, rh_child, constraint, data_set); } else if (constraint->score >= INFINITY) { crm_notice("Cannot pair %s with instance of %s", rsc_lh->id, rsc_rh->id); assign_node(rsc_lh, NULL, TRUE); } else { pe_rsc_debug(rsc_rh, "Cannot pair %s with instance of %s", rsc_lh->id, rsc_rh->id); } return; } else if (constraint->score >= INFINITY) { GListPtr rhs = NULL; gIter = rsc_rh->children; for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; pe_node_t *chosen = child_rsc->fns->location(child_rsc, NULL, FALSE); if (chosen != NULL && is_set_recursive(child_rsc, pe_rsc_block, TRUE) == FALSE) { pe_rsc_trace(rsc_rh, "Allowing %s: %s %d", constraint->id, chosen->details->uname, chosen->weight); rhs = g_list_prepend(rhs, chosen); } } node_list_exclude(rsc_lh->allowed_nodes, rhs, FALSE); g_list_free(rhs); return; } gIter = rsc_rh->children; for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; child_rsc->cmds->rsc_colocation_rh(rsc_lh, child_rsc, constraint, data_set); } } enum action_tasks clone_child_action(pe_action_t * action) { enum action_tasks result = no_action; pe_resource_t *child = (pe_resource_t *) action->rsc->children->data; if (pcmk__strcase_any_of(action->task, "notify", "notified", NULL)) { /* Find the action we're notifying about instead */ int stop = 0; char *key = action->uuid; int lpc = strlen(key); for (; lpc > 0; lpc--) { if (key[lpc] == '_' && stop == 0) { stop = lpc; } else if (key[lpc] == '_') { char *task_mutable = NULL; lpc++; task_mutable = strdup(key + lpc); task_mutable[stop - lpc] = 0; crm_trace("Extracted action '%s' from '%s'", task_mutable, key); result = get_complex_task(child, task_mutable, TRUE); free(task_mutable); break; } } } else { result = get_complex_task(child, action->task, TRUE); } return result; } #define pe__clear_action_summary_flags(flags, action, flag) do { \ flags = pcmk__clear_flags_as(__func__, __LINE__, LOG_TRACE, \ "Action summary", action->rsc->id, \ flags, flag, #flag); \ } while (0) enum pe_action_flags summary_action_flags(pe_action_t * action, GListPtr children, pe_node_t * node) { GListPtr gIter = NULL; gboolean any_runnable = FALSE; gboolean check_runnable = TRUE; enum action_tasks task = clone_child_action(action); enum pe_action_flags flags = (pe_action_optional | pe_action_runnable | pe_action_pseudo); const char *task_s = task2text(task); for (gIter = children; gIter != NULL; gIter = gIter->next) { pe_action_t *child_action = NULL; pe_resource_t *child = (pe_resource_t *) gIter->data; child_action = find_first_action(child->actions, NULL, task_s, child->children ? NULL : node); pe_rsc_trace(action->rsc, "Checking for %s in %s on %s (%s)", task_s, child->id, node ? node->details->uname : "none", child_action?child_action->uuid:"NA"); if (child_action) { enum pe_action_flags child_flags = child->cmds->action_flags(child_action, node); if (pcmk_is_set(flags, pe_action_optional) && !pcmk_is_set(child_flags, pe_action_optional)) { pe_rsc_trace(child, "%s is mandatory because of %s", action->uuid, child_action->uuid); pe__clear_action_summary_flags(flags, action, pe_action_optional); pe__clear_action_flags(action, pe_action_optional); } if (pcmk_is_set(child_flags, pe_action_runnable)) { any_runnable = TRUE; } } } if (check_runnable && any_runnable == FALSE) { pe_rsc_trace(action->rsc, "%s is not runnable because no children are", action->uuid); pe__clear_action_summary_flags(flags, action, pe_action_runnable); if (node == NULL) { pe__clear_action_flags(action, pe_action_runnable); } } return flags; } enum pe_action_flags clone_action_flags(pe_action_t * action, pe_node_t * node) { return summary_action_flags(action, action->rsc->children, node); } void clone_rsc_location(pe_resource_t *rsc, pe__location_t *constraint) { GListPtr gIter = rsc->children; pe_rsc_trace(rsc, "Processing location constraint %s for %s", constraint->id, rsc->id); native_rsc_location(rsc, constraint); for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; child_rsc->cmds->rsc_location(child_rsc, constraint); } } void clone_expand(pe_resource_t * rsc, pe_working_set_t * data_set) { GListPtr gIter = NULL; clone_variant_data_t *clone_data = NULL; get_clone_variant_data(clone_data, rsc); gIter = rsc->actions; for (; gIter != NULL; gIter = gIter->next) { pe_action_t *op = (pe_action_t *) gIter->data; rsc->cmds->action_flags(op, NULL); } if (clone_data->start_notify) { collect_notification_data(rsc, TRUE, TRUE, clone_data->start_notify); pcmk__create_notification_keys(rsc, clone_data->start_notify, data_set); create_notifications(rsc, clone_data->start_notify, data_set); } if (clone_data->stop_notify) { collect_notification_data(rsc, TRUE, TRUE, clone_data->stop_notify); pcmk__create_notification_keys(rsc, clone_data->stop_notify, data_set); create_notifications(rsc, clone_data->stop_notify, data_set); } if (clone_data->promote_notify) { collect_notification_data(rsc, TRUE, TRUE, clone_data->promote_notify); pcmk__create_notification_keys(rsc, clone_data->promote_notify, data_set); create_notifications(rsc, clone_data->promote_notify, data_set); } if (clone_data->demote_notify) { collect_notification_data(rsc, TRUE, TRUE, clone_data->demote_notify); pcmk__create_notification_keys(rsc, clone_data->demote_notify, data_set); create_notifications(rsc, clone_data->demote_notify, data_set); } /* Now that the notifcations have been created we can expand the children */ gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; child_rsc->cmds->expand(child_rsc, data_set); } native_expand(rsc, data_set); /* The notifications are in the graph now, we can destroy the notify_data */ free_notification_data(clone_data->demote_notify); clone_data->demote_notify = NULL; free_notification_data(clone_data->stop_notify); clone_data->stop_notify = NULL; free_notification_data(clone_data->start_notify); clone_data->start_notify = NULL; free_notification_data(clone_data->promote_notify); clone_data->promote_notify = NULL; } // Check whether a resource or any of its children is known on node static bool rsc_known_on(const pe_resource_t *rsc, const pe_node_t *node) { if (rsc->children) { for (GList *child_iter = rsc->children; child_iter != NULL; child_iter = child_iter->next) { pe_resource_t *child = (pe_resource_t *) child_iter->data; if (rsc_known_on(child, node)) { return TRUE; } } } else if (rsc->known_on) { GHashTableIter iter; pe_node_t *known_node = NULL; g_hash_table_iter_init(&iter, rsc->known_on); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &known_node)) { if (node->details == known_node->details) { return TRUE; } } } return FALSE; } // Look for an instance of clone that is known on node static pe_resource_t * find_instance_on(const pe_resource_t *clone, const pe_node_t *node) { for (GList *gIter = clone->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child = (pe_resource_t *) gIter->data; if (rsc_known_on(child, node)) { return child; } } return NULL; } // For unique clones, probe each instance separately static gboolean probe_unique_clone(pe_resource_t *rsc, pe_node_t *node, pe_action_t *complete, gboolean force, pe_working_set_t *data_set) { gboolean any_created = FALSE; for (GList *child_iter = rsc->children; child_iter != NULL; child_iter = child_iter->next) { pe_resource_t *child = (pe_resource_t *) child_iter->data; any_created |= child->cmds->create_probe(child, node, complete, force, data_set); } return any_created; } // For anonymous clones, only a single instance needs to be probed static gboolean probe_anonymous_clone(pe_resource_t *rsc, pe_node_t *node, pe_action_t *complete, gboolean force, pe_working_set_t *data_set) { // First, check if we probed an instance on this node last time pe_resource_t *child = find_instance_on(rsc, node); // Otherwise, check if we plan to start an instance on this node if (child == NULL) { for (GList *child_iter = rsc->children; child_iter && !child; child_iter = child_iter->next) { pe_node_t *local_node = NULL; pe_resource_t *child_rsc = (pe_resource_t *) child_iter->data; if (child_rsc) { /* make clang analyzer happy */ local_node = child_rsc->fns->location(child_rsc, NULL, FALSE); if (local_node && (local_node->details == node->details)) { child = child_rsc; } } } } // Otherwise, use the first clone instance if (child == NULL) { child = rsc->children->data; } CRM_ASSERT(child); return child->cmds->create_probe(child, node, complete, force, data_set); } gboolean clone_create_probe(pe_resource_t * rsc, pe_node_t * node, pe_action_t * complete, gboolean force, pe_working_set_t * data_set) { gboolean any_created = FALSE; CRM_ASSERT(rsc); rsc->children = g_list_sort(rsc->children, sort_rsc_id); if (rsc->children == NULL) { pe_warn("Clone %s has no children", rsc->id); return FALSE; } if (rsc->exclusive_discover) { pe_node_t *allowed = g_hash_table_lookup(rsc->allowed_nodes, node->details->id); if (allowed && allowed->rsc_discover_mode != pe_discover_exclusive) { /* exclusive discover is enabled and this node is not marked * as a node this resource should be discovered on * * remove the node from allowed_nodes so that the * notification contains only nodes that we might ever run * on */ g_hash_table_remove(rsc->allowed_nodes, node->details->id); /* Bit of a shortcut - might as well take it */ return FALSE; } } if (pcmk_is_set(rsc->flags, pe_rsc_unique)) { any_created = probe_unique_clone(rsc, node, complete, force, data_set); } else { any_created = probe_anonymous_clone(rsc, node, complete, force, data_set); } return any_created; } void clone_append_meta(pe_resource_t * rsc, xmlNode * xml) { char *name = NULL; clone_variant_data_t *clone_data = NULL; get_clone_variant_data(clone_data, rsc); name = crm_meta_name(XML_RSC_ATTR_UNIQUE); crm_xml_add(xml, name, pe__rsc_bool_str(rsc, pe_rsc_unique)); free(name); name = crm_meta_name(XML_RSC_ATTR_NOTIFY); crm_xml_add(xml, name, pe__rsc_bool_str(rsc, pe_rsc_notify)); free(name); name = crm_meta_name(XML_RSC_ATTR_INCARNATION_MAX); crm_xml_add_int(xml, name, clone_data->clone_max); free(name); name = crm_meta_name(XML_RSC_ATTR_INCARNATION_NODEMAX); crm_xml_add_int(xml, name, clone_data->clone_node_max); free(name); if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) { name = crm_meta_name(XML_RSC_ATTR_PROMOTED_MAX); crm_xml_add_int(xml, name, clone_data->promoted_max); free(name); name = crm_meta_name(XML_RSC_ATTR_PROMOTED_NODEMAX); crm_xml_add_int(xml, name, clone_data->promoted_node_max); free(name); /* @COMPAT Maintain backward compatibility with resource agents that * expect the old names (deprecated since 2.0.0). */ name = crm_meta_name(XML_RSC_ATTR_MASTER_MAX); crm_xml_add_int(xml, name, clone_data->promoted_max); free(name); name = crm_meta_name(XML_RSC_ATTR_MASTER_NODEMAX); crm_xml_add_int(xml, name, clone_data->promoted_node_max); free(name); } } diff --git a/lib/pacemaker/pcmk_sched_constraints.c b/lib/pacemaker/pcmk_sched_constraints.c index 121754d35e..0029ad742a 100644 --- a/lib/pacemaker/pcmk_sched_constraints.c +++ b/lib/pacemaker/pcmk_sched_constraints.c @@ -1,3065 +1,3065 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include enum pe_order_kind { pe_order_kind_optional, pe_order_kind_mandatory, pe_order_kind_serialize, }; #define EXPAND_CONSTRAINT_IDREF(__set, __rsc, __name) do { \ __rsc = pe_find_constraint_resource(data_set->resources, __name); \ if(__rsc == NULL) { \ pcmk__config_err("%s: No resource found for %s", __set, __name); \ return FALSE; \ } \ } while(0) enum pe_ordering get_flags(const char *id, enum pe_order_kind kind, const char *action_first, const char *action_then, gboolean invert); enum pe_ordering get_asymmetrical_flags(enum pe_order_kind kind); static pe__location_t *generate_location_rule(pe_resource_t *rsc, xmlNode *rule_xml, const char *discovery, crm_time_t *next_change, pe_working_set_t *data_set, pe_match_data_t *match_data); static void unpack_location(xmlNode *xml_obj, pe_working_set_t *data_set); +static void unpack_rsc_colocation(xmlNode *xml_obj, pe_working_set_t *data_set); static bool evaluate_lifetime(xmlNode *lifetime, pe_working_set_t *data_set) { bool result = FALSE; crm_time_t *next_change = crm_time_new_undefined(); result = pe_evaluate_rules(lifetime, NULL, data_set->now, next_change); if (crm_time_is_defined(next_change)) { time_t recheck = (time_t) crm_time_get_seconds_since_epoch(next_change); pe__update_recheck_time(recheck, data_set); } crm_time_free(next_change); return result; } gboolean unpack_constraints(xmlNode * xml_constraints, pe_working_set_t * data_set) { xmlNode *xml_obj = NULL; xmlNode *lifetime = NULL; for (xml_obj = pcmk__xe_first_child(xml_constraints); xml_obj != NULL; xml_obj = pcmk__xe_next(xml_obj)) { const char *id = crm_element_value(xml_obj, XML_ATTR_ID); const char *tag = crm_element_name(xml_obj); if (id == NULL) { pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID, tag); continue; } crm_trace("Processing constraint %s %s", tag, id); lifetime = first_named_child(xml_obj, "lifetime"); if (lifetime) { pcmk__config_warn("Support for 'lifetime' attribute (in %s) is " "deprecated (the rules it contains should " "instead be direct descendents of the " "constraint object)", id); } if (lifetime && !evaluate_lifetime(lifetime, data_set)) { crm_info("Constraint %s %s is not active", tag, id); } else if (pcmk__str_eq(XML_CONS_TAG_RSC_ORDER, tag, pcmk__str_casei)) { unpack_rsc_order(xml_obj, data_set); } else if (pcmk__str_eq(XML_CONS_TAG_RSC_DEPEND, tag, pcmk__str_casei)) { unpack_rsc_colocation(xml_obj, data_set); } else if (pcmk__str_eq(XML_CONS_TAG_RSC_LOCATION, tag, pcmk__str_casei)) { unpack_location(xml_obj, data_set); } else if (pcmk__str_eq(XML_CONS_TAG_RSC_TICKET, tag, pcmk__str_casei)) { unpack_rsc_ticket(xml_obj, data_set); } else { pe_err("Unsupported constraint type: %s", tag); } } return TRUE; } static const char * invert_action(const char *action) { if (pcmk__str_eq(action, RSC_START, pcmk__str_casei)) { return RSC_STOP; } else if (pcmk__str_eq(action, RSC_STOP, pcmk__str_casei)) { return RSC_START; } else if (pcmk__str_eq(action, RSC_PROMOTE, pcmk__str_casei)) { return RSC_DEMOTE; } else if (pcmk__str_eq(action, RSC_DEMOTE, pcmk__str_casei)) { return RSC_PROMOTE; } else if (pcmk__str_eq(action, RSC_PROMOTED, pcmk__str_casei)) { return RSC_DEMOTED; } else if (pcmk__str_eq(action, RSC_DEMOTED, pcmk__str_casei)) { return RSC_PROMOTED; } else if (pcmk__str_eq(action, RSC_STARTED, pcmk__str_casei)) { return RSC_STOPPED; } else if (pcmk__str_eq(action, RSC_STOPPED, pcmk__str_casei)) { return RSC_STARTED; } crm_warn("Unknown action '%s' specified in order constraint", action); return NULL; } static enum pe_order_kind get_ordering_type(xmlNode * xml_obj) { enum pe_order_kind kind_e = pe_order_kind_mandatory; const char *kind = crm_element_value(xml_obj, XML_ORDER_ATTR_KIND); if (kind == NULL) { const char *score = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE); kind_e = pe_order_kind_mandatory; if (score) { // @COMPAT deprecated informally since 1.0.7, formally since 2.0.1 int score_i = char2score(score); if (score_i == 0) { kind_e = pe_order_kind_optional; } pe_warn_once(pe_wo_order_score, "Support for 'score' in rsc_order is deprecated " "and will be removed in a future release (use 'kind' instead)"); } } else if (pcmk__str_eq(kind, "Mandatory", pcmk__str_casei)) { kind_e = pe_order_kind_mandatory; } else if (pcmk__str_eq(kind, "Optional", pcmk__str_casei)) { kind_e = pe_order_kind_optional; } else if (pcmk__str_eq(kind, "Serialize", pcmk__str_casei)) { kind_e = pe_order_kind_serialize; } else { pcmk__config_err("Resetting '" XML_ORDER_ATTR_KIND "' for constraint " "'%s' to Mandatory because '%s' is not valid", crm_str(ID(xml_obj)), kind); } return kind_e; } static pe_resource_t * pe_find_constraint_resource(GListPtr rsc_list, const char *id) { GListPtr rIter = NULL; for (rIter = rsc_list; id && rIter; rIter = rIter->next) { pe_resource_t *parent = rIter->data; pe_resource_t *match = parent->fns->find_rsc(parent, id, NULL, pe_find_renamed); if (match != NULL) { if(!pcmk__str_eq(match->id, id, pcmk__str_casei)) { /* We found an instance of a clone instead */ match = uber_parent(match); crm_debug("Found %s for %s", match->id, id); } return match; } } crm_trace("No match for %s", id); return NULL; } static gboolean pe_find_constraint_tag(pe_working_set_t * data_set, const char * id, pe_tag_t ** tag) { gboolean rc = FALSE; *tag = NULL; rc = g_hash_table_lookup_extended(data_set->template_rsc_sets, id, NULL, (gpointer*) tag); if (rc == FALSE) { rc = g_hash_table_lookup_extended(data_set->tags, id, NULL, (gpointer*) tag); if (rc == FALSE) { crm_warn("No template or tag named '%s'", id); return FALSE; } else if (*tag == NULL) { crm_warn("No resource is tagged with '%s'", id); return FALSE; } } else if (*tag == NULL) { crm_warn("No resource is derived from template '%s'", id); return FALSE; } return rc; } static gboolean valid_resource_or_tag(pe_working_set_t * data_set, const char * id, pe_resource_t ** rsc, pe_tag_t ** tag) { gboolean rc = FALSE; if (rsc) { *rsc = NULL; *rsc = pe_find_constraint_resource(data_set->resources, id); if (*rsc) { return TRUE; } } if (tag) { *tag = NULL; rc = pe_find_constraint_tag(data_set, id, tag); } return rc; } static gboolean order_is_symmetrical(xmlNode * xml_obj, enum pe_order_kind parent_kind, const char * parent_symmetrical_s) { const char *id = crm_element_value(xml_obj, XML_ATTR_ID); const char *kind_s = crm_element_value(xml_obj, XML_ORDER_ATTR_KIND); const char *score_s = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE); const char *symmetrical_s = crm_element_value(xml_obj, XML_CONS_ATTR_SYMMETRICAL); enum pe_order_kind kind = parent_kind; if (kind_s || score_s) { kind = get_ordering_type(xml_obj); } if (symmetrical_s == NULL) { symmetrical_s = parent_symmetrical_s; } if (symmetrical_s) { gboolean symmetrical = crm_is_true(symmetrical_s); if (symmetrical && kind == pe_order_kind_serialize) { pcmk__config_warn("Ignoring " XML_CONS_ATTR_SYMMETRICAL " for '%s' because not valid with " XML_ORDER_ATTR_KIND " of 'Serialize'", id); return FALSE; } return symmetrical; } else { if (kind == pe_order_kind_serialize) { return FALSE; } else { return TRUE; } } } static gboolean unpack_simple_rsc_order(xmlNode * xml_obj, pe_working_set_t * data_set) { int order_id = 0; pe_resource_t *rsc_then = NULL; pe_resource_t *rsc_first = NULL; gboolean invert_bool = TRUE; int min_required_before = 0; enum pe_order_kind kind = pe_order_kind_mandatory; enum pe_ordering cons_weight = pe_order_optional; const char *id_first = NULL; const char *id_then = NULL; const char *action_then = NULL; const char *action_first = NULL; const char *instance_then = NULL; const char *instance_first = NULL; const char *id = NULL; CRM_CHECK(xml_obj != NULL, return FALSE); id = crm_element_value(xml_obj, XML_ATTR_ID); if (id == NULL) { pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID, crm_element_name(xml_obj)); return FALSE; } invert_bool = order_is_symmetrical(xml_obj, kind, NULL); id_then = crm_element_value(xml_obj, XML_ORDER_ATTR_THEN); id_first = crm_element_value(xml_obj, XML_ORDER_ATTR_FIRST); action_then = crm_element_value(xml_obj, XML_ORDER_ATTR_THEN_ACTION); action_first = crm_element_value(xml_obj, XML_ORDER_ATTR_FIRST_ACTION); instance_then = crm_element_value(xml_obj, XML_ORDER_ATTR_THEN_INSTANCE); instance_first = crm_element_value(xml_obj, XML_ORDER_ATTR_FIRST_INSTANCE); if (action_first == NULL) { action_first = RSC_START; } if (action_then == NULL) { action_then = action_first; } if (id_first == NULL) { pcmk__config_err("Ignoring constraint '%s' without " XML_ORDER_ATTR_FIRST, id); return FALSE; } if (id_then == NULL) { pcmk__config_err("Ignoring constraint '%s' without " XML_ORDER_ATTR_THEN, id); return FALSE; } rsc_then = pe_find_constraint_resource(data_set->resources, id_then); rsc_first = pe_find_constraint_resource(data_set->resources, id_first); if (rsc_then == NULL) { pcmk__config_err("Ignoring constraint '%s' because resource '%s' " "does not exist", id, id_then); return FALSE; } else if (rsc_first == NULL) { pcmk__config_err("Ignoring constraint '%s' because resource '%s' " "does not exist", id, id_first); return FALSE; } else if (instance_then && pe_rsc_is_clone(rsc_then) == FALSE) { pcmk__config_err("Ignoring constraint '%s' because resource '%s' " "is not a clone but instance '%s' was requested", id, id_then, instance_then); return FALSE; } else if (instance_first && pe_rsc_is_clone(rsc_first) == FALSE) { pcmk__config_err("Ignoring constraint '%s' because resource '%s' " "is not a clone but instance '%s' was requested", id, id_first, instance_first); return FALSE; } if (instance_then) { rsc_then = find_clone_instance(rsc_then, instance_then, data_set); if (rsc_then == NULL) { pcmk__config_warn("Ignoring constraint '%s' because resource '%s' " "does not have an instance '%s'", id, id_then, instance_then); return FALSE; } } if (instance_first) { rsc_first = find_clone_instance(rsc_first, instance_first, data_set); if (rsc_first == NULL) { pcmk__config_warn("Ignoring constraint '%s' because resource '%s' " "does not have an instance '%s'", "'%s'", id, id_first, instance_first); return FALSE; } } cons_weight = pe_order_optional; kind = get_ordering_type(xml_obj); if (kind == pe_order_kind_optional && rsc_then->restart_type == pe_restart_restart) { crm_trace("Upgrade : recovery - implies right"); pe__set_order_flags(cons_weight, pe_order_implies_then); } if (invert_bool == FALSE) { pe__set_order_flags(cons_weight, get_asymmetrical_flags(kind)); } else { pe__set_order_flags(cons_weight, get_flags(id, kind, action_first, action_then, FALSE)); } if (pe_rsc_is_clone(rsc_first)) { /* If clone-min is set, require at least that number of instances to be * runnable before allowing dependencies to be runnable. */ const char *min_clones_s = g_hash_table_lookup(rsc_first->meta, XML_RSC_ATTR_INCARNATION_MIN); // @COMPAT 1.1.13: deprecated const char *require_all_s = crm_element_value(xml_obj, "require-all"); if (min_clones_s) { min_required_before = crm_parse_int(min_clones_s, "0"); } else if (require_all_s) { pe_warn_once(pe_wo_require_all, "Support for require-all in ordering constraints " "is deprecated and will be removed in a future release" " (use clone-min clone meta-attribute instead)"); if (crm_is_true(require_all_s) == FALSE) { // require-all=false is deprecated equivalent of clone-min=1 min_required_before = 1; } } } /* If there is a minimum number of instances that must be runnable before * the 'then' action is runnable, we use a pseudo action as an intermediate step * start min number of clones -> pseudo action is runnable -> dependency runnable. */ if (min_required_before) { GListPtr rIter = NULL; char *task = crm_strdup_printf(CRM_OP_RELAXED_CLONE ":%s", id); pe_action_t *unordered_action = get_pseudo_op(task, data_set); free(task); /* require the pseudo action to have "min_required_before" number of * actions to be considered runnable before allowing the pseudo action * to be runnable. */ unordered_action->required_runnable_before = min_required_before; update_action_flags(unordered_action, pe_action_requires_any, __func__, __LINE__); for (rIter = rsc_first->children; id && rIter; rIter = rIter->next) { pe_resource_t *child = rIter->data; /* order each clone instance before the pseudo action */ custom_action_order(child, pcmk__op_key(child->id, action_first, 0), NULL, NULL, NULL, unordered_action, pe_order_one_or_more|pe_order_implies_then_printed, data_set); } /* order the "then" dependency to occur after the pseudo action only if * the pseudo action is runnable */ order_id = custom_action_order(NULL, NULL, unordered_action, rsc_then, pcmk__op_key(rsc_then->id, action_then, 0), NULL, cons_weight|pe_order_runnable_left, data_set); } else { order_id = new_rsc_order(rsc_first, action_first, rsc_then, action_then, cons_weight, data_set); } pe_rsc_trace(rsc_first, "order-%d (%s): %s_%s before %s_%s flags=0x%.6x", order_id, id, rsc_first->id, action_first, rsc_then->id, action_then, cons_weight); if (invert_bool == FALSE) { return TRUE; } action_then = invert_action(action_then); action_first = invert_action(action_first); if (action_then == NULL || action_first == NULL) { pcmk__config_err("Cannot invert constraint '%s' " "(please specify inverse manually)", id); return TRUE; } cons_weight = pe_order_optional; if (kind == pe_order_kind_optional && rsc_then->restart_type == pe_restart_restart) { crm_trace("Upgrade : recovery - implies left"); pe__set_order_flags(cons_weight, pe_order_implies_first); } pe__set_order_flags(cons_weight, get_flags(id, kind, action_first, action_then, TRUE)); order_id = new_rsc_order(rsc_then, action_then, rsc_first, action_first, cons_weight, data_set); pe_rsc_trace(rsc_then, "order-%d (%s): %s_%s before %s_%s flags=0x%.6x", order_id, id, rsc_then->id, action_then, rsc_first->id, action_first, cons_weight); return TRUE; } static gboolean expand_tags_in_sets(xmlNode * xml_obj, xmlNode ** expanded_xml, pe_working_set_t * data_set) { xmlNode *new_xml = NULL; xmlNode *set = NULL; gboolean any_refs = FALSE; const char *cons_id = NULL; *expanded_xml = NULL; CRM_CHECK(xml_obj != NULL, return FALSE); new_xml = copy_xml(xml_obj); cons_id = ID(new_xml); for (set = pcmk__xe_first_child(new_xml); set != NULL; set = pcmk__xe_next(set)) { xmlNode *xml_rsc = NULL; GListPtr tag_refs = NULL; GListPtr gIter = NULL; if (!pcmk__str_eq((const char *)set->name, XML_CONS_TAG_RSC_SET, pcmk__str_casei)) { continue; } for (xml_rsc = pcmk__xe_first_child(set); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { pe_resource_t *rsc = NULL; pe_tag_t *tag = NULL; const char *id = ID(xml_rsc); if (!pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_casei)) { continue; } if (valid_resource_or_tag(data_set, id, &rsc, &tag) == FALSE) { pcmk__config_err("Ignoring resource sets for constraint '%s' " "because '%s' is not a valid resource or tag", cons_id, id); free_xml(new_xml); return FALSE; } else if (rsc) { continue; } else if (tag) { /* The resource_ref under the resource_set references a template/tag */ xmlNode *last_ref = xml_rsc; /* A sample: Original XML: Now we are appending rsc2 and rsc3 which are tagged with tag1 right after it: */ for (gIter = tag->refs; gIter != NULL; gIter = gIter->next) { const char *obj_ref = (const char *) gIter->data; xmlNode *new_rsc_ref = NULL; new_rsc_ref = xmlNewDocRawNode(getDocPtr(set), NULL, (pcmkXmlStr) XML_TAG_RESOURCE_REF, NULL); crm_xml_add(new_rsc_ref, XML_ATTR_ID, obj_ref); xmlAddNextSibling(last_ref, new_rsc_ref); last_ref = new_rsc_ref; } any_refs = TRUE; /* Do not directly free '', which would * break the further pcmk__xe_next(xml_rsc)) and * cause "Invalid read" seen by valgrind. Just remember it for * freeing later. */ tag_refs = g_list_append(tag_refs, xml_rsc); } } /* Now free '', and finally get: */ for (gIter = tag_refs; gIter != NULL; gIter = gIter->next) { xmlNode *tag_ref = gIter->data; free_xml(tag_ref); } g_list_free(tag_refs); } if (any_refs) { *expanded_xml = new_xml; } else { free_xml(new_xml); } return TRUE; } static gboolean tag_to_set(xmlNode * xml_obj, xmlNode ** rsc_set, const char * attr, gboolean convert_rsc, pe_working_set_t * data_set) { const char *cons_id = NULL; const char *id = NULL; pe_resource_t *rsc = NULL; pe_tag_t *tag = NULL; *rsc_set = NULL; CRM_CHECK((xml_obj != NULL) && (attr != NULL), return FALSE); cons_id = ID(xml_obj); if (cons_id == NULL) { pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID, crm_element_name(xml_obj)); return FALSE; } id = crm_element_value(xml_obj, attr); if (id == NULL) { return TRUE; } if (valid_resource_or_tag(data_set, id, &rsc, &tag) == FALSE) { pcmk__config_err("Ignoring constraint '%s' because '%s' is not a " "valid resource or tag", cons_id, id); return FALSE; } else if (tag) { GListPtr gIter = NULL; /* A template/tag is referenced by the "attr" attribute (first, then, rsc or with-rsc). Add the template/tag's corresponding "resource_set" which contains the resources derived from it or tagged with it under the constraint. */ *rsc_set = create_xml_node(xml_obj, XML_CONS_TAG_RSC_SET); crm_xml_add(*rsc_set, XML_ATTR_ID, id); for (gIter = tag->refs; gIter != NULL; gIter = gIter->next) { const char *obj_ref = (const char *) gIter->data; xmlNode *rsc_ref = NULL; rsc_ref = create_xml_node(*rsc_set, XML_TAG_RESOURCE_REF); crm_xml_add(rsc_ref, XML_ATTR_ID, obj_ref); } /* Set sequential="false" for the resource_set */ crm_xml_add(*rsc_set, "sequential", XML_BOOLEAN_FALSE); } else if (rsc && convert_rsc) { /* Even a regular resource is referenced by "attr", convert it into a resource_set. Because the other side of the constraint could be a template/tag reference. */ xmlNode *rsc_ref = NULL; *rsc_set = create_xml_node(xml_obj, XML_CONS_TAG_RSC_SET); crm_xml_add(*rsc_set, XML_ATTR_ID, id); rsc_ref = create_xml_node(*rsc_set, XML_TAG_RESOURCE_REF); crm_xml_add(rsc_ref, XML_ATTR_ID, id); } else { return TRUE; } /* Remove the "attr" attribute referencing the template/tag */ if (*rsc_set) { xml_remove_prop(xml_obj, attr); } return TRUE; } static void unpack_rsc_location(xmlNode *xml_obj, pe_resource_t *rsc_lh, const char *role, const char *score, pe_working_set_t *data_set, pe_match_data_t *match_data); static void unpack_simple_location(xmlNode *xml_obj, pe_working_set_t *data_set) { const char *id = crm_element_value(xml_obj, XML_ATTR_ID); const char *value = crm_element_value(xml_obj, XML_LOC_ATTR_SOURCE); if(value) { pe_resource_t *rsc_lh = pe_find_constraint_resource(data_set->resources, value); unpack_rsc_location(xml_obj, rsc_lh, NULL, NULL, data_set, NULL); } value = crm_element_value(xml_obj, XML_LOC_ATTR_SOURCE_PATTERN); if(value) { regex_t *r_patt = calloc(1, sizeof(regex_t)); bool invert = FALSE; GListPtr rIter = NULL; if(value[0] == '!') { value++; invert = TRUE; } if (regcomp(r_patt, value, REG_EXTENDED)) { pcmk__config_err("Ignoring constraint '%s' because " XML_LOC_ATTR_SOURCE_PATTERN " has invalid value '%s'", id, value); regfree(r_patt); free(r_patt); return; } for (rIter = data_set->resources; rIter; rIter = rIter->next) { pe_resource_t *r = rIter->data; int nregs = 0; regmatch_t *pmatch = NULL; int status; if(r_patt->re_nsub > 0) { nregs = r_patt->re_nsub + 1; } else { nregs = 1; } pmatch = calloc(nregs, sizeof(regmatch_t)); status = regexec(r_patt, r->id, nregs, pmatch, 0); if(invert == FALSE && status == 0) { pe_re_match_data_t re_match_data = { .string = r->id, .nregs = nregs, .pmatch = pmatch }; pe_match_data_t match_data = { .re = &re_match_data, .params = r->parameters, .meta = r->meta, }; crm_debug("'%s' matched '%s' for %s", r->id, value, id); unpack_rsc_location(xml_obj, r, NULL, NULL, data_set, &match_data); } else if (invert && (status != 0)) { crm_debug("'%s' is an inverted match of '%s' for %s", r->id, value, id); unpack_rsc_location(xml_obj, r, NULL, NULL, data_set, NULL); } else { crm_trace("'%s' does not match '%s' for %s", r->id, value, id); } free(pmatch); } regfree(r_patt); free(r_patt); } } static void unpack_rsc_location(xmlNode *xml_obj, pe_resource_t *rsc_lh, const char *role, const char *score, pe_working_set_t *data_set, pe_match_data_t *match_data) { pe__location_t *location = NULL; const char *id_lh = crm_element_value(xml_obj, XML_LOC_ATTR_SOURCE); const char *id = crm_element_value(xml_obj, XML_ATTR_ID); const char *node = crm_element_value(xml_obj, XML_CIB_TAG_NODE); const char *discovery = crm_element_value(xml_obj, XML_LOCATION_ATTR_DISCOVERY); if (rsc_lh == NULL) { pcmk__config_warn("Ignoring constraint '%s' because resource '%s' " "does not exist", id, id_lh); return; } if (score == NULL) { score = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE); } if (node != NULL && score != NULL) { int score_i = char2score(score); pe_node_t *match = pe_find_node(data_set->nodes, node); if (!match) { return; } location = rsc2node_new(id, rsc_lh, score_i, discovery, match, data_set); } else { bool empty = TRUE; crm_time_t *next_change = crm_time_new_undefined(); /* This loop is logically parallel to pe_evaluate_rules(), except * instead of checking whether any rule is active, we set up location * constraints for each active rule. */ for (xmlNode *rule_xml = first_named_child(xml_obj, XML_TAG_RULE); rule_xml != NULL; rule_xml = crm_next_same_xml(rule_xml)) { empty = FALSE; crm_trace("Unpacking %s/%s", id, ID(rule_xml)); generate_location_rule(rsc_lh, rule_xml, discovery, next_change, data_set, match_data); } if (empty) { pcmk__config_err("Ignoring constraint '%s' because it contains " "no rules", id); } /* If there is a point in the future when the evaluation of a rule will * change, make sure the scheduler is re-run by that time. */ if (crm_time_is_defined(next_change)) { time_t t = (time_t) crm_time_get_seconds_since_epoch(next_change); pe__update_recheck_time(t, data_set); } crm_time_free(next_change); return; } if (role == NULL) { role = crm_element_value(xml_obj, XML_RULE_ATTR_ROLE); } if (location && role) { if (text2role(role) == RSC_ROLE_UNKNOWN) { pe_err("Invalid constraint %s: Bad role %s", id, role); return; } else { enum rsc_role_e r = text2role(role); switch(r) { case RSC_ROLE_UNKNOWN: case RSC_ROLE_STARTED: case RSC_ROLE_SLAVE: /* Applies to all */ location->role_filter = RSC_ROLE_UNKNOWN; break; default: location->role_filter = r; break; } } } } static gboolean unpack_location_tags(xmlNode * xml_obj, xmlNode ** expanded_xml, pe_working_set_t * data_set) { const char *id = NULL; const char *id_lh = NULL; const char *state_lh = NULL; pe_resource_t *rsc_lh = NULL; pe_tag_t *tag_lh = NULL; xmlNode *new_xml = NULL; xmlNode *rsc_set_lh = NULL; *expanded_xml = NULL; CRM_CHECK(xml_obj != NULL, return FALSE); id = ID(xml_obj); if (id == NULL) { pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID, crm_element_name(xml_obj)); return FALSE; } /* Attempt to expand any template/tag references in possible resource sets. */ expand_tags_in_sets(xml_obj, &new_xml, data_set); if (new_xml) { /* There are resource sets referencing templates. Return with the expanded XML. */ crm_log_xml_trace(new_xml, "Expanded rsc_location..."); *expanded_xml = new_xml; return TRUE; } id_lh = crm_element_value(xml_obj, XML_LOC_ATTR_SOURCE); if (id_lh == NULL) { return TRUE; } if (valid_resource_or_tag(data_set, id_lh, &rsc_lh, &tag_lh) == FALSE) { pcmk__config_err("Ignoring constraint '%s' because '%s' is not a " "valid resource or tag", id, id_lh); return FALSE; } else if (rsc_lh) { /* No template is referenced. */ return TRUE; } state_lh = crm_element_value(xml_obj, XML_RULE_ATTR_ROLE); new_xml = copy_xml(xml_obj); /* Convert the template/tag reference in "rsc" into a resource_set under the rsc_location constraint. */ if (tag_to_set(new_xml, &rsc_set_lh, XML_LOC_ATTR_SOURCE, FALSE, data_set) == FALSE) { free_xml(new_xml); return FALSE; } if (rsc_set_lh) { if (state_lh) { /* A "rsc-role" is specified. Move it into the converted resource_set as a "role"" attribute. */ crm_xml_add(rsc_set_lh, "role", state_lh); xml_remove_prop(new_xml, XML_RULE_ATTR_ROLE); } crm_log_xml_trace(new_xml, "Expanded rsc_location..."); *expanded_xml = new_xml; } else { /* No sets */ free_xml(new_xml); } return TRUE; } static gboolean unpack_location_set(xmlNode * location, xmlNode * set, pe_working_set_t * data_set) { xmlNode *xml_rsc = NULL; pe_resource_t *resource = NULL; const char *set_id; const char *role; const char *local_score; CRM_CHECK(set != NULL, return FALSE); set_id = ID(set); if (set_id == NULL) { pcmk__config_err("Ignoring " XML_CONS_TAG_RSC_SET " without " XML_ATTR_ID " in constraint '%s'", crm_str(ID(location))); return FALSE; } role = crm_element_value(set, "role"); local_score = crm_element_value(set, XML_RULE_ATTR_SCORE); for (xml_rsc = pcmk__xe_first_child(set); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { if (pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { EXPAND_CONSTRAINT_IDREF(set_id, resource, ID(xml_rsc)); unpack_rsc_location(location, resource, role, local_score, data_set, NULL); } } return TRUE; } static void unpack_location(xmlNode *xml_obj, pe_working_set_t *data_set) { xmlNode *set = NULL; gboolean any_sets = FALSE; xmlNode *orig_xml = NULL; xmlNode *expanded_xml = NULL; if (unpack_location_tags(xml_obj, &expanded_xml, data_set) == FALSE) { return; } if (expanded_xml) { orig_xml = xml_obj; xml_obj = expanded_xml; } for (set = pcmk__xe_first_child(xml_obj); set != NULL; set = pcmk__xe_next(set)) { if (pcmk__str_eq((const char *)set->name, XML_CONS_TAG_RSC_SET, pcmk__str_none)) { any_sets = TRUE; set = expand_idref(set, data_set->input); if (unpack_location_set(xml_obj, set, data_set) == FALSE) { if (expanded_xml) { free_xml(expanded_xml); } return; } } } if (expanded_xml) { free_xml(expanded_xml); xml_obj = orig_xml; } if (any_sets == FALSE) { unpack_simple_location(xml_obj, data_set); } } static int get_node_score(const char *rule, const char *score, gboolean raw, pe_node_t * node, pe_resource_t *rsc) { int score_f = 0; if (score == NULL) { pe_err("Rule %s: no score specified. Assuming 0.", rule); } else if (raw) { score_f = char2score(score); } else { const char *attr_score = pe_node_attribute_calculated(node, score, rsc); if (attr_score == NULL) { crm_debug("Rule %s: node %s did not have a value for %s", rule, node->details->uname, score); score_f = -INFINITY; } else { crm_debug("Rule %s: node %s had value %s for %s", rule, node->details->uname, attr_score, score); score_f = char2score(attr_score); } } return score_f; } static pe__location_t * generate_location_rule(pe_resource_t *rsc, xmlNode *rule_xml, const char *discovery, crm_time_t *next_change, pe_working_set_t *data_set, pe_match_data_t *match_data) { const char *rule_id = NULL; const char *score = NULL; const char *boolean = NULL; const char *role = NULL; GListPtr gIter = NULL; GListPtr match_L = NULL; gboolean do_and = TRUE; gboolean accept = TRUE; gboolean raw_score = TRUE; gboolean score_allocated = FALSE; pe__location_t *location_rule = NULL; rule_xml = expand_idref(rule_xml, data_set->input); rule_id = crm_element_value(rule_xml, XML_ATTR_ID); boolean = crm_element_value(rule_xml, XML_RULE_ATTR_BOOLEAN_OP); role = crm_element_value(rule_xml, XML_RULE_ATTR_ROLE); crm_trace("Processing rule: %s", rule_id); if (role != NULL && text2role(role) == RSC_ROLE_UNKNOWN) { pe_err("Bad role specified for %s: %s", rule_id, role); return NULL; } score = crm_element_value(rule_xml, XML_RULE_ATTR_SCORE); if (score == NULL) { score = crm_element_value(rule_xml, XML_RULE_ATTR_SCORE_ATTRIBUTE); if (score != NULL) { raw_score = FALSE; } } if (pcmk__str_eq(boolean, "or", pcmk__str_casei)) { do_and = FALSE; } location_rule = rsc2node_new(rule_id, rsc, 0, discovery, NULL, data_set); if (location_rule == NULL) { return NULL; } if (match_data && match_data->re && match_data->re->nregs > 0 && match_data->re->pmatch[0].rm_so != -1) { if (raw_score == FALSE) { char *result = pe_expand_re_matches(score, match_data->re); if (result) { score = (const char *) result; score_allocated = TRUE; } } } if (role != NULL) { crm_trace("Setting role filter: %s", role); location_rule->role_filter = text2role(role); if (location_rule->role_filter == RSC_ROLE_SLAVE) { /* Any promotable clone cannot be promoted without being a slave first * Ergo, any constraint for the slave role applies to every role */ location_rule->role_filter = RSC_ROLE_UNKNOWN; } } if (do_and) { GListPtr gIter = NULL; match_L = pcmk__copy_node_list(data_set->nodes, true); for (gIter = match_L; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; node->weight = get_node_score(rule_id, score, raw_score, node, rsc); } } for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { int score_f = 0; pe_node_t *node = (pe_node_t *) gIter->data; accept = pe_test_rule(rule_xml, node->details->attrs, RSC_ROLE_UNKNOWN, data_set->now, next_change, match_data); crm_trace("Rule %s %s on %s", ID(rule_xml), accept ? "passed" : "failed", node->details->uname); score_f = get_node_score(rule_id, score, raw_score, node, rsc); /* if(accept && score_f == -INFINITY) { */ /* accept = FALSE; */ /* } */ if (accept) { pe_node_t *local = pe_find_node_id(match_L, node->details->id); if (local == NULL && do_and) { continue; } else if (local == NULL) { local = pe__copy_node(node); match_L = g_list_append(match_L, local); } if (do_and == FALSE) { local->weight = pe__add_scores(local->weight, score_f); } crm_trace("node %s now has weight %d", node->details->uname, local->weight); } else if (do_and && !accept) { /* remove it */ pe_node_t *delete = pe_find_node_id(match_L, node->details->id); if (delete != NULL) { match_L = g_list_remove(match_L, delete); crm_trace("node %s did not match", node->details->uname); } free(delete); } } if (score_allocated == TRUE) { free((char *)score); } location_rule->node_list_rh = match_L; if (location_rule->node_list_rh == NULL) { crm_trace("No matching nodes for rule %s", rule_id); return NULL; } crm_trace("%s: %d nodes matched", rule_id, g_list_length(location_rule->node_list_rh)); return location_rule; } static gint sort_cons_priority_lh(gconstpointer a, gconstpointer b) { - const rsc_colocation_t *rsc_constraint1 = (const rsc_colocation_t *)a; - const rsc_colocation_t *rsc_constraint2 = (const rsc_colocation_t *)b; + const pcmk__colocation_t *rsc_constraint1 = (const pcmk__colocation_t *) a; + const pcmk__colocation_t *rsc_constraint2 = (const pcmk__colocation_t *) b; if (a == NULL) { return 1; } if (b == NULL) { return -1; } CRM_ASSERT(rsc_constraint1->rsc_lh != NULL); CRM_ASSERT(rsc_constraint1->rsc_rh != NULL); if (rsc_constraint1->rsc_lh->priority > rsc_constraint2->rsc_lh->priority) { return -1; } if (rsc_constraint1->rsc_lh->priority < rsc_constraint2->rsc_lh->priority) { return 1; } /* Process clones before primitives and groups */ if (rsc_constraint1->rsc_lh->variant > rsc_constraint2->rsc_lh->variant) { return -1; } else if (rsc_constraint1->rsc_lh->variant < rsc_constraint2->rsc_lh->variant) { return 1; } /* @COMPAT scheduler <2.0.0: Process promotable clones before nonpromotable * clones (probably unnecessary, but avoids having to update regression * tests) */ if (rsc_constraint1->rsc_lh->variant == pe_clone) { if (pcmk_is_set(rsc_constraint1->rsc_lh->flags, pe_rsc_promotable) && !pcmk_is_set(rsc_constraint2->rsc_lh->flags, pe_rsc_promotable)) { return -1; } else if (!pcmk_is_set(rsc_constraint1->rsc_lh->flags, pe_rsc_promotable) && pcmk_is_set(rsc_constraint2->rsc_lh->flags, pe_rsc_promotable)) { return 1; } } return strcmp(rsc_constraint1->rsc_lh->id, rsc_constraint2->rsc_lh->id); } static gint sort_cons_priority_rh(gconstpointer a, gconstpointer b) { - const rsc_colocation_t *rsc_constraint1 = (const rsc_colocation_t *)a; - const rsc_colocation_t *rsc_constraint2 = (const rsc_colocation_t *)b; + const pcmk__colocation_t *rsc_constraint1 = (const pcmk__colocation_t *) a; + const pcmk__colocation_t *rsc_constraint2 = (const pcmk__colocation_t *) b; if (a == NULL) { return 1; } if (b == NULL) { return -1; } CRM_ASSERT(rsc_constraint1->rsc_lh != NULL); CRM_ASSERT(rsc_constraint1->rsc_rh != NULL); if (rsc_constraint1->rsc_rh->priority > rsc_constraint2->rsc_rh->priority) { return -1; } if (rsc_constraint1->rsc_rh->priority < rsc_constraint2->rsc_rh->priority) { return 1; } /* Process clones before primitives and groups */ if (rsc_constraint1->rsc_rh->variant > rsc_constraint2->rsc_rh->variant) { return -1; } else if (rsc_constraint1->rsc_rh->variant < rsc_constraint2->rsc_rh->variant) { return 1; } /* @COMPAT scheduler <2.0.0: Process promotable clones before nonpromotable * clones (probably unnecessary, but avoids having to update regression * tests) */ if (rsc_constraint1->rsc_rh->variant == pe_clone) { if (pcmk_is_set(rsc_constraint1->rsc_rh->flags, pe_rsc_promotable) && !pcmk_is_set(rsc_constraint2->rsc_rh->flags, pe_rsc_promotable)) { return -1; } else if (!pcmk_is_set(rsc_constraint1->rsc_rh->flags, pe_rsc_promotable) && pcmk_is_set(rsc_constraint2->rsc_rh->flags, pe_rsc_promotable)) { return 1; } } return strcmp(rsc_constraint1->rsc_rh->id, rsc_constraint2->rsc_rh->id); } static void anti_colocation_order(pe_resource_t * first_rsc, int first_role, pe_resource_t * then_rsc, int then_role, pe_working_set_t * data_set) { const char *first_tasks[] = { NULL, NULL }; const char *then_tasks[] = { NULL, NULL }; int first_lpc = 0; int then_lpc = 0; /* Actions to make first_rsc lose first_role */ if (first_role == RSC_ROLE_MASTER) { first_tasks[0] = CRMD_ACTION_DEMOTE; } else { first_tasks[0] = CRMD_ACTION_STOP; if (first_role == RSC_ROLE_SLAVE) { first_tasks[1] = CRMD_ACTION_PROMOTE; } } /* Actions to make then_rsc gain then_role */ if (then_role == RSC_ROLE_MASTER) { then_tasks[0] = CRMD_ACTION_PROMOTE; } else { then_tasks[0] = CRMD_ACTION_START; if (then_role == RSC_ROLE_SLAVE) { then_tasks[1] = CRMD_ACTION_DEMOTE; } } for (first_lpc = 0; first_lpc <= 1 && first_tasks[first_lpc] != NULL; first_lpc++) { for (then_lpc = 0; then_lpc <= 1 && then_tasks[then_lpc] != NULL; then_lpc++) { new_rsc_order(first_rsc, first_tasks[first_lpc], then_rsc, then_tasks[then_lpc], pe_order_anti_colocation, data_set); } } } -gboolean -rsc_colocation_new(const char *id, const char *node_attr, int score, - pe_resource_t * rsc_lh, pe_resource_t * rsc_rh, - const char *state_lh, const char *state_rh, pe_working_set_t * data_set) +void +pcmk__new_colocation(const char *id, const char *node_attr, int score, + pe_resource_t *rsc_lh, pe_resource_t *rsc_rh, + const char *state_lh, const char *state_rh, + pe_working_set_t *data_set) { - rsc_colocation_t *new_con = NULL; + pcmk__colocation_t *new_con = NULL; if ((rsc_lh == NULL) || (rsc_rh == NULL)) { pcmk__config_err("Ignoring colocation '%s' because resource " "does not exist", id); - return FALSE; + return; } - new_con = calloc(1, sizeof(rsc_colocation_t)); + new_con = calloc(1, sizeof(pcmk__colocation_t)); if (new_con == NULL) { - return FALSE; + return; } if (pcmk__str_eq(state_lh, RSC_ROLE_STARTED_S, pcmk__str_null_matches | pcmk__str_casei)) { state_lh = RSC_ROLE_UNKNOWN_S; } if (pcmk__str_eq(state_rh, RSC_ROLE_STARTED_S, pcmk__str_null_matches | pcmk__str_casei)) { state_rh = RSC_ROLE_UNKNOWN_S; } new_con->id = id; new_con->rsc_lh = rsc_lh; new_con->rsc_rh = rsc_rh; new_con->score = score; new_con->role_lh = text2role(state_lh); new_con->role_rh = text2role(state_rh); new_con->node_attribute = node_attr; if (node_attr == NULL) { node_attr = CRM_ATTR_UNAME; } pe_rsc_trace(rsc_lh, "%s ==> %s (%s %d)", rsc_lh->id, rsc_rh->id, node_attr, score); rsc_lh->rsc_cons = g_list_insert_sorted(rsc_lh->rsc_cons, new_con, sort_cons_priority_rh); rsc_rh->rsc_cons_lhs = g_list_insert_sorted(rsc_rh->rsc_cons_lhs, new_con, sort_cons_priority_lh); data_set->colocation_constraints = g_list_append(data_set->colocation_constraints, new_con); if (score <= -INFINITY) { anti_colocation_order(rsc_lh, new_con->role_lh, rsc_rh, new_con->role_rh, data_set); anti_colocation_order(rsc_rh, new_con->role_rh, rsc_lh, new_con->role_lh, data_set); } - - return TRUE; } /* LHS before RHS */ int new_rsc_order(pe_resource_t * lh_rsc, const char *lh_task, pe_resource_t * rh_rsc, const char *rh_task, enum pe_ordering type, pe_working_set_t * data_set) { char *lh_key = NULL; char *rh_key = NULL; CRM_CHECK(lh_rsc != NULL, return -1); CRM_CHECK(lh_task != NULL, return -1); CRM_CHECK(rh_rsc != NULL, return -1); CRM_CHECK(rh_task != NULL, return -1); #if 0 /* We do not need to test if these reference stonith resources * because the fencer has access to them even when they're not "running" */ if (validate_order_resources(lh_rsc, lh_task, rh_rsc, rh_task)) { return -1; } #endif lh_key = pcmk__op_key(lh_rsc->id, lh_task, 0); rh_key = pcmk__op_key(rh_rsc->id, rh_task, 0); return custom_action_order(lh_rsc, lh_key, NULL, rh_rsc, rh_key, NULL, type, data_set); } static char * task_from_action_or_key(pe_action_t *action, const char *key) { char *res = NULL; if (action) { res = strdup(action->task); } else if (key) { parse_op_key(key, NULL, &res, NULL); } return res; } /* when order constraints are made between two resources start and stop actions * those constraints have to be mirrored against the corresponding * migration actions to ensure start/stop ordering is preserved during * a migration */ static void handle_migration_ordering(pe__ordering_t *order, pe_working_set_t *data_set) { char *lh_task = NULL; char *rh_task = NULL; gboolean rh_migratable; gboolean lh_migratable; if (order->lh_rsc == NULL || order->rh_rsc == NULL) { return; } else if (order->lh_rsc == order->rh_rsc) { return; /* don't mess with those constraints built between parent * resources and the children */ } else if (is_parent(order->lh_rsc, order->rh_rsc)) { return; } else if (is_parent(order->rh_rsc, order->lh_rsc)) { return; } lh_migratable = pcmk_is_set(order->lh_rsc->flags, pe_rsc_allow_migrate); rh_migratable = pcmk_is_set(order->rh_rsc->flags, pe_rsc_allow_migrate); /* one of them has to be migratable for * the migrate ordering logic to be applied */ if (lh_migratable == FALSE && rh_migratable == FALSE) { return; } /* at this point we have two resources which allow migrations that have an * order dependency set between them. If those order dependencies involve * start/stop actions, we need to mirror the corresponding migrate actions * so order will be preserved. */ lh_task = task_from_action_or_key(order->lh_action, order->lh_action_task); rh_task = task_from_action_or_key(order->rh_action, order->rh_action_task); if (lh_task == NULL || rh_task == NULL) { goto cleanup_order; } if (pcmk__str_eq(lh_task, RSC_START, pcmk__str_casei) && pcmk__str_eq(rh_task, RSC_START, pcmk__str_casei)) { int flags = pe_order_optional; if (lh_migratable && rh_migratable) { /* A start then B start * A migrate_from then B migrate_to */ custom_action_order(order->lh_rsc, pcmk__op_key(order->lh_rsc->id, RSC_MIGRATED, 0), NULL, order->rh_rsc, pcmk__op_key(order->rh_rsc->id, RSC_MIGRATE, 0), NULL, flags, data_set); } if (rh_migratable) { if (lh_migratable) { pe__set_order_flags(flags, pe_order_apply_first_non_migratable); } /* A start then B start * A start then B migrate_to... only if A start is not a part of a migration*/ custom_action_order(order->lh_rsc, pcmk__op_key(order->lh_rsc->id, RSC_START, 0), NULL, order->rh_rsc, pcmk__op_key(order->rh_rsc->id, RSC_MIGRATE, 0), NULL, flags, data_set); } } else if (rh_migratable == TRUE && pcmk__str_eq(lh_task, RSC_STOP, pcmk__str_casei) && pcmk__str_eq(rh_task, RSC_STOP, pcmk__str_casei)) { int flags = pe_order_optional; if (lh_migratable) { pe__set_order_flags(flags, pe_order_apply_first_non_migratable); } /* rh side is at the bottom of the stack during a stop. If we have a constraint * stop B then stop A, if B is migrating via stop/start, and A is migrating using migration actions, * we need to enforce that A's migrate_to action occurs after B's stop action. */ custom_action_order(order->lh_rsc, pcmk__op_key(order->lh_rsc->id, RSC_STOP, 0), NULL, order->rh_rsc, pcmk__op_key(order->rh_rsc->id, RSC_MIGRATE, 0), NULL, flags, data_set); /* We need to build the stop constraint against migrate_from as well * to account for partial migrations. */ if (order->rh_rsc->partial_migration_target) { custom_action_order(order->lh_rsc, pcmk__op_key(order->lh_rsc->id, RSC_STOP, 0), NULL, order->rh_rsc, pcmk__op_key(order->rh_rsc->id, RSC_MIGRATED, 0), NULL, flags, data_set); } } else if (pcmk__str_eq(lh_task, RSC_PROMOTE, pcmk__str_casei) && pcmk__str_eq(rh_task, RSC_START, pcmk__str_casei)) { int flags = pe_order_optional; if (rh_migratable) { /* A promote then B start * A promote then B migrate_to */ custom_action_order(order->lh_rsc, pcmk__op_key(order->lh_rsc->id, RSC_PROMOTE, 0), NULL, order->rh_rsc, pcmk__op_key(order->rh_rsc->id, RSC_MIGRATE, 0), NULL, flags, data_set); } } else if (pcmk__str_eq(lh_task, RSC_DEMOTE, pcmk__str_casei) && pcmk__str_eq(rh_task, RSC_STOP, pcmk__str_casei)) { int flags = pe_order_optional; if (rh_migratable) { /* A demote then B stop * A demote then B migrate_to */ custom_action_order(order->lh_rsc, pcmk__op_key(order->lh_rsc->id, RSC_DEMOTE, 0), NULL, order->rh_rsc, pcmk__op_key(order->rh_rsc->id, RSC_MIGRATE, 0), NULL, flags, data_set); /* We need to build the demote constraint against migrate_from as well * to account for partial migrations. */ if (order->rh_rsc->partial_migration_target) { custom_action_order(order->lh_rsc, pcmk__op_key(order->lh_rsc->id, RSC_DEMOTE, 0), NULL, order->rh_rsc, pcmk__op_key(order->rh_rsc->id, RSC_MIGRATED, 0), NULL, flags, data_set); } } } cleanup_order: free(lh_task); free(rh_task); } /* LHS before RHS */ int custom_action_order(pe_resource_t * lh_rsc, char *lh_action_task, pe_action_t * lh_action, pe_resource_t * rh_rsc, char *rh_action_task, pe_action_t * rh_action, enum pe_ordering type, pe_working_set_t * data_set) { pe__ordering_t *order = NULL; if (lh_rsc == NULL && lh_action) { lh_rsc = lh_action->rsc; } if (rh_rsc == NULL && rh_action) { rh_rsc = rh_action->rsc; } if ((lh_action == NULL && lh_rsc == NULL) || (rh_action == NULL && rh_rsc == NULL)) { crm_err("Invalid ordering (bug?)"); free(lh_action_task); free(rh_action_task); return -1; } order = calloc(1, sizeof(pe__ordering_t)); crm_trace("Creating[%d] %s %s %s - %s %s %s", data_set->order_id, lh_rsc?lh_rsc->id:"NA", lh_action_task?lh_action_task:"NA", lh_action?lh_action->uuid:"NA", rh_rsc?rh_rsc->id:"NA", rh_action_task?rh_action_task:"NA", rh_action?rh_action->uuid:"NA"); /* CRM_ASSERT(data_set->order_id != 291); */ order->id = data_set->order_id++; order->type = type; order->lh_rsc = lh_rsc; order->rh_rsc = rh_rsc; order->lh_action = lh_action; order->rh_action = rh_action; order->lh_action_task = lh_action_task; order->rh_action_task = rh_action_task; if (order->lh_action_task == NULL && lh_action) { order->lh_action_task = strdup(lh_action->uuid); } if (order->rh_action_task == NULL && rh_action) { order->rh_action_task = strdup(rh_action->uuid); } if (order->lh_rsc == NULL && lh_action) { order->lh_rsc = lh_action->rsc; } if (order->rh_rsc == NULL && rh_action) { order->rh_rsc = rh_action->rsc; } data_set->ordering_constraints = g_list_prepend(data_set->ordering_constraints, order); handle_migration_ordering(order, data_set); return order->id; } enum pe_ordering get_asymmetrical_flags(enum pe_order_kind kind) { enum pe_ordering flags = pe_order_optional; if (kind == pe_order_kind_mandatory) { pe__set_order_flags(flags, pe_order_asymmetrical); } else if (kind == pe_order_kind_serialize) { pe__set_order_flags(flags, pe_order_serialize_only); } return flags; } enum pe_ordering get_flags(const char *id, enum pe_order_kind kind, const char *action_first, const char *action_then, gboolean invert) { enum pe_ordering flags = pe_order_optional; if (invert && kind == pe_order_kind_mandatory) { crm_trace("Upgrade %s: implies left", id); pe__set_order_flags(flags, pe_order_implies_first); } else if (kind == pe_order_kind_mandatory) { crm_trace("Upgrade %s: implies right", id); pe__set_order_flags(flags, pe_order_implies_then); if (pcmk__strcase_any_of(action_first, RSC_START, RSC_PROMOTE, NULL)) { crm_trace("Upgrade %s: runnable", id); pe__set_order_flags(flags, pe_order_runnable_left); } } else if (kind == pe_order_kind_serialize) { pe__set_order_flags(flags, pe_order_serialize_only); } return flags; } static gboolean unpack_order_set(xmlNode * set, enum pe_order_kind parent_kind, pe_resource_t ** rsc, pe_action_t ** begin, pe_action_t ** end, pe_action_t ** inv_begin, pe_action_t ** inv_end, const char *parent_symmetrical_s, pe_working_set_t * data_set) { xmlNode *xml_rsc = NULL; GListPtr set_iter = NULL; GListPtr resources = NULL; pe_resource_t *last = NULL; pe_resource_t *resource = NULL; int local_kind = parent_kind; gboolean sequential = FALSE; enum pe_ordering flags = pe_order_optional; gboolean symmetrical = TRUE; char *key = NULL; const char *id = ID(set); const char *action = crm_element_value(set, "action"); const char *sequential_s = crm_element_value(set, "sequential"); const char *kind_s = crm_element_value(set, XML_ORDER_ATTR_KIND); /* char *pseudo_id = NULL; char *end_id = NULL; char *begin_id = NULL; */ if (action == NULL) { action = RSC_START; } if (kind_s) { local_kind = get_ordering_type(set); } if (sequential_s == NULL) { sequential_s = "1"; } sequential = crm_is_true(sequential_s); symmetrical = order_is_symmetrical(set, parent_kind, parent_symmetrical_s); if (symmetrical) { flags = get_flags(id, local_kind, action, action, FALSE); } else { flags = get_asymmetrical_flags(local_kind); } for (xml_rsc = pcmk__xe_first_child(set); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { if (pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { EXPAND_CONSTRAINT_IDREF(id, resource, ID(xml_rsc)); resources = g_list_append(resources, resource); } } if (pcmk__list_of_1(resources)) { crm_trace("Single set: %s", id); *rsc = resource; *end = NULL; *begin = NULL; *inv_end = NULL; *inv_begin = NULL; goto done; } /* pseudo_id = crm_strdup_printf("%s-%s", id, action); end_id = crm_strdup_printf("%s-%s", pseudo_id, "end"); begin_id = crm_strdup_printf("%s-%s", pseudo_id, "begin"); */ *rsc = NULL; /* *end = get_pseudo_op(end_id, data_set); *begin = get_pseudo_op(begin_id, data_set); free(pseudo_id); free(begin_id); free(end_id); */ set_iter = resources; while (set_iter != NULL) { resource = (pe_resource_t *) set_iter->data; set_iter = set_iter->next; key = pcmk__op_key(resource->id, action, 0); /* custom_action_order(NULL, NULL, *begin, resource, strdup(key), NULL, flags|pe_order_implies_first_printed, data_set); custom_action_order(resource, strdup(key), NULL, NULL, NULL, *end, flags|pe_order_implies_then_printed, data_set); */ if (local_kind == pe_order_kind_serialize) { /* Serialize before everything that comes after */ GListPtr gIter = NULL; for (gIter = set_iter; gIter != NULL; gIter = gIter->next) { pe_resource_t *then_rsc = (pe_resource_t *) gIter->data; char *then_key = pcmk__op_key(then_rsc->id, action, 0); custom_action_order(resource, strdup(key), NULL, then_rsc, then_key, NULL, flags, data_set); } } else if (sequential) { if (last != NULL) { new_rsc_order(last, action, resource, action, flags, data_set); } last = resource; } free(key); } if (symmetrical == FALSE) { goto done; } last = NULL; action = invert_action(action); /* pseudo_id = crm_strdup_printf("%s-%s", id, action); end_id = crm_strdup_printf("%s-%s", pseudo_id, "end"); begin_id = crm_strdup_printf("%s-%s", pseudo_id, "begin"); *inv_end = get_pseudo_op(end_id, data_set); *inv_begin = get_pseudo_op(begin_id, data_set); free(pseudo_id); free(begin_id); free(end_id); */ flags = get_flags(id, local_kind, action, action, TRUE); set_iter = resources; while (set_iter != NULL) { resource = (pe_resource_t *) set_iter->data; set_iter = set_iter->next; /* key = pcmk__op_key(resource->id, action, 0); custom_action_order(NULL, NULL, *inv_begin, resource, strdup(key), NULL, flags|pe_order_implies_first_printed, data_set); custom_action_order(resource, key, NULL, NULL, NULL, *inv_end, flags|pe_order_implies_then_printed, data_set); */ if (sequential) { if (last != NULL) { new_rsc_order(resource, action, last, action, flags, data_set); } last = resource; } } done: g_list_free(resources); return TRUE; } static gboolean order_rsc_sets(const char *id, xmlNode * set1, xmlNode * set2, enum pe_order_kind kind, pe_working_set_t * data_set, gboolean invert, gboolean symmetrical) { xmlNode *xml_rsc = NULL; xmlNode *xml_rsc_2 = NULL; pe_resource_t *rsc_1 = NULL; pe_resource_t *rsc_2 = NULL; const char *action_1 = crm_element_value(set1, "action"); const char *action_2 = crm_element_value(set2, "action"); const char *sequential_1 = crm_element_value(set1, "sequential"); const char *sequential_2 = crm_element_value(set2, "sequential"); const char *require_all_s = crm_element_value(set1, "require-all"); gboolean require_all = require_all_s ? crm_is_true(require_all_s) : TRUE; enum pe_ordering flags = pe_order_none; if (action_1 == NULL) { action_1 = RSC_START; }; if (action_2 == NULL) { action_2 = RSC_START; }; if (invert) { action_1 = invert_action(action_1); action_2 = invert_action(action_2); } if(pcmk__str_eq(RSC_STOP, action_1, pcmk__str_casei) || pcmk__str_eq(RSC_DEMOTE, action_1, pcmk__str_casei)) { /* Assuming: A -> ( B || C) -> D * The one-or-more logic only applies during the start/promote phase * During shutdown neither B nor can shutdown until D is down, so simply turn require_all back on. */ require_all = TRUE; } if (symmetrical == FALSE) { flags = get_asymmetrical_flags(kind); } else { flags = get_flags(id, kind, action_2, action_1, invert); } /* If we have an un-ordered set1, whether it is sequential or not is irrelevant in regards to set2. */ if (!require_all) { char *task = crm_strdup_printf(CRM_OP_RELAXED_SET ":%s", ID(set1)); pe_action_t *unordered_action = get_pseudo_op(task, data_set); free(task); update_action_flags(unordered_action, pe_action_requires_any, __func__, __LINE__); for (xml_rsc = pcmk__xe_first_child(set1); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { if (!pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { continue; } EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc)); /* Add an ordering constraint between every element in set1 and the pseudo action. * If any action in set1 is runnable the pseudo action will be runnable. */ custom_action_order(rsc_1, pcmk__op_key(rsc_1->id, action_1, 0), NULL, NULL, NULL, unordered_action, pe_order_one_or_more|pe_order_implies_then_printed, data_set); } for (xml_rsc_2 = pcmk__xe_first_child(set2); xml_rsc_2 != NULL; xml_rsc_2 = pcmk__xe_next(xml_rsc_2)) { if (!pcmk__str_eq((const char *)xml_rsc_2->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { continue; } EXPAND_CONSTRAINT_IDREF(id, rsc_2, ID(xml_rsc_2)); /* Add an ordering constraint between the pseudo action and every element in set2. * If the pseudo action is runnable, every action in set2 will be runnable */ custom_action_order(NULL, NULL, unordered_action, rsc_2, pcmk__op_key(rsc_2->id, action_2, 0), NULL, flags|pe_order_runnable_left, data_set); } return TRUE; } if (crm_is_true(sequential_1)) { if (invert == FALSE) { /* get the last one */ const char *rid = NULL; for (xml_rsc = pcmk__xe_first_child(set1); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { if (pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { rid = ID(xml_rsc); } } EXPAND_CONSTRAINT_IDREF(id, rsc_1, rid); } else { /* get the first one */ for (xml_rsc = pcmk__xe_first_child(set1); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { if (pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc)); break; } } } } if (crm_is_true(sequential_2)) { if (invert == FALSE) { /* get the first one */ for (xml_rsc = pcmk__xe_first_child(set2); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { if (pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { EXPAND_CONSTRAINT_IDREF(id, rsc_2, ID(xml_rsc)); break; } } } else { /* get the last one */ const char *rid = NULL; for (xml_rsc = pcmk__xe_first_child(set2); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { if (pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { rid = ID(xml_rsc); } } EXPAND_CONSTRAINT_IDREF(id, rsc_2, rid); } } if (rsc_1 != NULL && rsc_2 != NULL) { new_rsc_order(rsc_1, action_1, rsc_2, action_2, flags, data_set); } else if (rsc_1 != NULL) { for (xml_rsc = pcmk__xe_first_child(set2); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { if (pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { EXPAND_CONSTRAINT_IDREF(id, rsc_2, ID(xml_rsc)); new_rsc_order(rsc_1, action_1, rsc_2, action_2, flags, data_set); } } } else if (rsc_2 != NULL) { xmlNode *xml_rsc = NULL; for (xml_rsc = pcmk__xe_first_child(set1); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { if (pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc)); new_rsc_order(rsc_1, action_1, rsc_2, action_2, flags, data_set); } } } else { for (xml_rsc = pcmk__xe_first_child(set1); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { if (pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { xmlNode *xml_rsc_2 = NULL; EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc)); for (xml_rsc_2 = pcmk__xe_first_child(set2); xml_rsc_2 != NULL; xml_rsc_2 = pcmk__xe_next(xml_rsc_2)) { if (pcmk__str_eq((const char *)xml_rsc_2->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { EXPAND_CONSTRAINT_IDREF(id, rsc_2, ID(xml_rsc_2)); new_rsc_order(rsc_1, action_1, rsc_2, action_2, flags, data_set); } } } } } return TRUE; } static gboolean unpack_order_tags(xmlNode * xml_obj, xmlNode ** expanded_xml, pe_working_set_t * data_set) { const char *id = NULL; const char *id_first = NULL; const char *id_then = NULL; const char *action_first = NULL; const char *action_then = NULL; pe_resource_t *rsc_first = NULL; pe_resource_t *rsc_then = NULL; pe_tag_t *tag_first = NULL; pe_tag_t *tag_then = NULL; xmlNode *new_xml = NULL; xmlNode *rsc_set_first = NULL; xmlNode *rsc_set_then = NULL; gboolean any_sets = FALSE; *expanded_xml = NULL; CRM_CHECK(xml_obj != NULL, return FALSE); id = ID(xml_obj); if (id == NULL) { pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID, crm_element_name(xml_obj)); return FALSE; } /* Attempt to expand any template/tag references in possible resource sets. */ expand_tags_in_sets(xml_obj, &new_xml, data_set); if (new_xml) { /* There are resource sets referencing templates/tags. Return with the expanded XML. */ crm_log_xml_trace(new_xml, "Expanded rsc_order..."); *expanded_xml = new_xml; return TRUE; } id_first = crm_element_value(xml_obj, XML_ORDER_ATTR_FIRST); id_then = crm_element_value(xml_obj, XML_ORDER_ATTR_THEN); if (id_first == NULL || id_then == NULL) { return TRUE; } if (valid_resource_or_tag(data_set, id_first, &rsc_first, &tag_first) == FALSE) { pcmk__config_err("Ignoring constraint '%s' because '%s' is not a " "valid resource or tag", id, id_first); return FALSE; } if (valid_resource_or_tag(data_set, id_then, &rsc_then, &tag_then) == FALSE) { pcmk__config_err("Ignoring constraint '%s' because '%s' is not a " "valid resource or tag", id, id_then); return FALSE; } if (rsc_first && rsc_then) { /* Neither side references any template/tag. */ return TRUE; } action_first = crm_element_value(xml_obj, XML_ORDER_ATTR_FIRST_ACTION); action_then = crm_element_value(xml_obj, XML_ORDER_ATTR_THEN_ACTION); new_xml = copy_xml(xml_obj); /* Convert the template/tag reference in "first" into a resource_set under the order constraint. */ if (tag_to_set(new_xml, &rsc_set_first, XML_ORDER_ATTR_FIRST, TRUE, data_set) == FALSE) { free_xml(new_xml); return FALSE; } if (rsc_set_first) { if (action_first) { /* A "first-action" is specified. Move it into the converted resource_set as an "action" attribute. */ crm_xml_add(rsc_set_first, "action", action_first); xml_remove_prop(new_xml, XML_ORDER_ATTR_FIRST_ACTION); } any_sets = TRUE; } /* Convert the template/tag reference in "then" into a resource_set under the order constraint. */ if (tag_to_set(new_xml, &rsc_set_then, XML_ORDER_ATTR_THEN, TRUE, data_set) == FALSE) { free_xml(new_xml); return FALSE; } if (rsc_set_then) { if (action_then) { /* A "then-action" is specified. Move it into the converted resource_set as an "action" attribute. */ crm_xml_add(rsc_set_then, "action", action_then); xml_remove_prop(new_xml, XML_ORDER_ATTR_THEN_ACTION); } any_sets = TRUE; } if (any_sets) { crm_log_xml_trace(new_xml, "Expanded rsc_order..."); *expanded_xml = new_xml; } else { free_xml(new_xml); } return TRUE; } gboolean unpack_rsc_order(xmlNode * xml_obj, pe_working_set_t * data_set) { gboolean any_sets = FALSE; pe_resource_t *rsc = NULL; /* pe_resource_t *last_rsc = NULL; */ pe_action_t *set_end = NULL; pe_action_t *set_begin = NULL; pe_action_t *set_inv_end = NULL; pe_action_t *set_inv_begin = NULL; xmlNode *set = NULL; xmlNode *last = NULL; xmlNode *orig_xml = NULL; xmlNode *expanded_xml = NULL; /* pe_action_t *last_end = NULL; pe_action_t *last_begin = NULL; pe_action_t *last_inv_end = NULL; pe_action_t *last_inv_begin = NULL; */ const char *id = crm_element_value(xml_obj, XML_ATTR_ID); const char *invert = crm_element_value(xml_obj, XML_CONS_ATTR_SYMMETRICAL); enum pe_order_kind kind = get_ordering_type(xml_obj); gboolean invert_bool = order_is_symmetrical(xml_obj, kind, NULL); gboolean rc = TRUE; rc = unpack_order_tags(xml_obj, &expanded_xml, data_set); if (expanded_xml) { orig_xml = xml_obj; xml_obj = expanded_xml; } else if (rc == FALSE) { return FALSE; } for (set = pcmk__xe_first_child(xml_obj); set != NULL; set = pcmk__xe_next(set)) { if (pcmk__str_eq((const char *)set->name, XML_CONS_TAG_RSC_SET, pcmk__str_none)) { any_sets = TRUE; set = expand_idref(set, data_set->input); if (unpack_order_set(set, kind, &rsc, &set_begin, &set_end, &set_inv_begin, &set_inv_end, invert, data_set) == FALSE) { return FALSE; /* Expand orders in order_rsc_sets() instead of via pseudo actions. */ /* } else if(last) { const char *set_action = crm_element_value(set, "action"); const char *last_action = crm_element_value(last, "action"); enum pe_ordering flags = get_flags(id, kind, last_action, set_action, FALSE); if(!set_action) { set_action = RSC_START; } if(!last_action) { last_action = RSC_START; } if(rsc == NULL && last_rsc == NULL) { order_actions(last_end, set_begin, flags); } else { custom_action_order( last_rsc, null_or_opkey(last_rsc, last_action), last_end, rsc, null_or_opkey(rsc, set_action), set_begin, flags, data_set); } if(crm_is_true(invert)) { set_action = invert_action(set_action); last_action = invert_action(last_action); flags = get_flags(id, kind, last_action, set_action, TRUE); if(rsc == NULL && last_rsc == NULL) { order_actions(last_inv_begin, set_inv_end, flags); } else { custom_action_order( last_rsc, null_or_opkey(last_rsc, last_action), last_inv_begin, rsc, null_or_opkey(rsc, set_action), set_inv_end, flags, data_set); } } */ } else if ( /* never called -- Now call it for supporting clones in resource sets */ last) { if (order_rsc_sets(id, last, set, kind, data_set, FALSE, invert_bool) == FALSE) { return FALSE; } if (invert_bool && order_rsc_sets(id, set, last, kind, data_set, TRUE, invert_bool) == FALSE) { return FALSE; } } last = set; /* last_rsc = rsc; last_end = set_end; last_begin = set_begin; last_inv_end = set_inv_end; last_inv_begin = set_inv_begin; */ } } if (expanded_xml) { free_xml(expanded_xml); xml_obj = orig_xml; } if (any_sets == FALSE) { return unpack_simple_rsc_order(xml_obj, data_set); } return TRUE; } static gboolean unpack_colocation_set(xmlNode * set, int score, pe_working_set_t * data_set) { xmlNode *xml_rsc = NULL; pe_resource_t *with = NULL; pe_resource_t *resource = NULL; const char *set_id = ID(set); const char *role = crm_element_value(set, "role"); const char *sequential = crm_element_value(set, "sequential"); const char *ordering = crm_element_value(set, "ordering"); int local_score = score; const char *score_s = crm_element_value(set, XML_RULE_ATTR_SCORE); if (score_s) { local_score = char2score(score_s); } if(ordering == NULL) { ordering = "group"; } if (sequential != NULL && crm_is_true(sequential) == FALSE) { return TRUE; } else if ((local_score >= 0) && pcmk__str_eq(ordering, "group", pcmk__str_casei)) { for (xml_rsc = pcmk__xe_first_child(set); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { if (pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { EXPAND_CONSTRAINT_IDREF(set_id, resource, ID(xml_rsc)); if (with != NULL) { pe_rsc_trace(resource, "Colocating %s with %s", resource->id, with->id); - rsc_colocation_new(set_id, NULL, local_score, resource, with, role, role, - data_set); + pcmk__new_colocation(set_id, NULL, local_score, resource, + with, role, role, data_set); } with = resource; } } } else if (local_score >= 0) { pe_resource_t *last = NULL; for (xml_rsc = pcmk__xe_first_child(set); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { if (pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { EXPAND_CONSTRAINT_IDREF(set_id, resource, ID(xml_rsc)); if (last != NULL) { pe_rsc_trace(resource, "Colocating %s with %s", last->id, resource->id); - rsc_colocation_new(set_id, NULL, local_score, last, resource, role, role, - data_set); + pcmk__new_colocation(set_id, NULL, local_score, last, + resource, role, role, data_set); } last = resource; } } } else { /* Anti-colocating with every prior resource is * the only way to ensure the intuitive result * (i.e. that no one in the set can run with anyone else in the set) */ for (xml_rsc = pcmk__xe_first_child(set); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { if (pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { xmlNode *xml_rsc_with = NULL; EXPAND_CONSTRAINT_IDREF(set_id, resource, ID(xml_rsc)); for (xml_rsc_with = pcmk__xe_first_child(set); xml_rsc_with != NULL; xml_rsc_with = pcmk__xe_next(xml_rsc_with)) { if (pcmk__str_eq((const char *)xml_rsc_with->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { if (pcmk__str_eq(resource->id, ID(xml_rsc_with), pcmk__str_casei)) { break; } EXPAND_CONSTRAINT_IDREF(set_id, with, ID(xml_rsc_with)); pe_rsc_trace(resource, "Anti-Colocating %s with %s", resource->id, with->id); - rsc_colocation_new(set_id, NULL, local_score, resource, with, role, role, - data_set); + pcmk__new_colocation(set_id, NULL, local_score, + resource, with, role, role, + data_set); } } } } } return TRUE; } static gboolean colocate_rsc_sets(const char *id, xmlNode * set1, xmlNode * set2, int score, pe_working_set_t * data_set) { xmlNode *xml_rsc = NULL; pe_resource_t *rsc_1 = NULL; pe_resource_t *rsc_2 = NULL; const char *role_1 = crm_element_value(set1, "role"); const char *role_2 = crm_element_value(set2, "role"); const char *sequential_1 = crm_element_value(set1, "sequential"); const char *sequential_2 = crm_element_value(set2, "sequential"); if (sequential_1 == NULL || crm_is_true(sequential_1)) { /* get the first one */ for (xml_rsc = pcmk__xe_first_child(set1); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { if (pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc)); break; } } } if (sequential_2 == NULL || crm_is_true(sequential_2)) { /* get the last one */ const char *rid = NULL; for (xml_rsc = pcmk__xe_first_child(set2); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { if (pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { rid = ID(xml_rsc); } } EXPAND_CONSTRAINT_IDREF(id, rsc_2, rid); } if (rsc_1 != NULL && rsc_2 != NULL) { - rsc_colocation_new(id, NULL, score, rsc_1, rsc_2, role_1, role_2, data_set); + pcmk__new_colocation(id, NULL, score, rsc_1, rsc_2, role_1, role_2, + data_set); } else if (rsc_1 != NULL) { for (xml_rsc = pcmk__xe_first_child(set2); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { if (pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { EXPAND_CONSTRAINT_IDREF(id, rsc_2, ID(xml_rsc)); - rsc_colocation_new(id, NULL, score, rsc_1, rsc_2, role_1, role_2, data_set); + pcmk__new_colocation(id, NULL, score, rsc_1, rsc_2, role_1, + role_2, data_set); } } } else if (rsc_2 != NULL) { for (xml_rsc = pcmk__xe_first_child(set1); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { if (pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc)); - rsc_colocation_new(id, NULL, score, rsc_1, rsc_2, role_1, role_2, data_set); + pcmk__new_colocation(id, NULL, score, rsc_1, rsc_2, role_1, + role_2, data_set); } } } else { for (xml_rsc = pcmk__xe_first_child(set1); xml_rsc != NULL; xml_rsc = pcmk__xe_next(xml_rsc)) { if (pcmk__str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { xmlNode *xml_rsc_2 = NULL; EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc)); for (xml_rsc_2 = pcmk__xe_first_child(set2); xml_rsc_2 != NULL; xml_rsc_2 = pcmk__xe_next(xml_rsc_2)) { if (pcmk__str_eq((const char *)xml_rsc_2->name, XML_TAG_RESOURCE_REF, pcmk__str_none)) { EXPAND_CONSTRAINT_IDREF(id, rsc_2, ID(xml_rsc_2)); - rsc_colocation_new(id, NULL, score, rsc_1, rsc_2, role_1, role_2, data_set); + pcmk__new_colocation(id, NULL, score, rsc_1, rsc_2, + role_1, role_2, data_set); } } } } } return TRUE; } -static gboolean +static void unpack_simple_colocation(xmlNode * xml_obj, pe_working_set_t * data_set) { int score_i = 0; const char *id = crm_element_value(xml_obj, XML_ATTR_ID); const char *score = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE); const char *id_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE); const char *id_rh = crm_element_value(xml_obj, XML_COLOC_ATTR_TARGET); const char *state_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE_ROLE); const char *state_rh = crm_element_value(xml_obj, XML_COLOC_ATTR_TARGET_ROLE); const char *attr = crm_element_value(xml_obj, XML_COLOC_ATTR_NODE_ATTR); const char *symmetrical = crm_element_value(xml_obj, XML_CONS_ATTR_SYMMETRICAL); // experimental syntax from pacemaker-next (unlikely to be adopted as-is) const char *instance_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE_INSTANCE); const char *instance_rh = crm_element_value(xml_obj, XML_COLOC_ATTR_TARGET_INSTANCE); pe_resource_t *rsc_lh = pe_find_constraint_resource(data_set->resources, id_lh); pe_resource_t *rsc_rh = pe_find_constraint_resource(data_set->resources, id_rh); if (rsc_lh == NULL) { pcmk__config_err("Ignoring constraint '%s' because resource '%s' " "does not exist", id, id_lh); - return FALSE; + return; } else if (rsc_rh == NULL) { pcmk__config_err("Ignoring constraint '%s' because resource '%s' " "does not exist", id, id_rh); - return FALSE; + return; } else if (instance_lh && pe_rsc_is_clone(rsc_lh) == FALSE) { pcmk__config_err("Ignoring constraint '%s' because resource '%s' " "is not a clone but instance '%s' was requested", id, id_lh, instance_lh); - return FALSE; + return; } else if (instance_rh && pe_rsc_is_clone(rsc_rh) == FALSE) { pcmk__config_err("Ignoring constraint '%s' because resource '%s' " "is not a clone but instance '%s' was requested", id, id_rh, instance_rh); - return FALSE; + return; } if (instance_lh) { rsc_lh = find_clone_instance(rsc_lh, instance_lh, data_set); if (rsc_lh == NULL) { pcmk__config_warn("Ignoring constraint '%s' because resource '%s' " "does not have an instance '%s'", id, id_lh, instance_lh); - return FALSE; + return; } } if (instance_rh) { rsc_rh = find_clone_instance(rsc_rh, instance_rh, data_set); if (rsc_rh == NULL) { pcmk__config_warn("Ignoring constraint '%s' because resource '%s' " "does not have an instance '%s'", "'%s'", id, id_rh, instance_rh); - return FALSE; + return; } } if (crm_is_true(symmetrical)) { pcmk__config_warn("The colocation constraint '" XML_CONS_ATTR_SYMMETRICAL "' attribute has been removed"); } if (score) { score_i = char2score(score); } - rsc_colocation_new(id, attr, score_i, rsc_lh, rsc_rh, state_lh, state_rh, data_set); - return TRUE; + pcmk__new_colocation(id, attr, score_i, rsc_lh, rsc_rh, state_lh, state_rh, + data_set); } static gboolean unpack_colocation_tags(xmlNode * xml_obj, xmlNode ** expanded_xml, pe_working_set_t * data_set) { const char *id = NULL; const char *id_lh = NULL; const char *id_rh = NULL; const char *state_lh = NULL; const char *state_rh = NULL; pe_resource_t *rsc_lh = NULL; pe_resource_t *rsc_rh = NULL; pe_tag_t *tag_lh = NULL; pe_tag_t *tag_rh = NULL; xmlNode *new_xml = NULL; xmlNode *rsc_set_lh = NULL; xmlNode *rsc_set_rh = NULL; gboolean any_sets = FALSE; *expanded_xml = NULL; CRM_CHECK(xml_obj != NULL, return FALSE); id = ID(xml_obj); if (id == NULL) { pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID, crm_element_name(xml_obj)); return FALSE; } /* Attempt to expand any template/tag references in possible resource sets. */ expand_tags_in_sets(xml_obj, &new_xml, data_set); if (new_xml) { /* There are resource sets referencing templates/tags. Return with the expanded XML. */ crm_log_xml_trace(new_xml, "Expanded rsc_colocation..."); *expanded_xml = new_xml; return TRUE; } id_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE); id_rh = crm_element_value(xml_obj, XML_COLOC_ATTR_TARGET); if (id_lh == NULL || id_rh == NULL) { return TRUE; } if (valid_resource_or_tag(data_set, id_lh, &rsc_lh, &tag_lh) == FALSE) { pcmk__config_err("Ignoring constraint '%s' because '%s' is not a " "valid resource or tag", id, id_lh); return FALSE; } if (valid_resource_or_tag(data_set, id_rh, &rsc_rh, &tag_rh) == FALSE) { pcmk__config_err("Ignoring constraint '%s' because '%s' is not a " "valid resource or tag", id, id_rh); return FALSE; } if (rsc_lh && rsc_rh) { /* Neither side references any template/tag. */ return TRUE; } if (tag_lh && tag_rh) { /* A colocation constraint between two templates/tags makes no sense. */ pcmk__config_err("Ignoring constraint '%s' because two templates or " "tags cannot be colocated", id); return FALSE; } state_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE_ROLE); state_rh = crm_element_value(xml_obj, XML_COLOC_ATTR_TARGET_ROLE); new_xml = copy_xml(xml_obj); /* Convert the template/tag reference in "rsc" into a resource_set under the colocation constraint. */ if (tag_to_set(new_xml, &rsc_set_lh, XML_COLOC_ATTR_SOURCE, TRUE, data_set) == FALSE) { free_xml(new_xml); return FALSE; } if (rsc_set_lh) { if (state_lh) { /* A "rsc-role" is specified. Move it into the converted resource_set as a "role"" attribute. */ crm_xml_add(rsc_set_lh, "role", state_lh); xml_remove_prop(new_xml, XML_COLOC_ATTR_SOURCE_ROLE); } any_sets = TRUE; } /* Convert the template/tag reference in "with-rsc" into a resource_set under the colocation constraint. */ if (tag_to_set(new_xml, &rsc_set_rh, XML_COLOC_ATTR_TARGET, TRUE, data_set) == FALSE) { free_xml(new_xml); return FALSE; } if (rsc_set_rh) { if (state_rh) { /* A "with-rsc-role" is specified. Move it into the converted resource_set as a "role"" attribute. */ crm_xml_add(rsc_set_rh, "role", state_rh); xml_remove_prop(new_xml, XML_COLOC_ATTR_TARGET_ROLE); } any_sets = TRUE; } if (any_sets) { crm_log_xml_trace(new_xml, "Expanded rsc_colocation..."); *expanded_xml = new_xml; } else { free_xml(new_xml); } return TRUE; } -gboolean -unpack_rsc_colocation(xmlNode * xml_obj, pe_working_set_t * data_set) +static void +unpack_rsc_colocation(xmlNode *xml_obj, pe_working_set_t *data_set) { int score_i = 0; xmlNode *set = NULL; xmlNode *last = NULL; gboolean any_sets = FALSE; xmlNode *orig_xml = NULL; xmlNode *expanded_xml = NULL; const char *id = crm_element_value(xml_obj, XML_ATTR_ID); const char *score = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE); - gboolean rc = TRUE; - if (score) { score_i = char2score(score); } - rc = unpack_colocation_tags(xml_obj, &expanded_xml, data_set); + if (!unpack_colocation_tags(xml_obj, &expanded_xml, data_set)) { + return; + } if (expanded_xml) { orig_xml = xml_obj; xml_obj = expanded_xml; - - } else if (rc == FALSE) { - return FALSE; } for (set = pcmk__xe_first_child(xml_obj); set != NULL; set = pcmk__xe_next(set)) { if (pcmk__str_eq((const char *)set->name, XML_CONS_TAG_RSC_SET, pcmk__str_none)) { any_sets = TRUE; set = expand_idref(set, data_set->input); - if (unpack_colocation_set(set, score_i, data_set) == FALSE) { - return FALSE; - - } else if (last && colocate_rsc_sets(id, last, set, score_i, data_set) == FALSE) { - return FALSE; + if (!unpack_colocation_set(set, score_i, data_set)) { + return; + } + if (last && !colocate_rsc_sets(id, last, set, score_i, data_set)) { + return; } last = set; } } if (expanded_xml) { free_xml(expanded_xml); xml_obj = orig_xml; } - if (any_sets == FALSE) { - return unpack_simple_colocation(xml_obj, data_set); + if (!any_sets) { + unpack_simple_colocation(xml_obj, data_set); } - - return TRUE; } gboolean rsc_ticket_new(const char *id, pe_resource_t * rsc_lh, pe_ticket_t * ticket, const char *state_lh, const char *loss_policy, pe_working_set_t * data_set) { rsc_ticket_t *new_rsc_ticket = NULL; if (rsc_lh == NULL) { pcmk__config_err("Ignoring ticket '%s' because resource " "does not exist", id); return FALSE; } new_rsc_ticket = calloc(1, sizeof(rsc_ticket_t)); if (new_rsc_ticket == NULL) { return FALSE; } if (pcmk__str_eq(state_lh, RSC_ROLE_STARTED_S, pcmk__str_null_matches | pcmk__str_casei)) { state_lh = RSC_ROLE_UNKNOWN_S; } new_rsc_ticket->id = id; new_rsc_ticket->ticket = ticket; new_rsc_ticket->rsc_lh = rsc_lh; new_rsc_ticket->role_lh = text2role(state_lh); if (pcmk__str_eq(loss_policy, "fence", pcmk__str_casei)) { if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { new_rsc_ticket->loss_policy = loss_ticket_fence; } else { pcmk__config_err("Resetting '" XML_TICKET_ATTR_LOSS_POLICY "' for ticket '%s' to 'stop' " "because fencing is not configured", ticket->id); loss_policy = "stop"; } } if (new_rsc_ticket->loss_policy == loss_ticket_fence) { crm_debug("On loss of ticket '%s': Fence the nodes running %s (%s)", new_rsc_ticket->ticket->id, new_rsc_ticket->rsc_lh->id, role2text(new_rsc_ticket->role_lh)); } else if (pcmk__str_eq(loss_policy, "freeze", pcmk__str_casei)) { crm_debug("On loss of ticket '%s': Freeze %s (%s)", new_rsc_ticket->ticket->id, new_rsc_ticket->rsc_lh->id, role2text(new_rsc_ticket->role_lh)); new_rsc_ticket->loss_policy = loss_ticket_freeze; } else if (pcmk__str_eq(loss_policy, "demote", pcmk__str_casei)) { crm_debug("On loss of ticket '%s': Demote %s (%s)", new_rsc_ticket->ticket->id, new_rsc_ticket->rsc_lh->id, role2text(new_rsc_ticket->role_lh)); new_rsc_ticket->loss_policy = loss_ticket_demote; } else if (pcmk__str_eq(loss_policy, "stop", pcmk__str_casei)) { crm_debug("On loss of ticket '%s': Stop %s (%s)", new_rsc_ticket->ticket->id, new_rsc_ticket->rsc_lh->id, role2text(new_rsc_ticket->role_lh)); new_rsc_ticket->loss_policy = loss_ticket_stop; } else { if (new_rsc_ticket->role_lh == RSC_ROLE_MASTER) { crm_debug("On loss of ticket '%s': Default to demote %s (%s)", new_rsc_ticket->ticket->id, new_rsc_ticket->rsc_lh->id, role2text(new_rsc_ticket->role_lh)); new_rsc_ticket->loss_policy = loss_ticket_demote; } else { crm_debug("On loss of ticket '%s': Default to stop %s (%s)", new_rsc_ticket->ticket->id, new_rsc_ticket->rsc_lh->id, role2text(new_rsc_ticket->role_lh)); new_rsc_ticket->loss_policy = loss_ticket_stop; } } pe_rsc_trace(rsc_lh, "%s (%s) ==> %s", rsc_lh->id, role2text(new_rsc_ticket->role_lh), ticket->id); rsc_lh->rsc_tickets = g_list_append(rsc_lh->rsc_tickets, new_rsc_ticket); data_set->ticket_constraints = g_list_append(data_set->ticket_constraints, new_rsc_ticket); if (new_rsc_ticket->ticket->granted == FALSE || new_rsc_ticket->ticket->standby) { rsc_ticket_constraint(rsc_lh, new_rsc_ticket, data_set); } return TRUE; } static gboolean unpack_rsc_ticket_set(xmlNode * set, pe_ticket_t * ticket, const char *loss_policy, pe_working_set_t * data_set) { xmlNode *xml_rsc = NULL; pe_resource_t *resource = NULL; const char *set_id = NULL; const char *role = NULL; CRM_CHECK(set != NULL, return FALSE); CRM_CHECK(ticket != NULL, return FALSE); set_id = ID(set); if (set_id == NULL) { pcmk__config_err("Ignoring <" XML_CONS_TAG_RSC_SET "> without " XML_ATTR_ID); return FALSE; } role = crm_element_value(set, "role"); for (xml_rsc = first_named_child(set, XML_TAG_RESOURCE_REF); xml_rsc != NULL; xml_rsc = crm_next_same_xml(xml_rsc)) { EXPAND_CONSTRAINT_IDREF(set_id, resource, ID(xml_rsc)); pe_rsc_trace(resource, "Resource '%s' depends on ticket '%s'", resource->id, ticket->id); rsc_ticket_new(set_id, resource, ticket, role, loss_policy, data_set); } return TRUE; } static gboolean unpack_simple_rsc_ticket(xmlNode * xml_obj, pe_working_set_t * data_set) { const char *id = NULL; const char *ticket_str = crm_element_value(xml_obj, XML_TICKET_ATTR_TICKET); const char *loss_policy = crm_element_value(xml_obj, XML_TICKET_ATTR_LOSS_POLICY); pe_ticket_t *ticket = NULL; const char *id_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE); const char *state_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE_ROLE); // experimental syntax from pacemaker-next (unlikely to be adopted as-is) const char *instance_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE_INSTANCE); pe_resource_t *rsc_lh = NULL; CRM_CHECK(xml_obj != NULL, return FALSE); id = ID(xml_obj); if (id == NULL) { pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID, crm_element_name(xml_obj)); return FALSE; } if (ticket_str == NULL) { pcmk__config_err("Ignoring constraint '%s' without ticket specified", id); return FALSE; } else { ticket = g_hash_table_lookup(data_set->tickets, ticket_str); } if (ticket == NULL) { pcmk__config_err("Ignoring constraint '%s' because ticket '%s' " "does not exist", id, ticket_str); return FALSE; } if (id_lh == NULL) { pcmk__config_err("Ignoring constraint '%s' without resource", id); return FALSE; } else { rsc_lh = pe_find_constraint_resource(data_set->resources, id_lh); } if (rsc_lh == NULL) { pcmk__config_err("Ignoring constraint '%s' because resource '%s' " "does not exist", id, id_lh); return FALSE; } else if (instance_lh && pe_rsc_is_clone(rsc_lh) == FALSE) { pcmk__config_err("Ignoring constraint '%s' because resource '%s' " "is not a clone but instance '%s' was requested", id, id_lh, instance_lh); return FALSE; } if (instance_lh) { rsc_lh = find_clone_instance(rsc_lh, instance_lh, data_set); if (rsc_lh == NULL) { pcmk__config_warn("Ignoring constraint '%s' because resource '%s' " "does not have an instance '%s'", "'%s'", id, id_lh, instance_lh); return FALSE; } } rsc_ticket_new(id, rsc_lh, ticket, state_lh, loss_policy, data_set); return TRUE; } static gboolean unpack_rsc_ticket_tags(xmlNode * xml_obj, xmlNode ** expanded_xml, pe_working_set_t * data_set) { const char *id = NULL; const char *id_lh = NULL; const char *state_lh = NULL; pe_resource_t *rsc_lh = NULL; pe_tag_t *tag_lh = NULL; xmlNode *new_xml = NULL; xmlNode *rsc_set_lh = NULL; gboolean any_sets = FALSE; *expanded_xml = NULL; CRM_CHECK(xml_obj != NULL, return FALSE); id = ID(xml_obj); if (id == NULL) { pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID, crm_element_name(xml_obj)); return FALSE; } /* Attempt to expand any template/tag references in possible resource sets. */ expand_tags_in_sets(xml_obj, &new_xml, data_set); if (new_xml) { /* There are resource sets referencing templates/tags. Return with the expanded XML. */ crm_log_xml_trace(new_xml, "Expanded rsc_ticket..."); *expanded_xml = new_xml; return TRUE; } id_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE); if (id_lh == NULL) { return TRUE; } if (valid_resource_or_tag(data_set, id_lh, &rsc_lh, &tag_lh) == FALSE) { pcmk__config_err("Ignoring constraint '%s' because '%s' is not a " "valid resource or tag", id, id_lh); return FALSE; } else if (rsc_lh) { /* No template/tag is referenced. */ return TRUE; } state_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE_ROLE); new_xml = copy_xml(xml_obj); /* Convert the template/tag reference in "rsc" into a resource_set under the rsc_ticket constraint. */ if (tag_to_set(new_xml, &rsc_set_lh, XML_COLOC_ATTR_SOURCE, FALSE, data_set) == FALSE) { free_xml(new_xml); return FALSE; } if (rsc_set_lh) { if (state_lh) { /* A "rsc-role" is specified. Move it into the converted resource_set as a "role"" attribute. */ crm_xml_add(rsc_set_lh, "role", state_lh); xml_remove_prop(new_xml, XML_COLOC_ATTR_SOURCE_ROLE); } any_sets = TRUE; } if (any_sets) { crm_log_xml_trace(new_xml, "Expanded rsc_ticket..."); *expanded_xml = new_xml; } else { free_xml(new_xml); } return TRUE; } gboolean unpack_rsc_ticket(xmlNode * xml_obj, pe_working_set_t * data_set) { xmlNode *set = NULL; gboolean any_sets = FALSE; const char *id = NULL; const char *ticket_str = crm_element_value(xml_obj, XML_TICKET_ATTR_TICKET); const char *loss_policy = crm_element_value(xml_obj, XML_TICKET_ATTR_LOSS_POLICY); pe_ticket_t *ticket = NULL; xmlNode *orig_xml = NULL; xmlNode *expanded_xml = NULL; gboolean rc = TRUE; CRM_CHECK(xml_obj != NULL, return FALSE); id = ID(xml_obj); if (id == NULL) { pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID, crm_element_name(xml_obj)); return FALSE; } if (data_set->tickets == NULL) { data_set->tickets = g_hash_table_new_full(crm_str_hash, g_str_equal, free, destroy_ticket); } if (ticket_str == NULL) { pcmk__config_err("Ignoring constraint '%s' without ticket", id); return FALSE; } else { ticket = g_hash_table_lookup(data_set->tickets, ticket_str); } if (ticket == NULL) { ticket = ticket_new(ticket_str, data_set); if (ticket == NULL) { return FALSE; } } rc = unpack_rsc_ticket_tags(xml_obj, &expanded_xml, data_set); if (expanded_xml) { orig_xml = xml_obj; xml_obj = expanded_xml; } else if (rc == FALSE) { return FALSE; } for (set = pcmk__xe_first_child(xml_obj); set != NULL; set = pcmk__xe_next(set)) { if (pcmk__str_eq((const char *)set->name, XML_CONS_TAG_RSC_SET, pcmk__str_none)) { any_sets = TRUE; set = expand_idref(set, data_set->input); if (unpack_rsc_ticket_set(set, ticket, loss_policy, data_set) == FALSE) { return FALSE; } } } if (expanded_xml) { free_xml(expanded_xml); xml_obj = orig_xml; } if (any_sets == FALSE) { return unpack_simple_rsc_ticket(xml_obj, data_set); } return TRUE; } diff --git a/lib/pacemaker/pcmk_sched_graph.c b/lib/pacemaker/pcmk_sched_graph.c index 51f70636b3..c012d23ce6 100644 --- a/lib/pacemaker/pcmk_sched_graph.c +++ b/lib/pacemaker/pcmk_sched_graph.c @@ -1,1858 +1,1859 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include void update_colo_start_chain(pe_action_t *action, pe_working_set_t *data_set); gboolean rsc_update_action(pe_action_t * first, pe_action_t * then, enum pe_ordering type); static enum pe_action_flags get_action_flags(pe_action_t * action, pe_node_t * node) { enum pe_action_flags flags = action->flags; if (action->rsc) { flags = action->rsc->cmds->action_flags(action, NULL); if (pe_rsc_is_clone(action->rsc) && node) { /* We only care about activity on $node */ enum pe_action_flags clone_flags = action->rsc->cmds->action_flags(action, node); /* Go to great lengths to ensure the correct value for pe_action_runnable... * * If we are a clone, then for _ordering_ constraints, it's only relevant * if we are runnable _anywhere_. * * This only applies to _runnable_ though, and only for ordering constraints. * If this function is ever used during colocation, then we'll need additional logic * * Not very satisfying, but it's logical and appears to work well. */ if (!pcmk_is_set(clone_flags, pe_action_runnable) && pcmk_is_set(flags, pe_action_runnable)) { pe__set_raw_action_flags(clone_flags, action->rsc->id, pe_action_runnable); } flags = clone_flags; } } return flags; } static char * convert_non_atomic_uuid(char *old_uuid, pe_resource_t * rsc, gboolean allow_notify, gboolean free_original) { guint interval_ms = 0; char *uuid = NULL; char *rid = NULL; char *raw_task = NULL; int task = no_action; CRM_ASSERT(rsc); pe_rsc_trace(rsc, "Processing %s", old_uuid); if (old_uuid == NULL) { return NULL; } else if (strstr(old_uuid, "notify") != NULL) { goto done; /* no conversion */ } else if (rsc->variant < pe_group) { goto done; /* no conversion */ } CRM_ASSERT(parse_op_key(old_uuid, &rid, &raw_task, &interval_ms)); if (interval_ms > 0) { goto done; /* no conversion */ } task = text2task(raw_task); switch (task) { case stop_rsc: case start_rsc: case action_notify: case action_promote: case action_demote: break; case stopped_rsc: case started_rsc: case action_notified: case action_promoted: case action_demoted: task--; break; case monitor_rsc: case shutdown_crm: case stonith_node: task = no_action; break; default: crm_err("Unknown action: %s", raw_task); task = no_action; break; } if (task != no_action) { if (pcmk_is_set(rsc->flags, pe_rsc_notify) && allow_notify) { uuid = pcmk__notify_key(rid, "confirmed-post", task2text(task + 1)); } else { uuid = pcmk__op_key(rid, task2text(task + 1), 0); } pe_rsc_trace(rsc, "Converted %s -> %s", old_uuid, uuid); } done: if (uuid == NULL) { uuid = strdup(old_uuid); } if (free_original) { free(old_uuid); } free(raw_task); free(rid); return uuid; } static pe_action_t * rsc_expand_action(pe_action_t * action) { gboolean notify = FALSE; pe_action_t *result = action; pe_resource_t *rsc = action->rsc; if (rsc == NULL) { return action; } if ((rsc->parent == NULL) || (pe_rsc_is_clone(rsc) && (rsc->parent->variant == pe_container))) { /* Only outermost resources have notification actions. * The exception is those in bundles. */ notify = pcmk_is_set(rsc->flags, pe_rsc_notify); } if (rsc->variant >= pe_group) { /* Expand 'start' -> 'started' */ char *uuid = NULL; uuid = convert_non_atomic_uuid(action->uuid, rsc, notify, FALSE); if (uuid) { pe_rsc_trace(rsc, "Converting %s to %s %d", action->uuid, uuid, pcmk_is_set(rsc->flags, pe_rsc_notify)); result = find_first_action(rsc->actions, uuid, NULL, NULL); if (result == NULL) { crm_err("Couldn't expand %s to %s in %s", action->uuid, uuid, rsc->id); result = action; } free(uuid); } } return result; } static enum pe_graph_flags graph_update_action(pe_action_t * first, pe_action_t * then, pe_node_t * node, enum pe_action_flags first_flags, enum pe_action_flags then_flags, pe_action_wrapper_t *order, pe_working_set_t *data_set) { enum pe_graph_flags changed = pe_graph_none; enum pe_ordering type = order->type; gboolean processed = FALSE; /* TODO: Do as many of these in parallel as possible */ if (pcmk_is_set(type, pe_order_implies_then_on_node)) { /* Normally we want the _whole_ 'then' clone to * restart if 'first' is restarted, so then->node is * needed. * * However for unfencing, we want to limit this to * instances on the same node as 'first' (the * unfencing operation), so first->node is supplied. * * Swap the node, from then on we can can treat it * like any other 'pe_order_implies_then' */ pe__clear_order_flags(type, pe_order_implies_then_on_node); pe__set_order_flags(type, pe_order_implies_then); node = first->node; } pe__clear_raw_action_flags(first_flags, "first action update", pe_action_pseudo); if (type & pe_order_implies_then) { processed = TRUE; if (then->rsc) { changed |= then->rsc->cmds->update_actions(first, then, node, first_flags & pe_action_optional, pe_action_optional, pe_order_implies_then, data_set); } else if (!pcmk_is_set(first_flags, pe_action_optional)) { if (update_action_flags(then, pe_action_optional | pe_action_clear, __func__, __LINE__)) { pe__set_graph_flags(changed, first, pe_graph_updated_then); } } if (changed) { pe_rsc_trace(then->rsc, "implies right: %s then %s: changed", first->uuid, then->uuid); } else { crm_trace("implies right: %s then %s %p", first->uuid, then->uuid, then->rsc); } } if ((type & pe_order_restart) && then->rsc) { enum pe_action_flags restart = (pe_action_optional | pe_action_runnable); processed = TRUE; changed |= then->rsc->cmds->update_actions(first, then, node, first_flags, restart, pe_order_restart, data_set); if (changed) { pe_rsc_trace(then->rsc, "restart: %s then %s: changed", first->uuid, then->uuid); } else { crm_trace("restart: %s then %s", first->uuid, then->uuid); } } if (type & pe_order_implies_first) { processed = TRUE; if (first->rsc) { changed |= first->rsc->cmds->update_actions(first, then, node, first_flags, pe_action_optional, pe_order_implies_first, data_set); } else if (!pcmk_is_set(first_flags, pe_action_optional)) { pe_rsc_trace(first->rsc, "first unrunnable: %s (%d) then %s (%d)", first->uuid, pcmk_is_set(first_flags, pe_action_optional), then->uuid, pcmk_is_set(then_flags, pe_action_optional)); if (update_action_flags(first, pe_action_runnable | pe_action_clear, __func__, __LINE__)) { pe__set_graph_flags(changed, first, pe_graph_updated_first); } } if (changed) { pe_rsc_trace(then->rsc, "implies left: %s then %s: changed", first->uuid, then->uuid); } else { crm_trace("implies left: %s (%d) then %s (%d)", first->uuid, pcmk_is_set(first_flags, pe_action_optional), then->uuid, pcmk_is_set(then_flags, pe_action_optional)); } } if (type & pe_order_implies_first_master) { processed = TRUE; if (then->rsc) { changed |= then->rsc->cmds->update_actions(first, then, node, first_flags & pe_action_optional, pe_action_optional, pe_order_implies_first_master, data_set); } if (changed) { pe_rsc_trace(then->rsc, "implies left when right rsc is Master role: %s then %s: changed", first->uuid, then->uuid); } else { crm_trace("implies left when right rsc is Master role: %s then %s", first->uuid, then->uuid); } } if (type & pe_order_one_or_more) { processed = TRUE; if (then->rsc) { changed |= then->rsc->cmds->update_actions(first, then, node, first_flags, pe_action_runnable, pe_order_one_or_more, data_set); } else if (pcmk_is_set(first_flags, pe_action_runnable)) { /* alright. a "first" action is considered runnable, incremente * the 'runnable_before' counter */ then->runnable_before++; /* if the runnable before count for then exceeds the required number * of "before" runnable actions... mark then as runnable */ if (then->runnable_before >= then->required_runnable_before) { if (update_action_flags(then, pe_action_runnable, __func__, __LINE__)) { pe__set_graph_flags(changed, first, pe_graph_updated_then); } } } if (changed) { pe_rsc_trace(then->rsc, "runnable_one_or_more: %s then %s: changed", first->uuid, then->uuid); } else { crm_trace("runnable_one_or_more: %s then %s", first->uuid, then->uuid); } } if (then->rsc && pcmk_is_set(type, pe_order_probe)) { processed = TRUE; if (!pcmk_is_set(first_flags, pe_action_runnable) && (first->rsc->running_on != NULL)) { pe_rsc_trace(then->rsc, "Ignoring %s then %s - %s is about to be stopped", first->uuid, then->uuid, first->rsc->id); type = pe_order_none; order->type = pe_order_none; } else { pe_rsc_trace(then->rsc, "Enforcing %s then %s", first->uuid, then->uuid); changed |= then->rsc->cmds->update_actions(first, then, node, first_flags, pe_action_runnable, pe_order_runnable_left, data_set); } if (changed) { pe_rsc_trace(then->rsc, "runnable: %s then %s: changed", first->uuid, then->uuid); } else { crm_trace("runnable: %s then %s", first->uuid, then->uuid); } } if (type & pe_order_runnable_left) { processed = TRUE; if (then->rsc) { changed |= then->rsc->cmds->update_actions(first, then, node, first_flags, pe_action_runnable, pe_order_runnable_left, data_set); } else if (!pcmk_is_set(first_flags, pe_action_runnable)) { pe_rsc_trace(then->rsc, "then unrunnable: %s then %s", first->uuid, then->uuid); if (update_action_flags(then, pe_action_runnable | pe_action_clear, __func__, __LINE__)) { pe__set_graph_flags(changed, first, pe_graph_updated_then); } } if (changed) { pe_rsc_trace(then->rsc, "runnable: %s then %s: changed", first->uuid, then->uuid); } else { crm_trace("runnable: %s then %s", first->uuid, then->uuid); } } if (type & pe_order_implies_first_migratable) { processed = TRUE; if (then->rsc) { changed |= then->rsc->cmds->update_actions(first, then, node, first_flags, pe_action_optional, pe_order_implies_first_migratable, data_set); } if (changed) { pe_rsc_trace(then->rsc, "optional: %s then %s: changed", first->uuid, then->uuid); } else { crm_trace("optional: %s then %s", first->uuid, then->uuid); } } if (type & pe_order_pseudo_left) { processed = TRUE; if (then->rsc) { changed |= then->rsc->cmds->update_actions(first, then, node, first_flags, pe_action_optional, pe_order_pseudo_left, data_set); } if (changed) { pe_rsc_trace(then->rsc, "optional: %s then %s: changed", first->uuid, then->uuid); } else { crm_trace("optional: %s then %s", first->uuid, then->uuid); } } if (type & pe_order_optional) { processed = TRUE; if (then->rsc) { changed |= then->rsc->cmds->update_actions(first, then, node, first_flags, pe_action_runnable, pe_order_optional, data_set); } if (changed) { pe_rsc_trace(then->rsc, "optional: %s then %s: changed", first->uuid, then->uuid); } else { crm_trace("optional: %s then %s", first->uuid, then->uuid); } } if (type & pe_order_asymmetrical) { processed = TRUE; if (then->rsc) { changed |= then->rsc->cmds->update_actions(first, then, node, first_flags, pe_action_runnable, pe_order_asymmetrical, data_set); } if (changed) { pe_rsc_trace(then->rsc, "asymmetrical: %s then %s: changed", first->uuid, then->uuid); } else { crm_trace("asymmetrical: %s then %s", first->uuid, then->uuid); } } if ((first->flags & pe_action_runnable) && (type & pe_order_implies_then_printed) && (first_flags & pe_action_optional) == 0) { processed = TRUE; crm_trace("%s implies %s printed", first->uuid, then->uuid); update_action_flags(then, pe_action_print_always, __func__, __LINE__); /* don't care about changed */ } if (pcmk_is_set(type, pe_order_implies_first_printed) && !pcmk_is_set(then_flags, pe_action_optional)) { processed = TRUE; crm_trace("%s implies %s printed", then->uuid, first->uuid); update_action_flags(first, pe_action_print_always, __func__, __LINE__); /* don't care about changed */ } if ((type & pe_order_implies_then || type & pe_order_implies_first || type & pe_order_restart) && first->rsc && pcmk__str_eq(first->task, RSC_STOP, pcmk__str_casei) && !pcmk_is_set(first->rsc->flags, pe_rsc_managed) && pcmk_is_set(first->rsc->flags, pe_rsc_block) && !pcmk_is_set(first->flags, pe_action_runnable)) { if (update_action_flags(then, pe_action_runnable | pe_action_clear, __func__, __LINE__)) { pe__set_graph_flags(changed, first, pe_graph_updated_then); } if (changed) { pe_rsc_trace(then->rsc, "unmanaged left: %s then %s: changed", first->uuid, then->uuid); } else { crm_trace("unmanaged left: %s then %s", first->uuid, then->uuid); } } if (processed == FALSE) { crm_trace("Constraint 0x%.6x not applicable", type); } return changed; } static void mark_start_blocked(pe_resource_t *rsc, pe_resource_t *reason, pe_working_set_t *data_set) { GListPtr gIter = rsc->actions; char *reason_text = crm_strdup_printf("colocation with %s", reason->id); for (; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; if (!pcmk__str_eq(action->task, RSC_START, pcmk__str_casei)) { continue; } if (pcmk_is_set(action->flags, pe_action_runnable)) { pe_action_set_flag_reason(__func__, __LINE__, action, NULL, reason_text, pe_action_runnable, FALSE); update_colo_start_chain(action, data_set); update_action(action, data_set); } } free(reason_text); } void update_colo_start_chain(pe_action_t *action, pe_working_set_t *data_set) { GListPtr gIter = NULL; pe_resource_t *rsc = NULL; if (!pcmk_is_set(action->flags, pe_action_runnable) && pcmk__str_eq(action->task, RSC_START, pcmk__str_casei)) { rsc = uber_parent(action->rsc); if (rsc->parent) { /* For bundles, uber_parent() returns the clone/master, not the * bundle, so the existence of rsc->parent implies this is a bundle. * In this case, we need the bundle resource, so that we can check * if all containers are stopped/stopping. */ rsc = rsc->parent; } } if (rsc == NULL || rsc->rsc_cons_lhs == NULL) { return; } /* if rsc has children, all the children need to have start set to * unrunnable before we follow the colo chain for the parent. */ for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child = (pe_resource_t *)gIter->data; pe_action_t *start = find_first_action(child->actions, NULL, RSC_START, NULL); if ((start == NULL) || pcmk_is_set(start->flags, pe_action_runnable)) { return; } } for (gIter = rsc->rsc_cons_lhs; gIter != NULL; gIter = gIter->next) { - rsc_colocation_t *colocate_with = (rsc_colocation_t *)gIter->data; + pcmk__colocation_t *colocate_with = (pcmk__colocation_t *) gIter->data; + if (colocate_with->score == INFINITY) { mark_start_blocked(colocate_with->rsc_lh, action->rsc, data_set); } } } gboolean update_action(pe_action_t *then, pe_working_set_t *data_set) { GListPtr lpc = NULL; enum pe_graph_flags changed = pe_graph_none; int last_flags = then->flags; crm_trace("Processing %s (%s %s %s)", then->uuid, pcmk_is_set(then->flags, pe_action_optional)? "optional" : "required", pcmk_is_set(then->flags, pe_action_runnable)? "runnable" : "unrunnable", pcmk_is_set(then->flags, pe_action_pseudo)? "pseudo" : (then->node? then->node->details->uname : "")); if (pcmk_is_set(then->flags, pe_action_requires_any)) { /* initialize current known runnable before actions to 0 * from here as graph_update_action is called for each of * then's before actions, this number will increment as * runnable 'first' actions are encountered */ then->runnable_before = 0; /* for backwards compatibility with previous options that use * the 'requires_any' flag, initialize required to 1 if it is * not set. */ if (then->required_runnable_before == 0) { then->required_runnable_before = 1; } pe__clear_action_flags(then, pe_action_runnable); /* We are relying on the pe_order_one_or_more clause of * graph_update_action(), called as part of the: * * 'if (first == other->action)' * * block below, to set this back if appropriate */ } for (lpc = then->actions_before; lpc != NULL; lpc = lpc->next) { pe_action_wrapper_t *other = (pe_action_wrapper_t *) lpc->data; pe_action_t *first = other->action; pe_node_t *then_node = then->node; pe_node_t *first_node = first->node; enum pe_action_flags then_flags = 0; enum pe_action_flags first_flags = 0; if (first->rsc && first->rsc->variant == pe_group && pcmk__str_eq(first->task, RSC_START, pcmk__str_casei)) { first_node = first->rsc->fns->location(first->rsc, NULL, FALSE); if (first_node) { crm_trace("First: Found node %s for %s", first_node->details->uname, first->uuid); } } if (then->rsc && then->rsc->variant == pe_group && pcmk__str_eq(then->task, RSC_START, pcmk__str_casei)) { then_node = then->rsc->fns->location(then->rsc, NULL, FALSE); if (then_node) { crm_trace("Then: Found node %s for %s", then_node->details->uname, then->uuid); } } /* Disable constraint if it only applies when on same node, but isn't */ if (pcmk_is_set(other->type, pe_order_same_node) && (first_node != NULL) && (then_node != NULL) && (first_node->details != then_node->details)) { crm_trace("Disabled constraint %s on %s -> %s on %s", other->action->uuid, first_node->details->uname, then->uuid, then_node->details->uname); other->type = pe_order_none; continue; } pe__clear_graph_flags(changed, then, pe_graph_updated_first); if (first->rsc && pcmk_is_set(other->type, pe_order_then_cancels_first) && !pcmk_is_set(then->flags, pe_action_optional)) { /* 'then' is required, so we must abandon 'first' * (e.g. a required stop cancels any reload). */ pe__set_action_flags(other->action, pe_action_optional); if (!strcmp(first->task, CRMD_ACTION_RELOAD)) { pe__clear_resource_flags(first->rsc, pe_rsc_reload); } } if (first->rsc && then->rsc && (first->rsc != then->rsc) && (is_parent(then->rsc, first->rsc) == FALSE)) { first = rsc_expand_action(first); } if (first != other->action) { crm_trace("Ordering %s after %s instead of %s", then->uuid, first->uuid, other->action->uuid); } first_flags = get_action_flags(first, then_node); then_flags = get_action_flags(then, first_node); crm_trace("Checking %s (%s %s %s) against %s (%s %s %s) filter=0x%.6x type=0x%.6x", then->uuid, pcmk_is_set(then_flags, pe_action_optional)? "optional" : "required", pcmk_is_set(then_flags, pe_action_runnable)? "runnable" : "unrunnable", pcmk_is_set(then_flags, pe_action_pseudo)? "pseudo" : (then->node? then->node->details->uname : ""), first->uuid, pcmk_is_set(first_flags, pe_action_optional)? "optional" : "required", pcmk_is_set(first_flags, pe_action_runnable)? "runnable" : "unrunnable", pcmk_is_set(first_flags, pe_action_pseudo)? "pseudo" : (first->node? first->node->details->uname : ""), first_flags, other->type); if (first == other->action) { /* * 'first' was not expanded (e.g. from 'start' to 'running'), which could mean it: * - has no associated resource, * - was a primitive, * - was pre-expanded (e.g. 'running' instead of 'start') * * The third argument here to graph_update_action() is a node which is used under two conditions: * - Interleaving, in which case first->node and * then->node are equal (and NULL) * - If 'then' is a clone, to limit the scope of the * constraint to instances on the supplied node * */ pe_node_t *node = then->node; changed |= graph_update_action(first, then, node, first_flags, then_flags, other, data_set); /* 'first' was for a complex resource (clone, group, etc), * create a new dependency if necessary */ } else if (order_actions(first, then, other->type)) { /* This was the first time 'first' and 'then' were associated, * start again to get the new actions_before list */ pe__set_graph_flags(changed, then, pe_graph_updated_then|pe_graph_disable); } if (changed & pe_graph_disable) { crm_trace("Disabled constraint %s -> %s in favor of %s -> %s", other->action->uuid, then->uuid, first->uuid, then->uuid); pe__clear_graph_flags(changed, then, pe_graph_disable); other->type = pe_order_none; } if (changed & pe_graph_updated_first) { GListPtr lpc2 = NULL; crm_trace("Updated %s (first %s %s %s), processing dependents ", first->uuid, pcmk_is_set(first->flags, pe_action_optional)? "optional" : "required", pcmk_is_set(first->flags, pe_action_runnable)? "runnable" : "unrunnable", pcmk_is_set(first->flags, pe_action_pseudo)? "pseudo" : (first->node? first->node->details->uname : "")); for (lpc2 = first->actions_after; lpc2 != NULL; lpc2 = lpc2->next) { pe_action_wrapper_t *other = (pe_action_wrapper_t *) lpc2->data; update_action(other->action, data_set); } update_action(first, data_set); } } if (pcmk_is_set(then->flags, pe_action_requires_any)) { if (last_flags != then->flags) { pe__set_graph_flags(changed, then, pe_graph_updated_then); } else { pe__clear_graph_flags(changed, then, pe_graph_updated_then); } } if (changed & pe_graph_updated_then) { crm_trace("Updated %s (then %s %s %s), processing dependents ", then->uuid, pcmk_is_set(then->flags, pe_action_optional)? "optional" : "required", pcmk_is_set(then->flags, pe_action_runnable)? "runnable" : "unrunnable", pcmk_is_set(then->flags, pe_action_pseudo)? "pseudo" : (then->node? then->node->details-> uname : "")); if (pcmk_is_set(last_flags, pe_action_runnable) && !pcmk_is_set(then->flags, pe_action_runnable)) { update_colo_start_chain(then, data_set); } update_action(then, data_set); for (lpc = then->actions_after; lpc != NULL; lpc = lpc->next) { pe_action_wrapper_t *other = (pe_action_wrapper_t *) lpc->data; update_action(other->action, data_set); } } return FALSE; } gboolean shutdown_constraints(pe_node_t * node, pe_action_t * shutdown_op, pe_working_set_t * data_set) { /* add the stop to the before lists so it counts as a pre-req * for the shutdown */ GListPtr lpc = NULL; for (lpc = data_set->actions; lpc != NULL; lpc = lpc->next) { pe_action_t *action = (pe_action_t *) lpc->data; if (action->rsc == NULL || action->node == NULL) { continue; } else if (action->node->details != node->details) { continue; } else if (pcmk_is_set(action->rsc->flags, pe_rsc_maintenance)) { pe_rsc_trace(action->rsc, "Skipping %s: maintenance mode", action->uuid); continue; } else if (node->details->maintenance) { pe_rsc_trace(action->rsc, "Skipping %s: node %s is in maintenance mode", action->uuid, node->details->uname); continue; } else if (!pcmk__str_eq(action->task, RSC_STOP, pcmk__str_casei)) { continue; } else if (!pcmk_any_flags_set(action->rsc->flags, pe_rsc_managed|pe_rsc_block)) { /* * If another action depends on this one, we may still end up blocking */ pe_rsc_trace(action->rsc, "Skipping %s: unmanaged", action->uuid); continue; } pe_rsc_trace(action->rsc, "Ordering %s before shutdown on %s", action->uuid, node->details->uname); pe__clear_action_flags(action, pe_action_optional); custom_action_order(action->rsc, NULL, action, NULL, strdup(CRM_OP_SHUTDOWN), shutdown_op, pe_order_optional | pe_order_runnable_left, data_set); } return TRUE; } /*! * \internal * \brief Order all actions appropriately relative to a fencing operation * * Ensure start operations of affected resources are ordered after fencing, * imply stop and demote operations of affected resources by marking them as * pseudo-actions, etc. * * \param[in] stonith_op Fencing operation * \param[in,out] data_set Working set of cluster */ void pcmk__order_vs_fence(pe_action_t *stonith_op, pe_working_set_t *data_set) { CRM_CHECK(stonith_op && data_set, return); for (GList *r = data_set->resources; r != NULL; r = r->next) { rsc_stonith_ordering((pe_resource_t *) r->data, stonith_op, data_set); } } static pe_node_t * get_router_node(pe_action_t *action) { pe_node_t *began_on = NULL; pe_node_t *ended_on = NULL; pe_node_t *router_node = NULL; bool partial_migration = FALSE; const char *task = action->task; if (pcmk__str_eq(task, CRM_OP_FENCE, pcmk__str_casei) || !pe__is_guest_or_remote_node(action->node)) { return NULL; } CRM_ASSERT(action->node->details->remote_rsc != NULL); began_on = pe__current_node(action->node->details->remote_rsc); ended_on = action->node->details->remote_rsc->allocated_to; if (action->node->details->remote_rsc && (action->node->details->remote_rsc->container == NULL) && action->node->details->remote_rsc->partial_migration_target) { partial_migration = TRUE; } /* if there is only one location to choose from, * this is easy. Check for those conditions first */ if (!began_on || !ended_on) { /* remote rsc is either shutting down or starting up */ return began_on ? began_on : ended_on; } else if (began_on->details == ended_on->details) { /* remote rsc didn't move nodes. */ return began_on; } /* If we have get here, we know the remote resource * began on one node and is moving to another node. * * This means some actions will get routed through the cluster * node the connection rsc began on, and others are routed through * the cluster node the connection rsc ends up on. * * 1. stop, demote, migrate actions of resources living in the remote * node _MUST_ occur _BEFORE_ the connection can move (these actions * are all required before the remote rsc stop action can occur.) In * this case, we know these actions have to be routed through the initial * cluster node the connection resource lived on before the move takes place. * The exception is a partial migration of a (non-guest) remote * connection resource; in that case, all actions (even these) will be * ordered after the connection's pseudo-start on the migration target, * so the target is the router node. * * 2. Everything else (start, promote, monitor, probe, refresh, clear failcount * delete ....) must occur after the resource starts on the node it is * moving to. */ if (pcmk__str_eq(task, "notify", pcmk__str_casei)) { task = g_hash_table_lookup(action->meta, "notify_operation"); } /* 1. before connection rsc moves. */ if (pcmk__strcase_any_of(task, "stop", "demote", "migrate_from", "migrate_to", NULL) && !partial_migration) { router_node = began_on; /* 2. after connection rsc moves. */ } else { router_node = ended_on; } return router_node; } /*! * \internal * \brief Add an XML node tag for a specified ID * * \param[in] id Node UUID to add * \param[in,out] xml Parent XML tag to add to */ static xmlNode* add_node_to_xml_by_id(const char *id, xmlNode *xml) { xmlNode *node_xml; node_xml = create_xml_node(xml, XML_CIB_TAG_NODE); crm_xml_add(node_xml, XML_ATTR_UUID, id); return node_xml; } /*! * \internal * \brief Add an XML node tag for a specified node * * \param[in] node Node to add * \param[in,out] xml XML to add node to */ static void add_node_to_xml(const pe_node_t *node, void *xml) { add_node_to_xml_by_id(node->details->id, (xmlNode *) xml); } /*! * \internal * \brief Add XML with nodes that need an update of their maintenance state * * \param[in,out] xml Parent XML tag to add to * \param[in] data_set Working set for cluster */ static int add_maintenance_nodes(xmlNode *xml, const pe_working_set_t *data_set) { GListPtr gIter = NULL; xmlNode *maintenance = xml?create_xml_node(xml, XML_GRAPH_TAG_MAINTENANCE):NULL; int count = 0; for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; struct pe_node_shared_s *details = node->details; if (!pe__is_guest_or_remote_node(node)) { continue; /* just remote nodes need to know atm */ } if (details->maintenance != details->remote_maintenance) { if (maintenance) { crm_xml_add( add_node_to_xml_by_id(node->details->id, maintenance), XML_NODE_IS_MAINTENANCE, details->maintenance?"1":"0"); } count++; } } crm_trace("%s %d nodes to adjust maintenance-mode " "to transition", maintenance?"Added":"Counted", count); return count; } /*! * \internal * \brief Add pseudo action with nodes needing maintenance state update * * \param[in,out] data_set Working set for cluster */ void add_maintenance_update(pe_working_set_t *data_set) { pe_action_t *action = NULL; if (add_maintenance_nodes(NULL, data_set)) { crm_trace("adding maintenance state update pseudo action"); action = get_pseudo_op(CRM_OP_MAINTENANCE_NODES, data_set); pe__set_action_flags(action, pe_action_print_always); } } /*! * \internal * \brief Add XML with nodes that an action is expected to bring down * * If a specified action is expected to bring any nodes down, add an XML block * with their UUIDs. When a node is lost, this allows the controller to * determine whether it was expected. * * \param[in,out] xml Parent XML tag to add to * \param[in] action Action to check for downed nodes * \param[in] data_set Working set for cluster */ static void add_downed_nodes(xmlNode *xml, const pe_action_t *action, const pe_working_set_t *data_set) { CRM_CHECK(xml && action && action->node && data_set, return); if (pcmk__str_eq(action->task, CRM_OP_SHUTDOWN, pcmk__str_casei)) { /* Shutdown makes the action's node down */ xmlNode *downed = create_xml_node(xml, XML_GRAPH_TAG_DOWNED); add_node_to_xml_by_id(action->node->details->id, downed); } else if (pcmk__str_eq(action->task, CRM_OP_FENCE, pcmk__str_casei)) { /* Fencing makes the action's node and any hosted guest nodes down */ const char *fence = g_hash_table_lookup(action->meta, "stonith_action"); if (pcmk__strcase_any_of(fence, "off", "reboot", NULL)) { xmlNode *downed = create_xml_node(xml, XML_GRAPH_TAG_DOWNED); add_node_to_xml_by_id(action->node->details->id, downed); pe_foreach_guest_node(data_set, action->node, add_node_to_xml, downed); } } else if (action->rsc && action->rsc->is_remote_node && pcmk__str_eq(action->task, CRMD_ACTION_STOP, pcmk__str_casei)) { /* Stopping a remote connection resource makes connected node down, * unless it's part of a migration */ GListPtr iter; pe_action_t *input; gboolean migrating = FALSE; for (iter = action->actions_before; iter != NULL; iter = iter->next) { input = ((pe_action_wrapper_t *) iter->data)->action; if (input->rsc && pcmk__str_eq(action->rsc->id, input->rsc->id, pcmk__str_casei) && pcmk__str_eq(input->task, CRMD_ACTION_MIGRATED, pcmk__str_casei)) { migrating = TRUE; break; } } if (!migrating) { xmlNode *downed = create_xml_node(xml, XML_GRAPH_TAG_DOWNED); add_node_to_xml_by_id(action->rsc->id, downed); } } } static bool should_lock_action(pe_action_t *action) { // Only actions taking place on resource's lock node are locked if ((action->rsc->lock_node == NULL) || (action->node == NULL) || (action->node->details != action->rsc->lock_node->details)) { return false; } /* During shutdown, only stops are locked (otherwise, another action such as * a demote would cause the controller to clear the lock) */ if (action->node->details->shutdown && action->task && strcmp(action->task, RSC_STOP)) { return false; } return true; } static xmlNode * action2xml(pe_action_t * action, gboolean as_input, pe_working_set_t *data_set) { gboolean needs_node_info = TRUE; gboolean needs_maintenance_info = FALSE; xmlNode *action_xml = NULL; xmlNode *args_xml = NULL; #if ENABLE_VERSIONED_ATTRS pe_rsc_action_details_t *rsc_details = NULL; #endif if (action == NULL) { return NULL; } if (pcmk__str_eq(action->task, CRM_OP_FENCE, pcmk__str_casei)) { /* All fences need node info; guest node fences are pseudo-events */ action_xml = create_xml_node(NULL, pcmk_is_set(action->flags, pe_action_pseudo)? XML_GRAPH_TAG_PSEUDO_EVENT : XML_GRAPH_TAG_CRM_EVENT); } else if (pcmk__str_eq(action->task, CRM_OP_SHUTDOWN, pcmk__str_casei)) { action_xml = create_xml_node(NULL, XML_GRAPH_TAG_CRM_EVENT); } else if (pcmk__str_eq(action->task, CRM_OP_CLEAR_FAILCOUNT, pcmk__str_casei)) { action_xml = create_xml_node(NULL, XML_GRAPH_TAG_CRM_EVENT); } else if (pcmk__str_eq(action->task, CRM_OP_LRM_REFRESH, pcmk__str_casei)) { action_xml = create_xml_node(NULL, XML_GRAPH_TAG_CRM_EVENT); } else if (pcmk__str_eq(action->task, CRM_OP_LRM_DELETE, pcmk__str_casei)) { // CIB-only clean-up for shutdown locks action_xml = create_xml_node(NULL, XML_GRAPH_TAG_CRM_EVENT); crm_xml_add(action_xml, PCMK__XA_MODE, XML_TAG_CIB); /* } else if(pcmk__str_eq(action->task, RSC_PROBED, pcmk__str_casei)) { */ /* action_xml = create_xml_node(NULL, XML_GRAPH_TAG_CRM_EVENT); */ } else if (pcmk_is_set(action->flags, pe_action_pseudo)) { if (pcmk__str_eq(action->task, CRM_OP_MAINTENANCE_NODES, pcmk__str_casei)) { needs_maintenance_info = TRUE; } action_xml = create_xml_node(NULL, XML_GRAPH_TAG_PSEUDO_EVENT); needs_node_info = FALSE; } else { action_xml = create_xml_node(NULL, XML_GRAPH_TAG_RSC_OP); #if ENABLE_VERSIONED_ATTRS rsc_details = pe_rsc_action_details(action); #endif } crm_xml_add_int(action_xml, XML_ATTR_ID, action->id); crm_xml_add(action_xml, XML_LRM_ATTR_TASK, action->task); if (action->rsc != NULL && action->rsc->clone_name != NULL) { char *clone_key = NULL; guint interval_ms; if (pcmk__guint_from_hash(action->meta, XML_LRM_ATTR_INTERVAL_MS, 0, &interval_ms) != pcmk_rc_ok) { interval_ms = 0; } if (pcmk__str_eq(action->task, RSC_NOTIFY, pcmk__str_casei)) { const char *n_type = g_hash_table_lookup(action->meta, "notify_type"); const char *n_task = g_hash_table_lookup(action->meta, "notify_operation"); CRM_CHECK(n_type != NULL, crm_err("No notify type value found for %s", action->uuid)); CRM_CHECK(n_task != NULL, crm_err("No notify operation value found for %s", action->uuid)); clone_key = pcmk__notify_key(action->rsc->clone_name, n_type, n_task); } else if(action->cancel_task) { clone_key = pcmk__op_key(action->rsc->clone_name, action->cancel_task, interval_ms); } else { clone_key = pcmk__op_key(action->rsc->clone_name, action->task, interval_ms); } CRM_CHECK(clone_key != NULL, crm_err("Could not generate a key for %s", action->uuid)); crm_xml_add(action_xml, XML_LRM_ATTR_TASK_KEY, clone_key); crm_xml_add(action_xml, "internal_" XML_LRM_ATTR_TASK_KEY, action->uuid); free(clone_key); } else { crm_xml_add(action_xml, XML_LRM_ATTR_TASK_KEY, action->uuid); } if (needs_node_info && action->node != NULL) { pe_node_t *router_node = get_router_node(action); crm_xml_add(action_xml, XML_LRM_ATTR_TARGET, action->node->details->uname); crm_xml_add(action_xml, XML_LRM_ATTR_TARGET_UUID, action->node->details->id); if (router_node) { crm_xml_add(action_xml, XML_LRM_ATTR_ROUTER_NODE, router_node->details->uname); } g_hash_table_insert(action->meta, strdup(XML_LRM_ATTR_TARGET), strdup(action->node->details->uname)); g_hash_table_insert(action->meta, strdup(XML_LRM_ATTR_TARGET_UUID), strdup(action->node->details->id)); } /* No details if this action is only being listed in the inputs section */ if (as_input) { return action_xml; } if (action->rsc && !pcmk_is_set(action->flags, pe_action_pseudo)) { int lpc = 0; xmlNode *rsc_xml = NULL; const char *attr_list[] = { XML_AGENT_ATTR_CLASS, XML_AGENT_ATTR_PROVIDER, XML_ATTR_TYPE }; /* If a resource is locked to a node via shutdown-lock, mark its actions * so the controller can preserve the lock when the action completes. */ if (should_lock_action(action)) { crm_xml_add_ll(action_xml, XML_CONFIG_ATTR_SHUTDOWN_LOCK, (long long) action->rsc->lock_time); } // List affected resource rsc_xml = create_xml_node(action_xml, crm_element_name(action->rsc->xml)); if (pcmk_is_set(action->rsc->flags, pe_rsc_orphan) && action->rsc->clone_name) { /* Do not use the 'instance free' name here as that * might interfere with the instance we plan to keep. * Ie. if there are more than two named /anonymous/ * instances on a given node, we need to make sure the * command goes to the right one. * * Keep this block, even when everyone is using * 'instance free' anonymous clone names - it means * we'll do the right thing if anyone toggles the * unique flag to 'off' */ crm_debug("Using orphan clone name %s instead of %s", action->rsc->id, action->rsc->clone_name); crm_xml_add(rsc_xml, XML_ATTR_ID, action->rsc->clone_name); crm_xml_add(rsc_xml, XML_ATTR_ID_LONG, action->rsc->id); } else if (!pcmk_is_set(action->rsc->flags, pe_rsc_unique)) { const char *xml_id = ID(action->rsc->xml); crm_debug("Using anonymous clone name %s for %s (aka. %s)", xml_id, action->rsc->id, action->rsc->clone_name); /* ID is what we'd like client to use * ID_LONG is what they might know it as instead * * ID_LONG is only strictly needed /here/ during the * transition period until all nodes in the cluster * are running the new software /and/ have rebooted * once (meaning that they've only ever spoken to a DC * supporting this feature). * * If anyone toggles the unique flag to 'on', the * 'instance free' name will correspond to an orphan * and fall into the clause above instead */ crm_xml_add(rsc_xml, XML_ATTR_ID, xml_id); if (action->rsc->clone_name && !pcmk__str_eq(xml_id, action->rsc->clone_name, pcmk__str_casei)) { crm_xml_add(rsc_xml, XML_ATTR_ID_LONG, action->rsc->clone_name); } else { crm_xml_add(rsc_xml, XML_ATTR_ID_LONG, action->rsc->id); } } else { CRM_ASSERT(action->rsc->clone_name == NULL); crm_xml_add(rsc_xml, XML_ATTR_ID, action->rsc->id); } for (lpc = 0; lpc < DIMOF(attr_list); lpc++) { crm_xml_add(rsc_xml, attr_list[lpc], g_hash_table_lookup(action->rsc->meta, attr_list[lpc])); } } /* List any attributes in effect */ args_xml = create_xml_node(NULL, XML_TAG_ATTRS); crm_xml_add(args_xml, XML_ATTR_CRM_VERSION, CRM_FEATURE_SET); g_hash_table_foreach(action->extra, hash2field, args_xml); if (action->rsc != NULL && action->node) { GHashTable *p = crm_str_table_new(); get_rsc_attributes(p, action->rsc, action->node, data_set); g_hash_table_foreach(p, hash2smartfield, args_xml); g_hash_table_destroy(p); #if ENABLE_VERSIONED_ATTRS { xmlNode *versioned_parameters = create_xml_node(NULL, XML_TAG_RSC_VER_ATTRS); pe_get_versioned_attributes(versioned_parameters, action->rsc, action->node, data_set); if (xml_has_children(versioned_parameters)) { add_node_copy(action_xml, versioned_parameters); } free_xml(versioned_parameters); } #endif } else if(action->rsc && action->rsc->variant <= pe_native) { g_hash_table_foreach(action->rsc->parameters, hash2smartfield, args_xml); #if ENABLE_VERSIONED_ATTRS if (xml_has_children(action->rsc->versioned_parameters)) { add_node_copy(action_xml, action->rsc->versioned_parameters); } #endif } #if ENABLE_VERSIONED_ATTRS if (rsc_details) { if (xml_has_children(rsc_details->versioned_parameters)) { add_node_copy(action_xml, rsc_details->versioned_parameters); } if (xml_has_children(rsc_details->versioned_meta)) { add_node_copy(action_xml, rsc_details->versioned_meta); } } #endif g_hash_table_foreach(action->meta, hash2metafield, args_xml); if (action->rsc != NULL) { const char *value = g_hash_table_lookup(action->rsc->meta, "external-ip"); pe_resource_t *parent = action->rsc; while (parent != NULL) { parent->cmds->append_meta(parent, args_xml); parent = parent->parent; } if(value) { hash2smartfield((gpointer)"pcmk_external_ip", (gpointer)value, (gpointer)args_xml); } if (action->node && /* make clang analyzer happy */ pe__is_guest_node(action->node)) { pe_node_t *host = NULL; enum action_tasks task = text2task(action->task); if(task == action_notify || task == action_notified) { const char *n_task = g_hash_table_lookup(action->meta, "notify_operation"); task = text2task(n_task); } // Differentiate between up and down actions switch (task) { case stop_rsc: case stopped_rsc: case action_demote: case action_demoted: host = pe__current_node(action->node->details->remote_rsc->container); break; case start_rsc: case started_rsc: case monitor_rsc: case action_promote: case action_promoted: host = action->node->details->remote_rsc->container->allocated_to; break; default: break; } if(host) { hash2metafield((gpointer)XML_RSC_ATTR_TARGET, (gpointer)g_hash_table_lookup(action->rsc->meta, XML_RSC_ATTR_TARGET), (gpointer)args_xml); hash2metafield((gpointer) PCMK__ENV_PHYSICAL_HOST, (gpointer)host->details->uname, (gpointer)args_xml); } } } else if (pcmk__str_eq(action->task, CRM_OP_FENCE, pcmk__str_casei) && action->node) { /* Pass the node's attributes as meta-attributes. * * @TODO: Determine whether it is still necessary to do this. It was * added in 33d99707, probably for the libfence-based implementation in * c9a90bd, which is no longer used. */ g_hash_table_foreach(action->node->details->attrs, hash2metafield, args_xml); } sorted_xml(args_xml, action_xml, FALSE); free_xml(args_xml); /* List any nodes this action is expected to make down */ if (needs_node_info && (action->node != NULL)) { add_downed_nodes(action_xml, action, data_set); } if (needs_maintenance_info) { add_maintenance_nodes(action_xml, data_set); } crm_log_xml_trace(action_xml, "dumped action"); return action_xml; } static bool should_dump_action(pe_action_t *action) { CRM_CHECK(action != NULL, return false); if (pcmk_is_set(action->flags, pe_action_dumped)) { crm_trace("Action %s (%d) already dumped", action->uuid, action->id); return false; } else if (pcmk_is_set(action->flags, pe_action_pseudo) && pcmk__str_eq(action->task, CRM_OP_PROBED, pcmk__str_casei)) { GListPtr lpc = NULL; /* This is a horrible but convenient hack * * It mimimizes the number of actions with unsatisfied inputs * (i.e. not included in the graph) * * This in turn, means we can be more concise when printing * aborted/incomplete graphs. * * It also makes it obvious which node is preventing * probe_complete from running (presumably because it is only * partially up) * * For these reasons we tolerate such perversions */ for (lpc = action->actions_after; lpc != NULL; lpc = lpc->next) { pe_action_wrapper_t *wrapper = (pe_action_wrapper_t *) lpc->data; if (!pcmk_is_set(wrapper->action->flags, pe_action_runnable)) { /* Only interested in runnable operations */ } else if (!pcmk__str_eq(wrapper->action->task, RSC_START, pcmk__str_casei)) { /* Only interested in start operations */ } else if (pcmk_is_set(wrapper->action->flags, pe_action_dumped) || should_dump_action(wrapper->action)) { crm_trace("Action %s (%d) should be dumped: " "dependency of %s (%d)", action->uuid, action->id, wrapper->action->uuid, wrapper->action->id); return true; } } } if (!pcmk_is_set(action->flags, pe_action_runnable)) { crm_trace("Ignoring action %s (%d): unrunnable", action->uuid, action->id); return false; } else if (pcmk_is_set(action->flags, pe_action_optional) && !pcmk_is_set(action->flags, pe_action_print_always)) { crm_trace("Ignoring action %s (%d): optional", action->uuid, action->id); return false; // Monitors should be dumped even for unmanaged resources } else if (action->rsc && !pcmk_is_set(action->rsc->flags, pe_rsc_managed) && !pcmk__str_eq(action->task, RSC_STATUS, pcmk__str_casei)) { const char *interval_ms_s = g_hash_table_lookup(action->meta, XML_LRM_ATTR_INTERVAL_MS); // Cancellation of recurring monitors should still be dumped if (pcmk__str_eq(interval_ms_s, "0", pcmk__str_null_matches)) { crm_trace("Ignoring action %s (%d): for unmanaged resource (%s)", action->uuid, action->id, action->rsc->id); return false; } } if (pcmk_is_set(action->flags, pe_action_pseudo) || pcmk__strcase_any_of(action->task, CRM_OP_FENCE, CRM_OP_SHUTDOWN, NULL)) { /* skip the next checks */ return true; } if (action->node == NULL) { pe_err("Skipping action %s (%d) " "because it was not allocated to a node (bug?)", action->uuid, action->id); log_action(LOG_DEBUG, "Unallocated action", action, false); return false; } else if (pcmk_is_set(action->flags, pe_action_dc)) { crm_trace("Action %s (%d) should be dumped: " "can run on DC instead of %s", action->uuid, action->id, action->node->details->uname); } else if (pe__is_guest_node(action->node) && !action->node->details->remote_requires_reset) { crm_trace("Action %s (%d) should be dumped: " "assuming will be runnable on guest node %s", action->uuid, action->id, action->node->details->uname); } else if (action->node->details->online == false) { pe_err("Skipping action %s (%d) " "because it was scheduled for offline node (bug?)", action->uuid, action->id); log_action(LOG_DEBUG, "Action for offline node", action, FALSE); return false; #if 0 /* but this would also affect resources that can be safely * migrated before a fencing op */ } else if (action->node->details->unclean == false) { pe_err("Skipping action %s (%d) " "because it was scheduled for unclean node (bug?)", action->uuid, action->id); log_action(LOG_DEBUG, "Action for unclean node", action, false); return false; #endif } return true; } /* lowest to highest */ static gint sort_action_id(gconstpointer a, gconstpointer b) { const pe_action_wrapper_t *action_wrapper2 = (const pe_action_wrapper_t *)a; const pe_action_wrapper_t *action_wrapper1 = (const pe_action_wrapper_t *)b; if (a == NULL) { return 1; } if (b == NULL) { return -1; } if (action_wrapper1->action->id > action_wrapper2->action->id) { return -1; } if (action_wrapper1->action->id < action_wrapper2->action->id) { return 1; } return 0; } /*! * \internal * \brief Check whether an action input should be in the transition graph * * \param[in] action Action to check * \param[in,out] input Action input to check * * \return true if input should be in graph, false otherwise * \note This function may not only check an input, but disable it under certian * circumstances (load or anti-colocation orderings that are not needed). */ static bool check_dump_input(pe_action_t *action, pe_action_wrapper_t *input) { int type = input->type; if (input->state == pe_link_dumped) { return true; } pe__clear_order_flags(type, pe_order_implies_first_printed |pe_order_implies_then_printed |pe_order_optional); if (input->type == pe_order_none) { crm_trace("Ignoring %s (%d) input %s (%d): " "ordering disabled", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (!pcmk_is_set(input->action->flags, pe_action_runnable) && (type == pe_order_none) && !pcmk__str_eq(input->action->uuid, CRM_OP_PROBED, pcmk__str_casei)) { crm_trace("Ignoring %s (%d) input %s (%d): " "optional and input unrunnable", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (!pcmk_is_set(input->action->flags, pe_action_runnable) && pcmk_is_set(input->type, pe_order_one_or_more)) { crm_trace("Ignoring %s (%d) input %s (%d): " "one-or-more and input unrunnable", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (pcmk_is_set(action->flags, pe_action_pseudo) && pcmk_is_set(input->type, pe_order_stonith_stop)) { crm_trace("Ignoring %s (%d) input %s (%d): " "stonith stop but action is pseudo", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (pcmk_is_set(input->type, pe_order_implies_first_migratable) && !pcmk_is_set(input->action->flags, pe_action_runnable)) { crm_trace("Ignoring %s (%d) input %s (%d): " "implies input migratable but input unrunnable", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (pcmk_is_set(input->type, pe_order_apply_first_non_migratable) && pcmk_is_set(input->action->flags, pe_action_migrate_runnable)) { crm_trace("Ignoring %s (%d) input %s (%d): " "only if input unmigratable but input unrunnable", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if ((input->type == pe_order_optional) && pcmk_is_set(input->action->flags, pe_action_migrate_runnable) && pcmk__ends_with(input->action->uuid, "_stop_0")) { crm_trace("Ignoring %s (%d) input %s (%d): " "optional but stop in migration", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (input->type == pe_order_load) { pe_node_t *input_node = input->action->node; // load orderings are relevant only if actions are for same node if (action->rsc && pcmk__str_eq(action->task, RSC_MIGRATE, pcmk__str_casei)) { pe_node_t *allocated = action->rsc->allocated_to; /* For load_stopped -> migrate_to orderings, we care about where it * has been allocated to, not where it will be executed. */ if ((input_node == NULL) || (allocated == NULL) || (input_node->details != allocated->details)) { crm_trace("Ignoring %s (%d) input %s (%d): " "load ordering node mismatch %s vs %s", action->uuid, action->id, input->action->uuid, input->action->id, (allocated? allocated->details->uname : ""), (input_node? input_node->details->uname : "")); input->type = pe_order_none; return false; } } else if ((input_node == NULL) || (action->node == NULL) || (input_node->details != action->node->details)) { crm_trace("Ignoring %s (%d) input %s (%d): " "load ordering node mismatch %s vs %s", action->uuid, action->id, input->action->uuid, input->action->id, (action->node? action->node->details->uname : ""), (input_node? input_node->details->uname : "")); input->type = pe_order_none; return false; } else if (pcmk_is_set(input->action->flags, pe_action_optional)) { crm_trace("Ignoring %s (%d) input %s (%d): " "load ordering input optional", action->uuid, action->id, input->action->uuid, input->action->id); input->type = pe_order_none; return false; } } else if (input->type == pe_order_anti_colocation) { if (input->action->node && action->node && (input->action->node->details != action->node->details)) { crm_trace("Ignoring %s (%d) input %s (%d): " "anti-colocation node mismatch %s vs %s", action->uuid, action->id, input->action->uuid, input->action->id, action->node->details->uname, input->action->node->details->uname); input->type = pe_order_none; return false; } else if (pcmk_is_set(input->action->flags, pe_action_optional)) { crm_trace("Ignoring %s (%d) input %s (%d): " "anti-colocation input optional", action->uuid, action->id, input->action->uuid, input->action->id); input->type = pe_order_none; return false; } } else if (input->action->rsc && input->action->rsc != action->rsc && pcmk_is_set(input->action->rsc->flags, pe_rsc_failed) && !pcmk_is_set(input->action->rsc->flags, pe_rsc_managed) && pcmk__ends_with(input->action->uuid, "_stop_0") && action->rsc && pe_rsc_is_clone(action->rsc)) { crm_warn("Ignoring requirement that %s complete before %s:" " unmanaged failed resources cannot prevent clone shutdown", input->action->uuid, action->uuid); return false; } else if (pcmk_is_set(input->action->flags, pe_action_optional) && !pcmk_any_flags_set(input->action->flags, pe_action_print_always|pe_action_dumped) && !should_dump_action(input->action)) { crm_trace("Ignoring %s (%d) input %s (%d): " "input optional", action->uuid, action->id, input->action->uuid, input->action->id); return false; } crm_trace("%s (%d) input %s (%d) @ %s should be dumped: %s, %s, %s, 0x%.6x", action->uuid, action->id, input->action->uuid, input->action->id, input->action->node? input->action->node->details->uname : "no node", pcmk_is_set(input->action->flags, pe_action_pseudo)? "pseudo" : "real", pcmk_is_set(input->action->flags, pe_action_runnable)? "runnable" : "unrunnable", pcmk_is_set(input->action->flags, pe_action_optional)? "optional" : "required", input->type); return true; } static bool graph_has_loop(pe_action_t *init_action, pe_action_t *action, pe_action_wrapper_t *input) { bool has_loop = false; if (pcmk_is_set(input->action->flags, pe_action_tracking)) { crm_trace("Breaking tracking loop: %s@%s -> %s@%s (0x%.6x)", input->action->uuid, input->action->node? input->action->node->details->uname : "", action->uuid, action->node? action->node->details->uname : "", input->type); return false; } // Don't need to check inputs that won't be used if (!check_dump_input(action, input)) { return false; } if (input->action == init_action) { crm_debug("Input loop found in %s@%s ->...-> %s@%s", action->uuid, action->node? action->node->details->uname : "", init_action->uuid, init_action->node? init_action->node->details->uname : ""); return true; } pe__set_action_flags(input->action, pe_action_tracking); crm_trace("Checking inputs of action %s@%s input %s@%s (0x%.6x)" "for graph loop with %s@%s ", action->uuid, action->node? action->node->details->uname : "", input->action->uuid, input->action->node? input->action->node->details->uname : "", input->type, init_action->uuid, init_action->node? init_action->node->details->uname : ""); // Recursively check input itself for loops for (GList *iter = input->action->actions_before; iter != NULL; iter = iter->next) { if (graph_has_loop(init_action, input->action, (pe_action_wrapper_t *) iter->data)) { // Recursive call already logged a debug message has_loop = true; goto done; } } done: pe__clear_action_flags(input->action, pe_action_tracking); if (!has_loop) { crm_trace("No input loop found in %s@%s -> %s@%s (0x%.6x)", input->action->uuid, input->action->node? input->action->node->details->uname : "", action->uuid, action->node? action->node->details->uname : "", input->type); } return has_loop; } bool pcmk__ordering_is_invalid(pe_action_t *action, pe_action_wrapper_t *input) { /* Prevent user-defined ordering constraints between resources * running in a guest node and the resource that defines that node. */ if (!pcmk_is_set(input->type, pe_order_preserve) && action->rsc && action->rsc->fillers && input->action->rsc && input->action->node && input->action->node->details->remote_rsc && (input->action->node->details->remote_rsc->container == action->rsc)) { crm_warn("Invalid ordering constraint between %s and %s", input->action->rsc->id, action->rsc->id); return true; } /* If there's an order like * "rscB_stop node2"-> "load_stopped_node2" -> "rscA_migrate_to node1" * * then rscA is being migrated from node1 to node2, while rscB is being * migrated from node2 to node1. If there would be a graph loop, * break the order "load_stopped_node2" -> "rscA_migrate_to node1". */ if ((input->type == pe_order_load) && action->rsc && pcmk__str_eq(action->task, RSC_MIGRATE, pcmk__str_casei) && graph_has_loop(action, action, input)) { return true; } return false; } // Remove duplicate inputs (regardless of flags) static void deduplicate_inputs(pe_action_t *action) { GList *item = NULL; GList *next = NULL; pe_action_wrapper_t *last_input = NULL; action->actions_before = g_list_sort(action->actions_before, sort_action_id); for (item = action->actions_before; item != NULL; item = next) { pe_action_wrapper_t *input = (pe_action_wrapper_t *) item->data; next = item->next; if (last_input && (input->action->id == last_input->action->id)) { crm_trace("Input %s (%d) duplicate skipped for action %s (%d)", input->action->uuid, input->action->id, action->uuid, action->id); /* For the purposes of scheduling, the ordering flags no longer * matter, but crm_simulate looks at certain ones when creating a * dot graph. Combining the flags is sufficient for that purpose. */ last_input->type |= input->type; if (input->state == pe_link_dumped) { last_input->state = pe_link_dumped; } free(item->data); action->actions_before = g_list_delete_link(action->actions_before, item); } else { last_input = input; input->state = pe_link_not_dumped; } } } /*! * \internal * \brief Add an action to the transition graph XML if appropriate * * \param[in] action Action to possibly add * \param[in] data_set Cluster working set * * \note This will de-duplicate the action inputs, meaning that the * pe_action_wrapper_t:type flags can no longer be relied on to retain * their original settings. That means this MUST be called after stage7() * is complete, and nothing after this should rely on those type flags. * (For example, some code looks for type equal to some flag rather than * whether the flag is set, and some code looks for particular * combinations of flags -- such code must be done before stage8().) */ void graph_element_from_action(pe_action_t *action, pe_working_set_t *data_set) { GList *lpc = NULL; int synapse_priority = 0; xmlNode *syn = NULL; xmlNode *set = NULL; xmlNode *in = NULL; xmlNode *xml_action = NULL; pe_action_wrapper_t *input = NULL; /* If we haven't already, de-duplicate inputs -- even if we won't be dumping * the action, so that crm_simulate dot graphs don't have duplicates. */ if (!pcmk_is_set(action->flags, pe_action_dedup)) { deduplicate_inputs(action); pe__set_action_flags(action, pe_action_dedup); } if (should_dump_action(action) == FALSE) { return; } pe__set_action_flags(action, pe_action_dumped); syn = create_xml_node(data_set->graph, "synapse"); set = create_xml_node(syn, "action_set"); in = create_xml_node(syn, "inputs"); crm_xml_add_int(syn, XML_ATTR_ID, data_set->num_synapse); data_set->num_synapse++; if (action->rsc != NULL) { synapse_priority = action->rsc->priority; } if (action->priority > synapse_priority) { synapse_priority = action->priority; } if (synapse_priority > 0) { crm_xml_add_int(syn, XML_CIB_ATTR_PRIORITY, synapse_priority); } xml_action = action2xml(action, FALSE, data_set); add_node_nocopy(set, crm_element_name(xml_action), xml_action); for (lpc = action->actions_before; lpc != NULL; lpc = lpc->next) { input = (pe_action_wrapper_t *) lpc->data; if (check_dump_input(action, input)) { xmlNode *input_xml = create_xml_node(in, "trigger"); input->state = pe_link_dumped; xml_action = action2xml(input->action, TRUE, data_set); add_node_nocopy(input_xml, crm_element_name(xml_action), xml_action); } } } diff --git a/lib/pacemaker/pcmk_sched_group.c b/lib/pacemaker/pcmk_sched_group.c index 35f25af8a0..5334f2369d 100644 --- a/lib/pacemaker/pcmk_sched_group.c +++ b/lib/pacemaker/pcmk_sched_group.c @@ -1,535 +1,535 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #define VARIANT_GROUP 1 #include pe_node_t * pcmk__group_allocate(pe_resource_t *rsc, pe_node_t *prefer, pe_working_set_t *data_set) { pe_node_t *node = NULL; pe_node_t *group_node = NULL; GListPtr gIter = NULL; group_variant_data_t *group_data = NULL; get_group_variant_data(group_data, rsc); if (!pcmk_is_set(rsc->flags, pe_rsc_provisional)) { return rsc->allocated_to; } if (pcmk_is_set(rsc->flags, pe_rsc_allocating)) { pe_rsc_debug(rsc, "Dependency loop detected involving %s", rsc->id); return NULL; } if (group_data->first_child == NULL) { // Nothing to allocate pe__clear_resource_flags(rsc, pe_rsc_provisional); return NULL; } pe__set_resource_flags(rsc, pe_rsc_allocating); rsc->role = group_data->first_child->role; group_data->first_child->rsc_cons = g_list_concat(group_data->first_child->rsc_cons, rsc->rsc_cons); rsc->rsc_cons = NULL; group_data->last_child->rsc_cons_lhs = g_list_concat(group_data->last_child->rsc_cons_lhs, rsc->rsc_cons_lhs); rsc->rsc_cons_lhs = NULL; pe__show_node_weights(!show_scores, rsc, __func__, rsc->allowed_nodes); gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; pe_rsc_trace(rsc, "Allocating group %s member %s", rsc->id, child_rsc->id); node = child_rsc->cmds->allocate(child_rsc, prefer, data_set); if (group_node == NULL) { group_node = node; } } rsc->next_role = group_data->first_child->next_role; pe__clear_resource_flags(rsc, pe_rsc_allocating|pe_rsc_provisional); if (group_data->colocated) { return group_node; } return NULL; } void group_update_pseudo_status(pe_resource_t * parent, pe_resource_t * child); void group_create_actions(pe_resource_t * rsc, pe_working_set_t * data_set) { pe_action_t *op = NULL; const char *value = NULL; GListPtr gIter = rsc->children; pe_rsc_trace(rsc, "Creating actions for %s", rsc->id); for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; child_rsc->cmds->create_actions(child_rsc, data_set); group_update_pseudo_status(rsc, child_rsc); } op = start_action(rsc, NULL, TRUE /* !group_data->child_starting */ ); pe__set_action_flags(op, pe_action_pseudo|pe_action_runnable); op = custom_action(rsc, started_key(rsc), RSC_STARTED, NULL, TRUE /* !group_data->child_starting */ , TRUE, data_set); pe__set_action_flags(op, pe_action_pseudo|pe_action_runnable); op = stop_action(rsc, NULL, TRUE /* !group_data->child_stopping */ ); pe__set_action_flags(op, pe_action_pseudo|pe_action_runnable); op = custom_action(rsc, stopped_key(rsc), RSC_STOPPED, NULL, TRUE /* !group_data->child_stopping */ , TRUE, data_set); pe__set_action_flags(op, pe_action_pseudo|pe_action_runnable); value = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_PROMOTABLE); if (crm_is_true(value)) { op = custom_action(rsc, demote_key(rsc), RSC_DEMOTE, NULL, TRUE, TRUE, data_set); pe__set_action_flags(op, pe_action_pseudo|pe_action_runnable); op = custom_action(rsc, demoted_key(rsc), RSC_DEMOTED, NULL, TRUE, TRUE, data_set); pe__set_action_flags(op, pe_action_pseudo|pe_action_runnable); op = custom_action(rsc, promote_key(rsc), RSC_PROMOTE, NULL, TRUE, TRUE, data_set); pe__set_action_flags(op, pe_action_pseudo|pe_action_runnable); op = custom_action(rsc, promoted_key(rsc), RSC_PROMOTED, NULL, TRUE, TRUE, data_set); pe__set_action_flags(op, pe_action_pseudo|pe_action_runnable); } } void group_update_pseudo_status(pe_resource_t * parent, pe_resource_t * child) { GListPtr gIter = child->actions; group_variant_data_t *group_data = NULL; get_group_variant_data(group_data, parent); if (group_data->ordered == FALSE) { /* If this group is not ordered, then leave the meta-actions as optional */ return; } if (group_data->child_stopping && group_data->child_starting) { return; } for (; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; if (pcmk_is_set(action->flags, pe_action_optional)) { continue; } if (pcmk__str_eq(RSC_STOP, action->task, pcmk__str_casei) && pcmk_is_set(action->flags, pe_action_runnable)) { group_data->child_stopping = TRUE; pe_rsc_trace(action->rsc, "Based on %s the group is stopping", action->uuid); } else if (pcmk__str_eq(RSC_START, action->task, pcmk__str_casei) && pcmk_is_set(action->flags, pe_action_runnable)) { group_data->child_starting = TRUE; pe_rsc_trace(action->rsc, "Based on %s the group is starting", action->uuid); } } } void group_internal_constraints(pe_resource_t * rsc, pe_working_set_t * data_set) { GListPtr gIter = rsc->children; pe_resource_t *last_rsc = NULL; pe_resource_t *last_active = NULL; pe_resource_t *top = uber_parent(rsc); group_variant_data_t *group_data = NULL; get_group_variant_data(group_data, rsc); new_rsc_order(rsc, RSC_STOPPED, rsc, RSC_START, pe_order_optional, data_set); new_rsc_order(rsc, RSC_START, rsc, RSC_STARTED, pe_order_runnable_left, data_set); new_rsc_order(rsc, RSC_STOP, rsc, RSC_STOPPED, pe_order_runnable_left, data_set); for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; int stop = pe_order_none; int stopped = pe_order_implies_then_printed; int start = pe_order_implies_then | pe_order_runnable_left; int started = pe_order_runnable_left | pe_order_implies_then | pe_order_implies_then_printed; child_rsc->cmds->internal_constraints(child_rsc, data_set); if (last_rsc == NULL) { if (group_data->ordered) { pe__set_order_flags(stop, pe_order_optional); stopped = pe_order_implies_then; } } else if (group_data->colocated) { - rsc_colocation_new("group:internal_colocation", NULL, INFINITY, - child_rsc, last_rsc, NULL, NULL, data_set); + pcmk__new_colocation("group:internal_colocation", NULL, INFINITY, + child_rsc, last_rsc, NULL, NULL, data_set); } if (pcmk_is_set(top->flags, pe_rsc_promotable)) { new_rsc_order(rsc, RSC_DEMOTE, child_rsc, RSC_DEMOTE, stop | pe_order_implies_first_printed, data_set); new_rsc_order(child_rsc, RSC_DEMOTE, rsc, RSC_DEMOTED, stopped, data_set); new_rsc_order(child_rsc, RSC_PROMOTE, rsc, RSC_PROMOTED, started, data_set); new_rsc_order(rsc, RSC_PROMOTE, child_rsc, RSC_PROMOTE, pe_order_implies_first_printed, data_set); } order_start_start(rsc, child_rsc, pe_order_implies_first_printed); order_stop_stop(rsc, child_rsc, stop | pe_order_implies_first_printed); new_rsc_order(child_rsc, RSC_STOP, rsc, RSC_STOPPED, stopped, data_set); new_rsc_order(child_rsc, RSC_START, rsc, RSC_STARTED, started, data_set); if (group_data->ordered == FALSE) { order_start_start(rsc, child_rsc, start | pe_order_implies_first_printed); if (pcmk_is_set(top->flags, pe_rsc_promotable)) { new_rsc_order(rsc, RSC_PROMOTE, child_rsc, RSC_PROMOTE, start | pe_order_implies_first_printed, data_set); } } else if (last_rsc != NULL) { child_rsc->restart_type = pe_restart_restart; order_start_start(last_rsc, child_rsc, start); order_stop_stop(child_rsc, last_rsc, pe_order_optional | pe_order_restart); if (pcmk_is_set(top->flags, pe_rsc_promotable)) { new_rsc_order(last_rsc, RSC_PROMOTE, child_rsc, RSC_PROMOTE, start, data_set); new_rsc_order(child_rsc, RSC_DEMOTE, last_rsc, RSC_DEMOTE, pe_order_optional, data_set); } } else { /* If anyone in the group is starting, then * pe_order_implies_then will cause _everyone_ in the group * to be sent a start action * But this is safe since starting something that is already * started is required to be "safe" */ int flags = pe_order_none; order_start_start(rsc, child_rsc, flags); if (pcmk_is_set(top->flags, pe_rsc_promotable)) { new_rsc_order(rsc, RSC_PROMOTE, child_rsc, RSC_PROMOTE, flags, data_set); } } /* Look for partially active groups * Make sure they still shut down in sequence */ if (child_rsc->running_on) { if (group_data->ordered && last_rsc && last_rsc->running_on == NULL && last_active && last_active->running_on) { order_stop_stop(child_rsc, last_active, pe_order_optional); } last_active = child_rsc; } last_rsc = child_rsc; } if (group_data->ordered && last_rsc != NULL) { int stop_stop_flags = pe_order_implies_then; int stop_stopped_flags = pe_order_optional; order_stop_stop(rsc, last_rsc, stop_stop_flags); new_rsc_order(last_rsc, RSC_STOP, rsc, RSC_STOPPED, stop_stopped_flags, data_set); if (pcmk_is_set(top->flags, pe_rsc_promotable)) { new_rsc_order(rsc, RSC_DEMOTE, last_rsc, RSC_DEMOTE, stop_stop_flags, data_set); new_rsc_order(last_rsc, RSC_DEMOTE, rsc, RSC_DEMOTED, stop_stopped_flags, data_set); } } } void group_rsc_colocation_lh(pe_resource_t *rsc_lh, pe_resource_t *rsc_rh, - rsc_colocation_t *constraint, + pcmk__colocation_t *constraint, pe_working_set_t *data_set) { GListPtr gIter = NULL; group_variant_data_t *group_data = NULL; if (rsc_lh == NULL) { pe_err("rsc_lh was NULL for %s", constraint->id); return; } else if (rsc_rh == NULL) { pe_err("rsc_rh was NULL for %s", constraint->id); return; } if (constraint->score == 0) { return; } gIter = rsc_lh->children; pe_rsc_trace(rsc_lh, "Processing constraints from %s", rsc_lh->id); get_group_variant_data(group_data, rsc_lh); if (group_data->colocated) { group_data->first_child->cmds->rsc_colocation_lh(group_data->first_child, rsc_rh, constraint, data_set); return; } else if (constraint->score >= INFINITY) { pcmk__config_err("%s: Cannot perform mandatory colocation " "between non-colocated group and %s", rsc_lh->id, rsc_rh->id); return; } for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; child_rsc->cmds->rsc_colocation_lh(child_rsc, rsc_rh, constraint, data_set); } } void group_rsc_colocation_rh(pe_resource_t *rsc_lh, pe_resource_t *rsc_rh, - rsc_colocation_t *constraint, + pcmk__colocation_t *constraint, pe_working_set_t *data_set) { GListPtr gIter = rsc_rh->children; group_variant_data_t *group_data = NULL; get_group_variant_data(group_data, rsc_rh); CRM_CHECK(rsc_lh->variant == pe_native, return); if (constraint->score == 0) { return; } pe_rsc_trace(rsc_rh, "Processing RH %s of constraint %s (LH is %s)", rsc_rh->id, constraint->id, rsc_lh->id); if (pcmk_is_set(rsc_rh->flags, pe_rsc_provisional)) { return; } else if (group_data->colocated && group_data->first_child) { if (constraint->score >= INFINITY) { /* Ensure RHS is _fully_ up before can start LHS */ group_data->last_child->cmds->rsc_colocation_rh(rsc_lh, group_data->last_child, constraint, data_set); } else { /* A partially active RHS is fine */ group_data->first_child->cmds->rsc_colocation_rh(rsc_lh, group_data->first_child, constraint, data_set); } return; } else if (constraint->score >= INFINITY) { pcmk__config_err("%s: Cannot perform mandatory colocation with" " non-colocated group %s", rsc_lh->id, rsc_rh->id); return; } for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; child_rsc->cmds->rsc_colocation_rh(rsc_lh, child_rsc, constraint, data_set); } } enum pe_action_flags group_action_flags(pe_action_t * action, pe_node_t * node) { GListPtr gIter = NULL; enum pe_action_flags flags = (pe_action_optional | pe_action_runnable | pe_action_pseudo); for (gIter = action->rsc->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child = (pe_resource_t *) gIter->data; enum action_tasks task = get_complex_task(child, action->task, TRUE); const char *task_s = task2text(task); pe_action_t *child_action = find_first_action(child->actions, NULL, task_s, node); if (child_action) { enum pe_action_flags child_flags = child->cmds->action_flags(child_action, node); if (pcmk_is_set(flags, pe_action_optional) && !pcmk_is_set(child_flags, pe_action_optional)) { pe_rsc_trace(action->rsc, "%s is mandatory because of %s", action->uuid, child_action->uuid); pe__clear_raw_action_flags(flags, "group action", pe_action_optional); pe__clear_action_flags(action, pe_action_optional); } if (!pcmk__str_eq(task_s, action->task, pcmk__str_casei) && pcmk_is_set(flags, pe_action_runnable) && !pcmk_is_set(child_flags, pe_action_runnable)) { pe_rsc_trace(action->rsc, "%s is not runnable because of %s", action->uuid, child_action->uuid); pe__clear_raw_action_flags(flags, "group action", pe_action_runnable); pe__clear_action_flags(action, pe_action_runnable); } } else if (task != stop_rsc && task != action_demote) { pe_rsc_trace(action->rsc, "%s is not runnable because of %s (not found in %s)", action->uuid, task_s, child->id); pe__clear_raw_action_flags(flags, "group action", pe_action_runnable); } } return flags; } enum pe_graph_flags group_update_actions(pe_action_t *first, pe_action_t *then, pe_node_t *node, enum pe_action_flags flags, enum pe_action_flags filter, enum pe_ordering type, pe_working_set_t *data_set) { GListPtr gIter = then->rsc->children; enum pe_graph_flags changed = pe_graph_none; CRM_ASSERT(then->rsc != NULL); changed |= native_update_actions(first, then, node, flags, filter, type, data_set); for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child = (pe_resource_t *) gIter->data; pe_action_t *child_action = find_first_action(child->actions, NULL, then->task, node); if (child_action) { changed |= child->cmds->update_actions(first, child_action, node, flags, filter, type, data_set); } } return changed; } void group_rsc_location(pe_resource_t *rsc, pe__location_t *constraint) { GListPtr gIter = rsc->children; GListPtr saved = constraint->node_list_rh; GListPtr zero = pcmk__copy_node_list(constraint->node_list_rh, true); gboolean reset_scores = TRUE; group_variant_data_t *group_data = NULL; get_group_variant_data(group_data, rsc); pe_rsc_debug(rsc, "Processing rsc_location %s for %s", constraint->id, rsc->id); native_rsc_location(rsc, constraint); for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; child_rsc->cmds->rsc_location(child_rsc, constraint); if (group_data->colocated && reset_scores) { reset_scores = FALSE; constraint->node_list_rh = zero; } } constraint->node_list_rh = saved; g_list_free_full(zero, free); } void group_expand(pe_resource_t * rsc, pe_working_set_t * data_set) { CRM_CHECK(rsc != NULL, return); pe_rsc_trace(rsc, "Processing actions from %s", rsc->id); native_expand(rsc, data_set); for (GListPtr gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; child_rsc->cmds->expand(child_rsc, data_set); } } GHashTable * pcmk__group_merge_weights(pe_resource_t *rsc, const char *rhs, GHashTable *nodes, const char *attr, float factor, uint32_t flags) { GListPtr gIter = rsc->rsc_cons_lhs; group_variant_data_t *group_data = NULL; get_group_variant_data(group_data, rsc); if (pcmk_is_set(rsc->flags, pe_rsc_merging)) { pe_rsc_info(rsc, "Breaking dependency loop with %s at %s", rsc->id, rhs); return nodes; } pe__set_resource_flags(rsc, pe_rsc_merging); nodes = group_data->first_child->cmds->merge_weights(group_data->first_child, rhs, nodes, attr, factor, flags); for (; gIter != NULL; gIter = gIter->next) { - rsc_colocation_t *constraint = (rsc_colocation_t *) gIter->data; - - if (constraint->score == 0) { - continue; + pcmk__colocation_t *constraint = (pcmk__colocation_t *) gIter->data; + + if (pcmk__colocation_applies(rsc, constraint, false)) { + nodes = pcmk__native_merge_weights(constraint->rsc_lh, rsc->id, + nodes, + constraint->node_attribute, + constraint->score / (float) INFINITY, + flags); } - nodes = pcmk__native_merge_weights(constraint->rsc_lh, rsc->id, nodes, - constraint->node_attribute, - constraint->score / (float) INFINITY, - flags); } pe__clear_resource_flags(rsc, pe_rsc_merging); return nodes; } void group_append_meta(pe_resource_t * rsc, xmlNode * xml) { } diff --git a/lib/pacemaker/pcmk_sched_native.c b/lib/pacemaker/pcmk_sched_native.c index 9196f590d3..c302db6f57 100644 --- a/lib/pacemaker/pcmk_sched_native.c +++ b/lib/pacemaker/pcmk_sched_native.c @@ -1,3497 +1,3493 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include // The controller removes the resource from the CIB, making this redundant // #define DELETE_THEN_REFRESH 1 #define INFINITY_HACK (INFINITY * -100) #define VARIANT_NATIVE 1 #include static void Recurring(pe_resource_t *rsc, pe_action_t *start, pe_node_t *node, pe_working_set_t *data_set); static void RecurringOp(pe_resource_t *rsc, pe_action_t *start, pe_node_t *node, xmlNode *operation, pe_working_set_t *data_set); static void Recurring_Stopped(pe_resource_t *rsc, pe_action_t *start, pe_node_t *node, pe_working_set_t *data_set); static void RecurringOp_Stopped(pe_resource_t *rsc, pe_action_t *start, pe_node_t *node, xmlNode *operation, pe_working_set_t *data_set); void ReloadRsc(pe_resource_t * rsc, pe_node_t *node, pe_working_set_t * data_set); gboolean DeleteRsc(pe_resource_t * rsc, pe_node_t * node, gboolean optional, pe_working_set_t * data_set); gboolean StopRsc(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set); gboolean StartRsc(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set); gboolean DemoteRsc(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set); gboolean PromoteRsc(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set); gboolean RoleError(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set); gboolean NullOp(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set); /* This array says what the *next* role should be when transitioning from one * role to another. For example going from Stopped to Master, the next role is * RSC_ROLE_SLAVE, because the resource must be started before being promoted. * The current state then becomes Started, which is fed into this array again, * giving a next role of RSC_ROLE_MASTER. */ static enum rsc_role_e rsc_state_matrix[RSC_ROLE_MAX][RSC_ROLE_MAX] = { /* Current state Next state*/ /* Unknown Stopped Started Slave Master */ /* Unknown */ { RSC_ROLE_UNKNOWN, RSC_ROLE_STOPPED, RSC_ROLE_STOPPED, RSC_ROLE_STOPPED, RSC_ROLE_STOPPED, }, /* Stopped */ { RSC_ROLE_STOPPED, RSC_ROLE_STOPPED, RSC_ROLE_STARTED, RSC_ROLE_SLAVE, RSC_ROLE_SLAVE, }, /* Started */ { RSC_ROLE_STOPPED, RSC_ROLE_STOPPED, RSC_ROLE_STARTED, RSC_ROLE_SLAVE, RSC_ROLE_MASTER, }, /* Slave */ { RSC_ROLE_STOPPED, RSC_ROLE_STOPPED, RSC_ROLE_STOPPED, RSC_ROLE_SLAVE, RSC_ROLE_MASTER, }, /* Master */ { RSC_ROLE_STOPPED, RSC_ROLE_SLAVE, RSC_ROLE_SLAVE, RSC_ROLE_SLAVE, RSC_ROLE_MASTER, }, }; typedef gboolean (*rsc_transition_fn)(pe_resource_t *rsc, pe_node_t *next, gboolean optional, pe_working_set_t *data_set); // This array picks the function needed to transition from one role to another static rsc_transition_fn rsc_action_matrix[RSC_ROLE_MAX][RSC_ROLE_MAX] = { /* Current state Next state */ /* Unknown Stopped Started Slave Master */ /* Unknown */ { RoleError, StopRsc, RoleError, RoleError, RoleError, }, /* Stopped */ { RoleError, NullOp, StartRsc, StartRsc, RoleError, }, /* Started */ { RoleError, StopRsc, NullOp, NullOp, PromoteRsc, }, /* Slave */ { RoleError, StopRsc, StopRsc, NullOp, PromoteRsc, }, /* Master */ { RoleError, DemoteRsc, DemoteRsc, DemoteRsc, NullOp , }, }; #define clear_node_weights_flags(nw_flags, nw_rsc, flags_to_clear) do { \ flags = pcmk__clear_flags_as(__func__, __LINE__, LOG_TRACE, \ "Node weight", (nw_rsc)->id, (flags), \ (flags_to_clear), #flags_to_clear); \ } while (0) static gboolean native_choose_node(pe_resource_t * rsc, pe_node_t * prefer, pe_working_set_t * data_set) { GListPtr nodes = NULL; pe_node_t *chosen = NULL; pe_node_t *best = NULL; int multiple = 1; int length = 0; gboolean result = FALSE; process_utilization(rsc, &prefer, data_set); if (!pcmk_is_set(rsc->flags, pe_rsc_provisional)) { return rsc->allocated_to ? TRUE : FALSE; } // Sort allowed nodes by weight if (rsc->allowed_nodes) { length = g_hash_table_size(rsc->allowed_nodes); } if (length > 0) { nodes = g_hash_table_get_values(rsc->allowed_nodes); nodes = sort_nodes_by_weight(nodes, pe__current_node(rsc), data_set); // First node in sorted list has the best score best = g_list_nth_data(nodes, 0); } if (prefer && nodes) { chosen = g_hash_table_lookup(rsc->allowed_nodes, prefer->details->id); if (chosen == NULL) { pe_rsc_trace(rsc, "Preferred node %s for %s was unknown", prefer->details->uname, rsc->id); /* Favor the preferred node as long as its weight is at least as good as * the best allowed node's. * * An alternative would be to favor the preferred node even if the best * node is better, when the best node's weight is less than INFINITY. */ } else if ((chosen->weight < 0) || (chosen->weight < best->weight)) { pe_rsc_trace(rsc, "Preferred node %s for %s was unsuitable", chosen->details->uname, rsc->id); chosen = NULL; } else if (!can_run_resources(chosen)) { pe_rsc_trace(rsc, "Preferred node %s for %s was unavailable", chosen->details->uname, rsc->id); chosen = NULL; } else { pe_rsc_trace(rsc, "Chose preferred node %s for %s (ignoring %d candidates)", chosen->details->uname, rsc->id, length); } } if ((chosen == NULL) && nodes) { /* Either there is no preferred node, or the preferred node is not * available, but there are other nodes allowed to run the resource. */ chosen = best; pe_rsc_trace(rsc, "Chose node %s for %s from %d candidates", chosen ? chosen->details->uname : "", rsc->id, length); if (!pe_rsc_is_unique_clone(rsc->parent) && chosen && (chosen->weight > 0) && can_run_resources(chosen)) { /* If the resource is already running on a node, prefer that node if * it is just as good as the chosen node. * * We don't do this for unique clone instances, because * distribute_children() has already assigned instances to their * running nodes when appropriate, and if we get here, we don't want * remaining unallocated instances to prefer a node that's already * running another instance. */ pe_node_t *running = pe__current_node(rsc); if (running && (can_run_resources(running) == FALSE)) { pe_rsc_trace(rsc, "Current node for %s (%s) can't run resources", rsc->id, running->details->uname); } else if (running) { for (GList *iter = nodes->next; iter; iter = iter->next) { pe_node_t *tmp = (pe_node_t *) iter->data; if (tmp->weight != chosen->weight) { // The nodes are sorted by weight, so no more are equal break; } if (tmp->details == running->details) { // Scores are equal, so prefer the current node chosen = tmp; } multiple++; } } } } if (multiple > 1) { static char score[33]; int log_level = (chosen->weight >= INFINITY)? LOG_WARNING : LOG_INFO; score2char_stack(chosen->weight, score, sizeof(score)); do_crm_log(log_level, "Chose node %s for %s from %d nodes with score %s", chosen->details->uname, rsc->id, multiple, score); } result = native_assign_node(rsc, nodes, chosen, FALSE); g_list_free(nodes); return result; } /*! * \internal * \brief Find score of highest-scored node that matches colocation attribute * * \param[in] rsc Resource whose allowed nodes should be searched * \param[in] attr Colocation attribute name (must not be NULL) * \param[in] value Colocation attribute value to require */ static int best_node_score_matching_attr(const pe_resource_t *rsc, const char *attr, const char *value) { GHashTableIter iter; pe_node_t *node = NULL; int best_score = -INFINITY; const char *best_node = NULL; // Find best allowed node with matching attribute g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) { if ((node->weight > best_score) && can_run_resources(node) && pcmk__str_eq(value, pe_node_attribute_raw(node, attr), pcmk__str_casei)) { best_score = node->weight; best_node = node->details->uname; } } if (!pcmk__str_eq(attr, CRM_ATTR_UNAME, pcmk__str_casei)) { if (best_node == NULL) { crm_info("No allowed node for %s matches node attribute %s=%s", rsc->id, attr, value); } else { crm_info("Allowed node %s for %s had best score (%d) " "of those matching node attribute %s=%s", best_node, rsc->id, best_score, attr, value); } } return best_score; } /*! * \internal * \brief Add resource's colocation matches to current node allocation scores * * For each node in a given table, if any of a given resource's allowed nodes * have a matching value for the colocation attribute, add the highest of those * nodes' scores to the node's score. * * \param[in,out] nodes Hash table of nodes with allocation scores so far * \param[in] rsc Resource whose allowed nodes should be compared * \param[in] attr Colocation attribute that must match (NULL for default) * \param[in] factor Factor by which to multiply scores being added * \param[in] only_positive Whether to add only positive scores */ static void add_node_scores_matching_attr(GHashTable *nodes, const pe_resource_t *rsc, const char *attr, float factor, bool only_positive) { GHashTableIter iter; pe_node_t *node = NULL; if (attr == NULL) { attr = CRM_ATTR_UNAME; } // Iterate through each node g_hash_table_iter_init(&iter, nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { float weight_f = 0; int weight = 0; int score = 0; int new_score = 0; score = best_node_score_matching_attr(rsc, attr, pe_node_attribute_raw(node, attr)); if ((factor < 0) && (score < 0)) { /* Negative preference for a node with a negative score * should not become a positive preference. * * @TODO Consider filtering only if weight is -INFINITY */ crm_trace("%s: Filtering %d + %f * %d (double negative disallowed)", node->details->uname, node->weight, factor, score); continue; } if (node->weight == INFINITY_HACK) { crm_trace("%s: Filtering %d + %f * %d (node was marked unusable)", node->details->uname, node->weight, factor, score); continue; } weight_f = factor * score; // Round the number; see http://c-faq.com/fp/round.html weight = (int) ((weight_f < 0)? (weight_f - 0.5) : (weight_f + 0.5)); /* Small factors can obliterate the small scores that are often actually * used in configurations. If the score and factor are nonzero, ensure * that the result is nonzero as well. */ if ((weight == 0) && (score != 0)) { if (factor > 0.0) { weight = 1; } else if (factor < 0.0) { weight = -1; } } new_score = pe__add_scores(weight, node->weight); if (only_positive && (new_score < 0) && (node->weight > 0)) { crm_trace("%s: Filtering %d + %f * %d = %d " "(negative disallowed, marking node unusable)", node->details->uname, node->weight, factor, score, new_score); node->weight = INFINITY_HACK; continue; } if (only_positive && (new_score < 0) && (node->weight == 0)) { crm_trace("%s: Filtering %d + %f * %d = %d (negative disallowed)", node->details->uname, node->weight, factor, score, new_score); continue; } crm_trace("%s: %d + %f * %d = %d", node->details->uname, node->weight, factor, score, new_score); node->weight = new_score; } } static inline bool is_nonempty_group(pe_resource_t *rsc) { return rsc && (rsc->variant == pe_group) && (rsc->children != NULL); } /*! * \internal * \brief Incorporate colocation constraint scores into node weights * * \param[in,out] rsc Resource being placed * \param[in] rhs ID of 'with' resource * \param[in,out] nodes Nodes, with scores as of this point * \param[in] attr Colocation attribute (ID by default) * \param[in] factor Incorporate scores multiplied by this factor * \param[in] flags Bitmask of enum pe_weights values * * \return Nodes, with scores modified by this constraint * \note This function assumes ownership of the nodes argument. The caller * should free the returned copy rather than the original. */ GHashTable * pcmk__native_merge_weights(pe_resource_t *rsc, const char *rhs, GHashTable *nodes, const char *attr, float factor, uint32_t flags) { GHashTable *work = NULL; // Avoid infinite recursion if (pcmk_is_set(rsc->flags, pe_rsc_merging)) { pe_rsc_info(rsc, "%s: Breaking dependency loop at %s", rhs, rsc->id); return nodes; } pe__set_resource_flags(rsc, pe_rsc_merging); if (pcmk_is_set(flags, pe_weights_init)) { if (is_nonempty_group(rsc)) { GList *last = g_list_last(rsc->children); pe_resource_t *last_rsc = last->data; pe_rsc_trace(rsc, "%s: Merging scores from group %s " "using last member %s (at %.6f)", rhs, rsc->id, last_rsc->id, factor); work = pcmk__native_merge_weights(last_rsc, rhs, NULL, attr, factor, flags); } else { work = pcmk__copy_node_table(rsc->allowed_nodes); } clear_node_weights_flags(flags, rsc, pe_weights_init); } else if (is_nonempty_group(rsc)) { /* The first member of the group will recursively incorporate any * constraints involving other members (including the group internal * colocation). * * @TODO The indirect colocations from the dependent group's other * members will be incorporated at full strength rather than by * factor, so the group's combined stickiness will be treated as * (factor + (#members - 1)) * stickiness. It is questionable what * the right approach should be. */ pe_rsc_trace(rsc, "%s: Merging scores from first member of group %s " "(at %.6f)", rhs, rsc->id, factor); work = pcmk__copy_node_table(nodes); work = pcmk__native_merge_weights(rsc->children->data, rhs, work, attr, factor, flags); } else { pe_rsc_trace(rsc, "%s: Merging scores from %s (at %.6f)", rhs, rsc->id, factor); work = pcmk__copy_node_table(nodes); add_node_scores_matching_attr(work, rsc, attr, factor, pcmk_is_set(flags, pe_weights_positive)); } if (can_run_any(work)) { GListPtr gIter = NULL; int multiplier = (factor < 0)? -1 : 1; if (pcmk_is_set(flags, pe_weights_forward)) { gIter = rsc->rsc_cons; pe_rsc_trace(rsc, "Checking additional %d optional '%s with' constraints", g_list_length(gIter), rsc->id); } else if (is_nonempty_group(rsc)) { pe_resource_t *last_rsc = g_list_last(rsc->children)->data; gIter = last_rsc->rsc_cons_lhs; pe_rsc_trace(rsc, "Checking additional %d optional 'with group %s' " "constraints using last member %s", g_list_length(gIter), rsc->id, last_rsc->id); } else { gIter = rsc->rsc_cons_lhs; pe_rsc_trace(rsc, "Checking additional %d optional 'with %s' constraints", g_list_length(gIter), rsc->id); } for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *other = NULL; - rsc_colocation_t *constraint = (rsc_colocation_t *) gIter->data; + pcmk__colocation_t *constraint = (pcmk__colocation_t *) gIter->data; if (constraint->score == 0) { continue; } if (pcmk_is_set(flags, pe_weights_forward)) { other = constraint->rsc_rh; } else { other = constraint->rsc_lh; } pe_rsc_trace(rsc, "Optionally merging score of '%s' constraint (%s with %s)", constraint->id, constraint->rsc_lh->id, constraint->rsc_rh->id); work = pcmk__native_merge_weights(other, rhs, work, constraint->node_attribute, multiplier * constraint->score / (float) INFINITY, flags|pe_weights_rollback); pe__show_node_weights(true, NULL, rhs, work); } } else if (pcmk_is_set(flags, pe_weights_rollback)) { pe_rsc_info(rsc, "%s: Rolling back optional scores from %s", rhs, rsc->id); g_hash_table_destroy(work); pe__clear_resource_flags(rsc, pe_rsc_merging); return nodes; } if (pcmk_is_set(flags, pe_weights_positive)) { pe_node_t *node = NULL; GHashTableIter iter; g_hash_table_iter_init(&iter, work); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { if (node->weight == INFINITY_HACK) { node->weight = 1; } } } if (nodes) { g_hash_table_destroy(nodes); } pe__clear_resource_flags(rsc, pe_rsc_merging); return work; } static inline bool node_has_been_unfenced(pe_node_t *node) { const char *unfenced = pe_node_attribute_raw(node, CRM_ATTR_UNFENCED); return !pcmk__str_eq(unfenced, "0", pcmk__str_null_matches); } static inline bool is_unfence_device(pe_resource_t *rsc, pe_working_set_t *data_set) { return pcmk_is_set(rsc->flags, pe_rsc_fence_device) && pcmk_is_set(data_set->flags, pe_flag_enable_unfencing); } pe_node_t * pcmk__native_allocate(pe_resource_t *rsc, pe_node_t *prefer, pe_working_set_t *data_set) { GListPtr gIter = NULL; if (rsc->parent && !pcmk_is_set(rsc->parent->flags, pe_rsc_allocating)) { /* never allocate children on their own */ pe_rsc_debug(rsc, "Escalating allocation of %s to its parent: %s", rsc->id, rsc->parent->id); rsc->parent->cmds->allocate(rsc->parent, prefer, data_set); } if (!pcmk_is_set(rsc->flags, pe_rsc_provisional)) { return rsc->allocated_to; } if (pcmk_is_set(rsc->flags, pe_rsc_allocating)) { pe_rsc_debug(rsc, "Dependency loop detected involving %s", rsc->id); return NULL; } pe__set_resource_flags(rsc, pe_rsc_allocating); pe__show_node_weights(true, rsc, "Pre-alloc", rsc->allowed_nodes); for (gIter = rsc->rsc_cons; gIter != NULL; gIter = gIter->next) { - rsc_colocation_t *constraint = (rsc_colocation_t *) gIter->data; + pcmk__colocation_t *constraint = (pcmk__colocation_t *) gIter->data; GHashTable *archive = NULL; pe_resource_t *rsc_rh = constraint->rsc_rh; if (constraint->score == 0) { continue; } if (constraint->role_lh >= RSC_ROLE_MASTER || (constraint->score < 0 && constraint->score > -INFINITY)) { archive = pcmk__copy_node_table(rsc->allowed_nodes); } pe_rsc_trace(rsc, "%s: Allocating %s first (constraint=%s score=%d role=%s)", rsc->id, rsc_rh->id, constraint->id, constraint->score, role2text(constraint->role_lh)); rsc_rh->cmds->allocate(rsc_rh, NULL, data_set); rsc->cmds->rsc_colocation_lh(rsc, rsc_rh, constraint, data_set); if (archive && can_run_any(rsc->allowed_nodes) == FALSE) { pe_rsc_info(rsc, "%s: Rolling back scores from %s", rsc->id, rsc_rh->id); g_hash_table_destroy(rsc->allowed_nodes); rsc->allowed_nodes = archive; archive = NULL; } if (archive) { g_hash_table_destroy(archive); } } pe__show_node_weights(true, rsc, "Post-coloc", rsc->allowed_nodes); for (gIter = rsc->rsc_cons_lhs; gIter != NULL; gIter = gIter->next) { - rsc_colocation_t *constraint = (rsc_colocation_t *) gIter->data; + pcmk__colocation_t *constraint = (pcmk__colocation_t *) gIter->data; - if (constraint->score == 0) { - continue; + if (pcmk__colocation_applies(rsc, constraint, false)) { + rsc->allowed_nodes = constraint->rsc_lh->cmds->merge_weights(constraint->rsc_lh, + rsc->id, rsc->allowed_nodes, constraint->node_attribute, + constraint->score / (float) INFINITY, pe_weights_rollback); } - pe_rsc_trace(rsc, "Merging score of '%s' constraint (%s with %s)", - constraint->id, constraint->rsc_lh->id, - constraint->rsc_rh->id); - rsc->allowed_nodes = - constraint->rsc_lh->cmds->merge_weights(constraint->rsc_lh, rsc->id, rsc->allowed_nodes, - constraint->node_attribute, - (float)constraint->score / INFINITY, - pe_weights_rollback); } if (rsc->next_role == RSC_ROLE_STOPPED) { pe_rsc_trace(rsc, "Making sure %s doesn't get allocated", rsc->id); /* make sure it doesn't come up again */ resource_location(rsc, NULL, -INFINITY, XML_RSC_ATTR_TARGET_ROLE, data_set); } else if(rsc->next_role > rsc->role && !pcmk_is_set(data_set->flags, pe_flag_have_quorum) && data_set->no_quorum_policy == no_quorum_freeze) { crm_notice("Resource %s cannot be elevated from %s to %s: no-quorum-policy=freeze", rsc->id, role2text(rsc->role), role2text(rsc->next_role)); rsc->next_role = rsc->role; } pe__show_node_weights(!show_scores, rsc, __func__, rsc->allowed_nodes); if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled) && !pcmk_is_set(data_set->flags, pe_flag_have_stonith_resource)) { pe__clear_resource_flags(rsc, pe_rsc_managed); } if (!pcmk_is_set(rsc->flags, pe_rsc_managed)) { const char *reason = NULL; pe_node_t *assign_to = NULL; rsc->next_role = rsc->role; assign_to = pe__current_node(rsc); if (assign_to == NULL) { reason = "inactive"; } else if (rsc->role == RSC_ROLE_MASTER) { reason = "master"; } else if (pcmk_is_set(rsc->flags, pe_rsc_failed)) { reason = "failed"; } else { reason = "active"; } pe_rsc_info(rsc, "Unmanaged resource %s allocated to %s: %s", rsc->id, (assign_to? assign_to->details->uname : "no node"), reason); native_assign_node(rsc, NULL, assign_to, TRUE); } else if (pcmk_is_set(data_set->flags, pe_flag_stop_everything)) { pe_rsc_debug(rsc, "Forcing %s to stop", rsc->id); native_assign_node(rsc, NULL, NULL, TRUE); } else if (pcmk_is_set(rsc->flags, pe_rsc_provisional) && native_choose_node(rsc, prefer, data_set)) { pe_rsc_trace(rsc, "Allocated resource %s to %s", rsc->id, rsc->allocated_to->details->uname); } else if (rsc->allocated_to == NULL) { if (!pcmk_is_set(rsc->flags, pe_rsc_orphan)) { pe_rsc_info(rsc, "Resource %s cannot run anywhere", rsc->id); } else if (rsc->running_on != NULL) { pe_rsc_info(rsc, "Stopping orphan resource %s", rsc->id); } } else { pe_rsc_debug(rsc, "Pre-Allocated resource %s to %s", rsc->id, rsc->allocated_to->details->uname); } pe__clear_resource_flags(rsc, pe_rsc_allocating); if (rsc->is_remote_node) { pe_node_t *remote_node = pe_find_node(data_set->nodes, rsc->id); CRM_ASSERT(remote_node != NULL); if (rsc->allocated_to && rsc->next_role != RSC_ROLE_STOPPED) { crm_trace("Setting Pacemaker Remote node %s to ONLINE", remote_node->details->id); remote_node->details->online = TRUE; /* We shouldn't consider an unseen remote-node unclean if we are going * to try and connect to it. Otherwise we get an unnecessary fence */ if (remote_node->details->unseen == TRUE) { remote_node->details->unclean = FALSE; } } else { crm_trace("Setting Pacemaker Remote node %s to SHUTDOWN (next role %s, %sallocated)", remote_node->details->id, role2text(rsc->next_role), (rsc->allocated_to? "" : "un")); remote_node->details->shutdown = TRUE; } } return rsc->allocated_to; } static gboolean is_op_dup(pe_resource_t *rsc, const char *name, guint interval_ms) { gboolean dup = FALSE; const char *id = NULL; const char *value = NULL; xmlNode *operation = NULL; guint interval2_ms = 0; CRM_ASSERT(rsc); for (operation = pcmk__xe_first_child(rsc->ops_xml); operation != NULL; operation = pcmk__xe_next(operation)) { if (pcmk__str_eq((const char *)operation->name, "op", pcmk__str_none)) { value = crm_element_value(operation, "name"); if (!pcmk__str_eq(value, name, pcmk__str_casei)) { continue; } value = crm_element_value(operation, XML_LRM_ATTR_INTERVAL); interval2_ms = crm_parse_interval_spec(value); if (interval_ms != interval2_ms) { continue; } if (id == NULL) { id = ID(operation); } else { pcmk__config_err("Operation %s is duplicate of %s (do not use " "same name and interval combination more " "than once per resource)", ID(operation), id); dup = TRUE; } } } return dup; } static bool op_cannot_recur(const char *name) { return pcmk__strcase_any_of(name, RSC_STOP, RSC_START, RSC_DEMOTE, RSC_PROMOTE, NULL); } static void RecurringOp(pe_resource_t * rsc, pe_action_t * start, pe_node_t * node, xmlNode * operation, pe_working_set_t * data_set) { char *key = NULL; const char *name = NULL; const char *role = NULL; const char *interval_spec = NULL; const char *node_uname = node? node->details->uname : "n/a"; guint interval_ms = 0; pe_action_t *mon = NULL; gboolean is_optional = TRUE; GListPtr possible_matches = NULL; CRM_ASSERT(rsc); /* Only process for the operations without role="Stopped" */ role = crm_element_value(operation, "role"); if (role && text2role(role) == RSC_ROLE_STOPPED) { return; } interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL); interval_ms = crm_parse_interval_spec(interval_spec); if (interval_ms == 0) { return; } name = crm_element_value(operation, "name"); if (is_op_dup(rsc, name, interval_ms)) { crm_trace("Not creating duplicate recurring action %s for %dms %s", ID(operation), interval_ms, name); return; } if (op_cannot_recur(name)) { pcmk__config_err("Ignoring %s because action '%s' cannot be recurring", ID(operation), name); return; } key = pcmk__op_key(rsc->id, name, interval_ms); if (find_rsc_op_entry(rsc, key) == NULL) { crm_trace("Not creating recurring action %s for disabled resource %s", ID(operation), rsc->id); free(key); return; } pe_rsc_trace(rsc, "Creating recurring action %s for %s in role %s on %s", ID(operation), rsc->id, role2text(rsc->next_role), node_uname); if (start != NULL) { pe_rsc_trace(rsc, "Marking %s %s due to %s", key, pcmk_is_set(start->flags, pe_action_optional)? "optional" : "mandatory", start->uuid); is_optional = (rsc->cmds->action_flags(start, NULL) & pe_action_optional); } else { pe_rsc_trace(rsc, "Marking %s optional", key); is_optional = TRUE; } /* start a monitor for an already active resource */ possible_matches = find_actions_exact(rsc->actions, key, node); if (possible_matches == NULL) { is_optional = FALSE; pe_rsc_trace(rsc, "Marking %s mandatory: not active", key); } else { GListPtr gIter = NULL; for (gIter = possible_matches; gIter != NULL; gIter = gIter->next) { pe_action_t *op = (pe_action_t *) gIter->data; if (pcmk_is_set(op->flags, pe_action_reschedule)) { is_optional = FALSE; break; } } g_list_free(possible_matches); } if ((rsc->next_role == RSC_ROLE_MASTER && role == NULL) || (role != NULL && text2role(role) != rsc->next_role)) { int log_level = LOG_TRACE; const char *result = "Ignoring"; if (is_optional) { char *after_key = NULL; pe_action_t *cancel_op = NULL; // It's running, so cancel it log_level = LOG_INFO; result = "Cancelling"; cancel_op = pe_cancel_op(rsc, name, interval_ms, node, data_set); switch (rsc->role) { case RSC_ROLE_SLAVE: case RSC_ROLE_STARTED: if (rsc->next_role == RSC_ROLE_MASTER) { after_key = promote_key(rsc); } else if (rsc->next_role == RSC_ROLE_STOPPED) { after_key = stop_key(rsc); } break; case RSC_ROLE_MASTER: after_key = demote_key(rsc); break; default: break; } if (after_key) { custom_action_order(rsc, NULL, cancel_op, rsc, after_key, NULL, pe_order_runnable_left, data_set); } } do_crm_log(log_level, "%s action %s (%s vs. %s)", result, key, role ? role : role2text(RSC_ROLE_SLAVE), role2text(rsc->next_role)); free(key); return; } mon = custom_action(rsc, key, name, node, is_optional, TRUE, data_set); key = mon->uuid; if (is_optional) { pe_rsc_trace(rsc, "%s\t %s (optional)", node_uname, mon->uuid); } if ((start == NULL) || !pcmk_is_set(start->flags, pe_action_runnable)) { pe_rsc_debug(rsc, "%s\t %s (cancelled : start un-runnable)", node_uname, mon->uuid); update_action_flags(mon, pe_action_runnable | pe_action_clear, __func__, __LINE__); } else if (node == NULL || node->details->online == FALSE || node->details->unclean) { pe_rsc_debug(rsc, "%s\t %s (cancelled : no node available)", node_uname, mon->uuid); update_action_flags(mon, pe_action_runnable | pe_action_clear, __func__, __LINE__); } else if (!pcmk_is_set(mon->flags, pe_action_optional)) { pe_rsc_info(rsc, " Start recurring %s (%us) for %s on %s", mon->task, interval_ms / 1000, rsc->id, node_uname); } if (rsc->next_role == RSC_ROLE_MASTER) { char *running_master = crm_itoa(PCMK_OCF_RUNNING_MASTER); add_hash_param(mon->meta, XML_ATTR_TE_TARGET_RC, running_master); free(running_master); } if ((node == NULL) || pcmk_is_set(rsc->flags, pe_rsc_managed)) { custom_action_order(rsc, start_key(rsc), NULL, NULL, strdup(key), mon, pe_order_implies_then | pe_order_runnable_left, data_set); custom_action_order(rsc, reload_key(rsc), NULL, NULL, strdup(key), mon, pe_order_implies_then | pe_order_runnable_left, data_set); if (rsc->next_role == RSC_ROLE_MASTER) { custom_action_order(rsc, promote_key(rsc), NULL, rsc, NULL, mon, pe_order_optional | pe_order_runnable_left, data_set); } else if (rsc->role == RSC_ROLE_MASTER) { custom_action_order(rsc, demote_key(rsc), NULL, rsc, NULL, mon, pe_order_optional | pe_order_runnable_left, data_set); } } } static void Recurring(pe_resource_t * rsc, pe_action_t * start, pe_node_t * node, pe_working_set_t * data_set) { if (!pcmk_is_set(rsc->flags, pe_rsc_maintenance) && (node == NULL || node->details->maintenance == FALSE)) { xmlNode *operation = NULL; for (operation = pcmk__xe_first_child(rsc->ops_xml); operation != NULL; operation = pcmk__xe_next(operation)) { if (pcmk__str_eq((const char *)operation->name, "op", pcmk__str_none)) { RecurringOp(rsc, start, node, operation, data_set); } } } } static void RecurringOp_Stopped(pe_resource_t * rsc, pe_action_t * start, pe_node_t * node, xmlNode * operation, pe_working_set_t * data_set) { char *key = NULL; const char *name = NULL; const char *role = NULL; const char *interval_spec = NULL; const char *node_uname = node? node->details->uname : "n/a"; guint interval_ms = 0; GListPtr possible_matches = NULL; GListPtr gIter = NULL; /* Only process for the operations with role="Stopped" */ role = crm_element_value(operation, "role"); if (role == NULL || text2role(role) != RSC_ROLE_STOPPED) { return; } interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL); interval_ms = crm_parse_interval_spec(interval_spec); if (interval_ms == 0) { return; } name = crm_element_value(operation, "name"); if (is_op_dup(rsc, name, interval_ms)) { crm_trace("Not creating duplicate recurring action %s for %dms %s", ID(operation), interval_ms, name); return; } if (op_cannot_recur(name)) { pcmk__config_err("Ignoring %s because action '%s' cannot be recurring", ID(operation), name); return; } key = pcmk__op_key(rsc->id, name, interval_ms); if (find_rsc_op_entry(rsc, key) == NULL) { crm_trace("Not creating recurring action %s for disabled resource %s", ID(operation), rsc->id); free(key); return; } // @TODO add support if (!pcmk_is_set(rsc->flags, pe_rsc_unique)) { crm_notice("Ignoring %s (recurring monitors for Stopped role are " "not supported for anonymous clones)", ID(operation)); return; } pe_rsc_trace(rsc, "Creating recurring action %s for %s in role %s on nodes where it should not be running", ID(operation), rsc->id, role2text(rsc->next_role)); /* if the monitor exists on the node where the resource will be running, cancel it */ if (node != NULL) { possible_matches = find_actions_exact(rsc->actions, key, node); if (possible_matches) { pe_action_t *cancel_op = NULL; g_list_free(possible_matches); cancel_op = pe_cancel_op(rsc, name, interval_ms, node, data_set); if (rsc->next_role == RSC_ROLE_STARTED || rsc->next_role == RSC_ROLE_SLAVE) { /* rsc->role == RSC_ROLE_STOPPED: cancel the monitor before start */ /* rsc->role == RSC_ROLE_STARTED: for a migration, cancel the monitor on the target node before start */ custom_action_order(rsc, NULL, cancel_op, rsc, start_key(rsc), NULL, pe_order_runnable_left, data_set); } pe_rsc_info(rsc, "Cancel action %s (%s vs. %s) on %s", key, role, role2text(rsc->next_role), node_uname); } } for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { pe_node_t *stop_node = (pe_node_t *) gIter->data; const char *stop_node_uname = stop_node->details->uname; gboolean is_optional = TRUE; gboolean probe_is_optional = TRUE; gboolean stop_is_optional = TRUE; pe_action_t *stopped_mon = NULL; char *rc_inactive = NULL; GListPtr probe_complete_ops = NULL; GListPtr stop_ops = NULL; GListPtr local_gIter = NULL; if (node && pcmk__str_eq(stop_node_uname, node_uname, pcmk__str_casei)) { continue; } pe_rsc_trace(rsc, "Creating recurring action %s for %s on %s", ID(operation), rsc->id, crm_str(stop_node_uname)); /* start a monitor for an already stopped resource */ possible_matches = find_actions_exact(rsc->actions, key, stop_node); if (possible_matches == NULL) { pe_rsc_trace(rsc, "Marking %s mandatory on %s: not active", key, crm_str(stop_node_uname)); is_optional = FALSE; } else { pe_rsc_trace(rsc, "Marking %s optional on %s: already active", key, crm_str(stop_node_uname)); is_optional = TRUE; g_list_free(possible_matches); } stopped_mon = custom_action(rsc, strdup(key), name, stop_node, is_optional, TRUE, data_set); rc_inactive = crm_itoa(PCMK_OCF_NOT_RUNNING); add_hash_param(stopped_mon->meta, XML_ATTR_TE_TARGET_RC, rc_inactive); free(rc_inactive); if (pcmk_is_set(rsc->flags, pe_rsc_managed)) { GList *probes = pe__resource_actions(rsc, stop_node, RSC_STATUS, FALSE); GListPtr pIter = NULL; for (pIter = probes; pIter != NULL; pIter = pIter->next) { pe_action_t *probe = (pe_action_t *) pIter->data; order_actions(probe, stopped_mon, pe_order_runnable_left); crm_trace("%s then %s on %s", probe->uuid, stopped_mon->uuid, stop_node->details->uname); } g_list_free(probes); } if (probe_complete_ops) { g_list_free(probe_complete_ops); } stop_ops = pe__resource_actions(rsc, stop_node, RSC_STOP, TRUE); for (local_gIter = stop_ops; local_gIter != NULL; local_gIter = local_gIter->next) { pe_action_t *stop = (pe_action_t *) local_gIter->data; if (!pcmk_is_set(stop->flags, pe_action_optional)) { stop_is_optional = FALSE; } if (!pcmk_is_set(stop->flags, pe_action_runnable)) { crm_debug("%s\t %s (cancelled : stop un-runnable)", crm_str(stop_node_uname), stopped_mon->uuid); update_action_flags(stopped_mon, pe_action_runnable | pe_action_clear, __func__, __LINE__); } if (pcmk_is_set(rsc->flags, pe_rsc_managed)) { custom_action_order(rsc, stop_key(rsc), stop, NULL, strdup(key), stopped_mon, pe_order_implies_then | pe_order_runnable_left, data_set); } } if (stop_ops) { g_list_free(stop_ops); } if (is_optional == FALSE && probe_is_optional && stop_is_optional && !pcmk_is_set(rsc->flags, pe_rsc_managed)) { pe_rsc_trace(rsc, "Marking %s optional on %s due to unmanaged", key, crm_str(stop_node_uname)); update_action_flags(stopped_mon, pe_action_optional, __func__, __LINE__); } if (pcmk_is_set(stopped_mon->flags, pe_action_optional)) { pe_rsc_trace(rsc, "%s\t %s (optional)", crm_str(stop_node_uname), stopped_mon->uuid); } if (stop_node->details->online == FALSE || stop_node->details->unclean) { pe_rsc_debug(rsc, "%s\t %s (cancelled : no node available)", crm_str(stop_node_uname), stopped_mon->uuid); update_action_flags(stopped_mon, pe_action_runnable | pe_action_clear, __func__, __LINE__); } if (pcmk_is_set(stopped_mon->flags, pe_action_runnable) && !pcmk_is_set(stopped_mon->flags, pe_action_optional)) { crm_notice(" Start recurring %s (%us) for %s on %s", stopped_mon->task, interval_ms / 1000, rsc->id, crm_str(stop_node_uname)); } } free(key); } static void Recurring_Stopped(pe_resource_t * rsc, pe_action_t * start, pe_node_t * node, pe_working_set_t * data_set) { if (!pcmk_is_set(rsc->flags, pe_rsc_maintenance) && (node == NULL || node->details->maintenance == FALSE)) { xmlNode *operation = NULL; for (operation = pcmk__xe_first_child(rsc->ops_xml); operation != NULL; operation = pcmk__xe_next(operation)) { if (pcmk__str_eq((const char *)operation->name, "op", pcmk__str_none)) { RecurringOp_Stopped(rsc, start, node, operation, data_set); } } } } static void handle_migration_actions(pe_resource_t * rsc, pe_node_t *current, pe_node_t *chosen, pe_working_set_t * data_set) { pe_action_t *migrate_to = NULL; pe_action_t *migrate_from = NULL; pe_action_t *start = NULL; pe_action_t *stop = NULL; gboolean partial = rsc->partial_migration_target ? TRUE : FALSE; pe_rsc_trace(rsc, "Processing migration actions %s moving from %s to %s . partial migration = %s", rsc->id, current->details->id, chosen->details->id, partial ? "TRUE" : "FALSE"); start = start_action(rsc, chosen, TRUE); stop = stop_action(rsc, current, TRUE); if (partial == FALSE) { migrate_to = custom_action(rsc, pcmk__op_key(rsc->id, RSC_MIGRATE, 0), RSC_MIGRATE, current, TRUE, TRUE, data_set); } migrate_from = custom_action(rsc, pcmk__op_key(rsc->id, RSC_MIGRATED, 0), RSC_MIGRATED, chosen, TRUE, TRUE, data_set); if ((migrate_to && migrate_from) || (migrate_from && partial)) { pe__set_action_flags(start, pe_action_migrate_runnable); pe__set_action_flags(stop, pe_action_migrate_runnable); // This is easier than trying to delete it from the graph update_action_flags(start, pe_action_pseudo, __func__, __LINE__); /* order probes before migrations */ if (partial) { pe__set_action_flags(migrate_from, pe_action_migrate_runnable); migrate_from->needs = start->needs; custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_STATUS, 0), NULL, rsc, pcmk__op_key(rsc->id, RSC_MIGRATED, 0), NULL, pe_order_optional, data_set); } else { pe__set_action_flags(migrate_from, pe_action_migrate_runnable); pe__set_action_flags(migrate_to, pe_action_migrate_runnable); migrate_to->needs = start->needs; custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_STATUS, 0), NULL, rsc, pcmk__op_key(rsc->id, RSC_MIGRATE, 0), NULL, pe_order_optional, data_set); custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_MIGRATE, 0), NULL, rsc, pcmk__op_key(rsc->id, RSC_MIGRATED, 0), NULL, pe_order_optional|pe_order_implies_first_migratable, data_set); } custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_MIGRATED, 0), NULL, rsc, pcmk__op_key(rsc->id, RSC_STOP, 0), NULL, pe_order_optional|pe_order_implies_first_migratable, data_set); custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_MIGRATED, 0), NULL, rsc, pcmk__op_key(rsc->id, RSC_START, 0), NULL, pe_order_optional|pe_order_implies_first_migratable|pe_order_pseudo_left, data_set); } if (migrate_to) { add_hash_param(migrate_to->meta, XML_LRM_ATTR_MIGRATE_SOURCE, current->details->uname); add_hash_param(migrate_to->meta, XML_LRM_ATTR_MIGRATE_TARGET, chosen->details->uname); /* Pacemaker Remote connections don't require pending to be recorded in * the CIB. We can reduce CIB writes by not setting PENDING for them. */ if (rsc->is_remote_node == FALSE) { /* migrate_to takes place on the source node, but can * have an effect on the target node depending on how * the agent is written. Because of this, we have to maintain * a record that the migrate_to occurred, in case the source node * loses membership while the migrate_to action is still in-flight. */ add_hash_param(migrate_to->meta, XML_OP_ATTR_PENDING, "true"); } } if (migrate_from) { add_hash_param(migrate_from->meta, XML_LRM_ATTR_MIGRATE_SOURCE, current->details->uname); add_hash_param(migrate_from->meta, XML_LRM_ATTR_MIGRATE_TARGET, chosen->details->uname); } } void native_create_actions(pe_resource_t * rsc, pe_working_set_t * data_set) { pe_action_t *start = NULL; pe_node_t *chosen = NULL; pe_node_t *current = NULL; gboolean need_stop = FALSE; bool need_promote = FALSE; gboolean is_moving = FALSE; gboolean allow_migrate = pcmk_is_set(rsc->flags, pe_rsc_allow_migrate)? TRUE : FALSE; GListPtr gIter = NULL; unsigned int num_all_active = 0; unsigned int num_clean_active = 0; bool multiply_active = FALSE; enum rsc_role_e role = RSC_ROLE_UNKNOWN; enum rsc_role_e next_role = RSC_ROLE_UNKNOWN; CRM_ASSERT(rsc); chosen = rsc->allocated_to; next_role = rsc->next_role; if (next_role == RSC_ROLE_UNKNOWN) { rsc->next_role = (chosen == NULL)? RSC_ROLE_STOPPED : RSC_ROLE_STARTED; } pe_rsc_trace(rsc, "Creating all actions for %s transition from %s to %s (%s) on %s", rsc->id, role2text(rsc->role), role2text(rsc->next_role), ((next_role == RSC_ROLE_UNKNOWN)? "implicit" : "explicit"), ((chosen == NULL)? "no node" : chosen->details->uname)); current = pe__find_active_on(rsc, &num_all_active, &num_clean_active); for (gIter = rsc->dangling_migrations; gIter != NULL; gIter = gIter->next) { pe_node_t *dangling_source = (pe_node_t *) gIter->data; pe_action_t *stop = NULL; pe_rsc_trace(rsc, "Creating stop action %s cleanup for %s on %s due to dangling migration", (pcmk_is_set(data_set->flags, pe_flag_remove_after_stop)? "and" : "without"), rsc->id, dangling_source->details->uname); stop = stop_action(rsc, dangling_source, FALSE); pe__set_action_flags(stop, pe_action_dangle); if (pcmk_is_set(data_set->flags, pe_flag_remove_after_stop)) { DeleteRsc(rsc, dangling_source, FALSE, data_set); } } if ((num_all_active == 2) && (num_clean_active == 2) && chosen && rsc->partial_migration_source && rsc->partial_migration_target && (current->details == rsc->partial_migration_source->details) && (chosen->details == rsc->partial_migration_target->details)) { /* The chosen node is still the migration target from a partial * migration. Attempt to continue the migration instead of recovering * by stopping the resource everywhere and starting it on a single node. */ pe_rsc_trace(rsc, "Will attempt to continue with partial migration " "to target %s from %s", rsc->partial_migration_target->details->id, rsc->partial_migration_source->details->id); } else if (!pcmk_is_set(rsc->flags, pe_rsc_needs_fencing)) { /* If a resource has "requires" set to nothing or quorum, don't consider * it active on unclean nodes (similar to how all resources behave when * stonith-enabled is false). We can start such resources elsewhere * before fencing completes, and if we considered the resource active on * the failed node, we would attempt recovery for being active on * multiple nodes. */ multiply_active = (num_clean_active > 1); } else { multiply_active = (num_all_active > 1); } if (multiply_active) { if (rsc->partial_migration_target && rsc->partial_migration_source) { // Migration was in progress, but we've chosen a different target crm_notice("Resource %s can no longer migrate from %s to %s " "(will stop on both nodes)", rsc->id, rsc->partial_migration_source->details->uname, rsc->partial_migration_target->details->uname); } else { const char *class = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS); // Resource was incorrectly multiply active pe_proc_err("%s resource %s is active on %u nodes (%s)", crm_str(class), rsc->id, num_all_active, recovery2text(rsc->recovery_type)); crm_notice("See https://wiki.clusterlabs.org/wiki/FAQ#Resource_is_Too_Active for more information"); } if (rsc->recovery_type == recovery_stop_start) { need_stop = TRUE; } /* If by chance a partial migration is in process, but the migration * target is not chosen still, clear all partial migration data. */ rsc->partial_migration_source = rsc->partial_migration_target = NULL; allow_migrate = FALSE; } if (pcmk_is_set(rsc->flags, pe_rsc_start_pending)) { pe_rsc_trace(rsc, "Creating start action for %s to represent already pending start", rsc->id); start = start_action(rsc, chosen, TRUE); pe__set_action_flags(start, pe_action_print_always); } if (current && chosen && current->details != chosen->details) { pe_rsc_trace(rsc, "Moving %s from %s to %s", rsc->id, crm_str(current->details->uname), crm_str(chosen->details->uname)); is_moving = TRUE; need_stop = TRUE; } else if (pcmk_is_set(rsc->flags, pe_rsc_failed)) { if (pcmk_is_set(rsc->flags, pe_rsc_stop)) { need_stop = TRUE; pe_rsc_trace(rsc, "Recovering %s", rsc->id); } else { pe_rsc_trace(rsc, "Recovering %s by demotion", rsc->id); if (rsc->next_role == RSC_ROLE_MASTER) { need_promote = TRUE; } } } else if (pcmk_is_set(rsc->flags, pe_rsc_block)) { pe_rsc_trace(rsc, "Blocking further actions on %s", rsc->id); need_stop = TRUE; } else if (rsc->role > RSC_ROLE_STARTED && current != NULL && chosen != NULL) { pe_rsc_trace(rsc, "Creating start action for promoted resource %s", rsc->id); start = start_action(rsc, chosen, TRUE); if (!pcmk_is_set(start->flags, pe_action_optional)) { // Recovery of a promoted resource pe_rsc_trace(rsc, "%s restart is required for recovery", rsc->id); need_stop = TRUE; } } /* Create any additional actions required when bringing resource down and * back up to same level. */ role = rsc->role; while (role != RSC_ROLE_STOPPED) { next_role = rsc_state_matrix[role][RSC_ROLE_STOPPED]; pe_rsc_trace(rsc, "Creating %s action to take %s down from %s to %s", (need_stop? "required" : "optional"), rsc->id, role2text(role), role2text(next_role)); if (rsc_action_matrix[role][next_role] (rsc, current, !need_stop, data_set) == FALSE) { break; } role = next_role; } while ((rsc->role <= rsc->next_role) && (role != rsc->role) && !pcmk_is_set(rsc->flags, pe_rsc_block)) { bool required = need_stop; next_role = rsc_state_matrix[role][rsc->role]; if ((next_role == RSC_ROLE_MASTER) && need_promote) { required = true; } pe_rsc_trace(rsc, "Creating %s action to take %s up from %s to %s", (required? "required" : "optional"), rsc->id, role2text(role), role2text(next_role)); if (rsc_action_matrix[role][next_role](rsc, chosen, !required, data_set) == FALSE) { break; } role = next_role; } role = rsc->role; /* Required steps from this role to the next */ while (role != rsc->next_role) { next_role = rsc_state_matrix[role][rsc->next_role]; pe_rsc_trace(rsc, "Creating action to take %s from %s to %s (ending at %s)", rsc->id, role2text(role), role2text(next_role), role2text(rsc->next_role)); if (rsc_action_matrix[role][next_role] (rsc, chosen, FALSE, data_set) == FALSE) { break; } role = next_role; } if (pcmk_is_set(rsc->flags, pe_rsc_block)) { pe_rsc_trace(rsc, "Not creating recurring monitors for blocked resource %s", rsc->id); } else if ((rsc->next_role != RSC_ROLE_STOPPED) || !pcmk_is_set(rsc->flags, pe_rsc_managed)) { pe_rsc_trace(rsc, "Creating recurring monitors for %s resource %s", ((rsc->next_role == RSC_ROLE_STOPPED)? "unmanaged" : "active"), rsc->id); start = start_action(rsc, chosen, TRUE); Recurring(rsc, start, chosen, data_set); Recurring_Stopped(rsc, start, chosen, data_set); } else { pe_rsc_trace(rsc, "Creating recurring monitors for inactive resource %s", rsc->id); Recurring_Stopped(rsc, NULL, NULL, data_set); } /* if we are stuck in a partial migration, where the target * of the partial migration no longer matches the chosen target. * A full stop/start is required */ if (rsc->partial_migration_target && (chosen == NULL || rsc->partial_migration_target->details != chosen->details)) { pe_rsc_trace(rsc, "Not allowing partial migration of %s to continue", rsc->id); allow_migrate = FALSE; } else if (!is_moving || !pcmk_is_set(rsc->flags, pe_rsc_managed) || pcmk_any_flags_set(rsc->flags, pe_rsc_failed|pe_rsc_start_pending) || (current && current->details->unclean) || rsc->next_role < RSC_ROLE_STARTED) { allow_migrate = FALSE; } if (allow_migrate) { handle_migration_actions(rsc, current, chosen, data_set); } } static void rsc_avoids_remote_nodes(pe_resource_t *rsc) { GHashTableIter iter; pe_node_t *node = NULL; g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { if (node->details->remote_rsc) { node->weight = -INFINITY; } } } /*! * \internal * \brief Return allowed nodes as (possibly sorted) list * * Convert a resource's hash table of allowed nodes to a list. If printing to * stdout, sort the list, to keep action ID numbers consistent for regression * test output (while avoiding the performance hit on a live cluster). * * \param[in] rsc Resource to check for allowed nodes * \param[in] data_set Cluster working set * * \return List of resource's allowed nodes * \note Callers should take care not to rely on the list being sorted. */ static GList * allowed_nodes_as_list(pe_resource_t *rsc, pe_working_set_t *data_set) { GList *allowed_nodes = NULL; if (rsc->allowed_nodes) { allowed_nodes = g_hash_table_get_values(rsc->allowed_nodes); } if (pcmk_is_set(data_set->flags, pe_flag_stdout)) { allowed_nodes = g_list_sort(allowed_nodes, sort_node_uname); } return allowed_nodes; } void native_internal_constraints(pe_resource_t * rsc, pe_working_set_t * data_set) { /* This function is on the critical path and worth optimizing as much as possible */ pe_resource_t *top = NULL; GList *allowed_nodes = NULL; bool check_unfencing = FALSE; bool check_utilization = FALSE; if (!pcmk_is_set(rsc->flags, pe_rsc_managed)) { pe_rsc_trace(rsc, "Skipping native constraints for unmanaged resource: %s", rsc->id); return; } top = uber_parent(rsc); // Whether resource requires unfencing check_unfencing = !pcmk_is_set(rsc->flags, pe_rsc_fence_device) && pcmk_is_set(data_set->flags, pe_flag_enable_unfencing) && pcmk_is_set(rsc->flags, pe_rsc_needs_unfencing); // Whether a non-default placement strategy is used check_utilization = (g_hash_table_size(rsc->utilization) > 0) && !pcmk__str_eq(data_set->placement_strategy, "default", pcmk__str_casei); // Order stops before starts (i.e. restart) custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_STOP, 0), NULL, rsc, pcmk__op_key(rsc->id, RSC_START, 0), NULL, pe_order_optional|pe_order_implies_then|pe_order_restart, data_set); // Promotable ordering: demote before stop, start before promote if (pcmk_is_set(top->flags, pe_rsc_promotable) || (rsc->role > RSC_ROLE_SLAVE)) { custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_DEMOTE, 0), NULL, rsc, pcmk__op_key(rsc->id, RSC_STOP, 0), NULL, pe_order_implies_first_master, data_set); custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_START, 0), NULL, rsc, pcmk__op_key(rsc->id, RSC_PROMOTE, 0), NULL, pe_order_runnable_left, data_set); } // Don't clear resource history if probing on same node custom_action_order(rsc, pcmk__op_key(rsc->id, CRM_OP_LRM_DELETE, 0), NULL, rsc, pcmk__op_key(rsc->id, RSC_STATUS, 0), NULL, pe_order_same_node|pe_order_then_cancels_first, data_set); // Certain checks need allowed nodes if (check_unfencing || check_utilization || rsc->container) { allowed_nodes = allowed_nodes_as_list(rsc, data_set); } if (check_unfencing) { /* Check if the node needs to be unfenced first */ for (GList *item = allowed_nodes; item; item = item->next) { pe_node_t *node = item->data; pe_action_t *unfence = pe_fence_op(node, "on", TRUE, NULL, FALSE, data_set); crm_debug("Ordering any stops of %s before %s, and any starts after", rsc->id, unfence->uuid); /* * It would be more efficient to order clone resources once, * rather than order each instance, but ordering the instance * allows us to avoid unnecessary dependencies that might conflict * with user constraints. * * @TODO: This constraint can still produce a transition loop if the * resource has a stop scheduled on the node being unfenced, and * there is a user ordering constraint to start some other resource * (which will be ordered after the unfence) before stopping this * resource. An example is "start some slow-starting cloned service * before stopping an associated virtual IP that may be moving to * it": * stop this -> unfencing -> start that -> stop this */ custom_action_order(rsc, stop_key(rsc), NULL, NULL, strdup(unfence->uuid), unfence, pe_order_optional|pe_order_same_node, data_set); custom_action_order(NULL, strdup(unfence->uuid), unfence, rsc, start_key(rsc), NULL, pe_order_implies_then_on_node|pe_order_same_node, data_set); } } if (check_utilization) { GListPtr gIter = NULL; pe_rsc_trace(rsc, "Creating utilization constraints for %s - strategy: %s", rsc->id, data_set->placement_strategy); for (gIter = rsc->running_on; gIter != NULL; gIter = gIter->next) { pe_node_t *current = (pe_node_t *) gIter->data; char *load_stopped_task = crm_strdup_printf(LOAD_STOPPED "_%s", current->details->uname); pe_action_t *load_stopped = get_pseudo_op(load_stopped_task, data_set); if (load_stopped->node == NULL) { load_stopped->node = pe__copy_node(current); update_action_flags(load_stopped, pe_action_optional | pe_action_clear, __func__, __LINE__); } custom_action_order(rsc, stop_key(rsc), NULL, NULL, load_stopped_task, load_stopped, pe_order_load, data_set); } for (GList *item = allowed_nodes; item; item = item->next) { pe_node_t *next = item->data; char *load_stopped_task = crm_strdup_printf(LOAD_STOPPED "_%s", next->details->uname); pe_action_t *load_stopped = get_pseudo_op(load_stopped_task, data_set); if (load_stopped->node == NULL) { load_stopped->node = pe__copy_node(next); update_action_flags(load_stopped, pe_action_optional | pe_action_clear, __func__, __LINE__); } custom_action_order(NULL, strdup(load_stopped_task), load_stopped, rsc, start_key(rsc), NULL, pe_order_load, data_set); custom_action_order(NULL, strdup(load_stopped_task), load_stopped, rsc, pcmk__op_key(rsc->id, RSC_MIGRATE, 0), NULL, pe_order_load, data_set); free(load_stopped_task); } } if (rsc->container) { pe_resource_t *remote_rsc = NULL; if (rsc->is_remote_node) { // rsc is the implicit remote connection for a guest or bundle node /* Do not allow a guest resource to live on a Pacemaker Remote node, * to avoid nesting remotes. However, allow bundles to run on remote * nodes. */ if (!pcmk_is_set(rsc->flags, pe_rsc_allow_remote_remotes)) { rsc_avoids_remote_nodes(rsc->container); } /* If someone cleans up a guest or bundle node's container, we will * likely schedule a (re-)probe of the container and recovery of the * connection. Order the connection stop after the container probe, * so that if we detect the container running, we will trigger a new * transition and avoid the unnecessary recovery. */ new_rsc_order(rsc->container, RSC_STATUS, rsc, RSC_STOP, pe_order_optional, data_set); /* A user can specify that a resource must start on a Pacemaker Remote * node by explicitly configuring it with the container=NODENAME * meta-attribute. This is of questionable merit, since location * constraints can accomplish the same thing. But we support it, so here * we check whether a resource (that is not itself a remote connection) * has container set to a remote node or guest node resource. */ } else if (rsc->container->is_remote_node) { remote_rsc = rsc->container; } else { remote_rsc = pe__resource_contains_guest_node(data_set, rsc->container); } if (remote_rsc) { /* Force the resource on the Pacemaker Remote node instead of * colocating the resource with the container resource. */ for (GList *item = allowed_nodes; item; item = item->next) { pe_node_t *node = item->data; if (node->details->remote_rsc != remote_rsc) { node->weight = -INFINITY; } } } else { /* This resource is either a filler for a container that does NOT * represent a Pacemaker Remote node, or a Pacemaker Remote * connection resource for a guest node or bundle. */ int score; crm_trace("Order and colocate %s relative to its container %s", rsc->id, rsc->container->id); custom_action_order(rsc->container, pcmk__op_key(rsc->container->id, RSC_START, 0), NULL, rsc, pcmk__op_key(rsc->id, RSC_START, 0), NULL, pe_order_implies_then|pe_order_runnable_left, data_set); custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_STOP, 0), NULL, rsc->container, pcmk__op_key(rsc->container->id, RSC_STOP, 0), NULL, pe_order_implies_first, data_set); if (pcmk_is_set(rsc->flags, pe_rsc_allow_remote_remotes)) { score = 10000; /* Highly preferred but not essential */ } else { score = INFINITY; /* Force them to run on the same host */ } - rsc_colocation_new("resource-with-container", NULL, score, rsc, - rsc->container, NULL, NULL, data_set); + pcmk__new_colocation("resource-with-container", NULL, score, rsc, + rsc->container, NULL, NULL, data_set); } } if (rsc->is_remote_node || pcmk_is_set(rsc->flags, pe_rsc_fence_device)) { /* don't allow remote nodes to run stonith devices * or remote connection resources.*/ rsc_avoids_remote_nodes(rsc); } g_list_free(allowed_nodes); } void native_rsc_colocation_lh(pe_resource_t *rsc_lh, pe_resource_t *rsc_rh, - rsc_colocation_t *constraint, + pcmk__colocation_t *constraint, pe_working_set_t *data_set) { if (rsc_lh == NULL) { pe_err("rsc_lh was NULL for %s", constraint->id); return; } else if (constraint->rsc_rh == NULL) { pe_err("rsc_rh was NULL for %s", constraint->id); return; } if (constraint->score == 0) { return; } pe_rsc_trace(rsc_lh, "Processing colocation constraint between %s and %s", rsc_lh->id, rsc_rh->id); rsc_rh->cmds->rsc_colocation_rh(rsc_lh, rsc_rh, constraint, data_set); } enum filter_colocation_res filter_colocation_constraint(pe_resource_t * rsc_lh, pe_resource_t * rsc_rh, - rsc_colocation_t * constraint, gboolean preview) + pcmk__colocation_t *constraint, gboolean preview) { if (constraint->score == 0) { return influence_nothing; } /* rh side must be allocated before we can process constraint */ if (!preview && pcmk_is_set(rsc_rh->flags, pe_rsc_provisional)) { return influence_nothing; } if ((constraint->role_lh >= RSC_ROLE_SLAVE) && rsc_lh->parent && pcmk_is_set(rsc_lh->parent->flags, pe_rsc_promotable) && !pcmk_is_set(rsc_lh->flags, pe_rsc_provisional)) { /* LH and RH resources have already been allocated, place the correct * priority on LH rsc for the given promotable clone resource role */ return influence_rsc_priority; } if (!preview && !pcmk_is_set(rsc_lh->flags, pe_rsc_provisional)) { // Log an error if we violated a mandatory colocation constraint const pe_node_t *rh_node = rsc_rh->allocated_to; if (rsc_lh->allocated_to == NULL) { // Dependent resource isn't allocated, so constraint doesn't matter return influence_nothing; } if (constraint->score >= INFINITY) { // Dependent resource must colocate with rh_node if ((rh_node == NULL) || (rh_node->details != rsc_lh->allocated_to->details)) { crm_err("%s must be colocated with %s but is not (%s vs. %s)", rsc_lh->id, rsc_rh->id, rsc_lh->allocated_to->details->uname, (rh_node? rh_node->details->uname : "unallocated")); } } else if (constraint->score <= -INFINITY) { // Dependent resource must anti-colocate with rh_node if ((rh_node != NULL) && (rsc_lh->allocated_to->details == rh_node->details)) { crm_err("%s and %s must be anti-colocated but are allocated " "to the same node (%s)", rsc_lh->id, rsc_rh->id, rh_node->details->uname); } } return influence_nothing; } if (constraint->score > 0 && constraint->role_lh != RSC_ROLE_UNKNOWN && constraint->role_lh != rsc_lh->next_role) { crm_trace("LH: Skipping constraint: \"%s\" state filter nextrole is %s", role2text(constraint->role_lh), role2text(rsc_lh->next_role)); return influence_nothing; } if (constraint->score > 0 && constraint->role_rh != RSC_ROLE_UNKNOWN && constraint->role_rh != rsc_rh->next_role) { crm_trace("RH: Skipping constraint: \"%s\" state filter", role2text(constraint->role_rh)); return influence_nothing; } if (constraint->score < 0 && constraint->role_lh != RSC_ROLE_UNKNOWN && constraint->role_lh == rsc_lh->next_role) { crm_trace("LH: Skipping negative constraint: \"%s\" state filter", role2text(constraint->role_lh)); return influence_nothing; } if (constraint->score < 0 && constraint->role_rh != RSC_ROLE_UNKNOWN && constraint->role_rh == rsc_rh->next_role) { crm_trace("RH: Skipping negative constraint: \"%s\" state filter", role2text(constraint->role_rh)); return influence_nothing; } return influence_rsc_location; } static void -influence_priority(pe_resource_t * rsc_lh, pe_resource_t * rsc_rh, rsc_colocation_t * constraint) +influence_priority(pe_resource_t *rsc_lh, pe_resource_t *rsc_rh, + pcmk__colocation_t *constraint) { const char *rh_value = NULL; const char *lh_value = NULL; const char *attribute = CRM_ATTR_ID; int score_multiplier = 1; if (constraint->score == 0) { return; } if (!rsc_rh->allocated_to || !rsc_lh->allocated_to) { return; } if (constraint->node_attribute != NULL) { attribute = constraint->node_attribute; } lh_value = pe_node_attribute_raw(rsc_lh->allocated_to, attribute); rh_value = pe_node_attribute_raw(rsc_rh->allocated_to, attribute); if (!pcmk__str_eq(lh_value, rh_value, pcmk__str_casei)) { if(constraint->score == INFINITY && constraint->role_lh == RSC_ROLE_MASTER) { rsc_lh->priority = -INFINITY; } return; } if (constraint->role_rh && (constraint->role_rh != rsc_rh->next_role)) { return; } if (constraint->role_lh == RSC_ROLE_SLAVE) { score_multiplier = -1; } rsc_lh->priority = pe__add_scores(score_multiplier * constraint->score, rsc_lh->priority); } static void -colocation_match(pe_resource_t * rsc_lh, pe_resource_t * rsc_rh, rsc_colocation_t * constraint) +colocation_match(pe_resource_t *rsc_lh, pe_resource_t *rsc_rh, + pcmk__colocation_t *constraint) { const char *attribute = CRM_ATTR_ID; const char *value = NULL; GHashTable *work = NULL; GHashTableIter iter; pe_node_t *node = NULL; if (constraint->score == 0) { return; } if (constraint->node_attribute != NULL) { attribute = constraint->node_attribute; } if (rsc_rh->allocated_to) { value = pe_node_attribute_raw(rsc_rh->allocated_to, attribute); } else if (constraint->score < 0) { // Nothing to do (anti-colocation with something that is not running) return; } work = pcmk__copy_node_table(rsc_lh->allowed_nodes); g_hash_table_iter_init(&iter, work); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { if (rsc_rh->allocated_to == NULL) { pe_rsc_trace(rsc_lh, "%s: %s@%s -= %d (%s inactive)", constraint->id, rsc_lh->id, node->details->uname, constraint->score, rsc_rh->id); node->weight = pe__add_scores(-constraint->score, node->weight); } else if (pcmk__str_eq(pe_node_attribute_raw(node, attribute), value, pcmk__str_casei)) { if (constraint->score < CRM_SCORE_INFINITY) { pe_rsc_trace(rsc_lh, "%s: %s@%s += %d", constraint->id, rsc_lh->id, node->details->uname, constraint->score); node->weight = pe__add_scores(constraint->score, node->weight); } } else if (constraint->score >= CRM_SCORE_INFINITY) { pe_rsc_trace(rsc_lh, "%s: %s@%s -= %d (%s mismatch)", constraint->id, rsc_lh->id, node->details->uname, constraint->score, attribute); node->weight = pe__add_scores(-constraint->score, node->weight); } } if (can_run_any(work) || constraint->score <= -INFINITY || constraint->score >= INFINITY) { g_hash_table_destroy(rsc_lh->allowed_nodes); rsc_lh->allowed_nodes = work; work = NULL; } else { pe_rsc_info(rsc_lh, "%s: Rolling back scores from %s (no available nodes)", rsc_lh->id, rsc_rh->id); } if (work) { g_hash_table_destroy(work); } } void native_rsc_colocation_rh(pe_resource_t *rsc_lh, pe_resource_t *rsc_rh, - rsc_colocation_t *constraint, + pcmk__colocation_t *constraint, pe_working_set_t *data_set) { enum filter_colocation_res filter_results; CRM_ASSERT(rsc_lh); CRM_ASSERT(rsc_rh); filter_results = filter_colocation_constraint(rsc_lh, rsc_rh, constraint, FALSE); pe_rsc_trace(rsc_lh, "%s %s with %s (%s, score=%d, filter=%d)", ((constraint->score >= 0)? "Colocating" : "Anti-colocating"), rsc_lh->id, rsc_rh->id, constraint->id, constraint->score, filter_results); switch (filter_results) { case influence_rsc_priority: influence_priority(rsc_lh, rsc_rh, constraint); break; case influence_rsc_location: colocation_match(rsc_lh, rsc_rh, constraint); break; case influence_nothing: default: return; } } static gboolean filter_rsc_ticket(pe_resource_t * rsc_lh, rsc_ticket_t * rsc_ticket) { if (rsc_ticket->role_lh != RSC_ROLE_UNKNOWN && rsc_ticket->role_lh != rsc_lh->role) { pe_rsc_trace(rsc_lh, "LH: Skipping constraint: \"%s\" state filter", role2text(rsc_ticket->role_lh)); return FALSE; } return TRUE; } void rsc_ticket_constraint(pe_resource_t * rsc_lh, rsc_ticket_t * rsc_ticket, pe_working_set_t * data_set) { if (rsc_ticket == NULL) { pe_err("rsc_ticket was NULL"); return; } if (rsc_lh == NULL) { pe_err("rsc_lh was NULL for %s", rsc_ticket->id); return; } if (rsc_ticket->ticket->granted && rsc_ticket->ticket->standby == FALSE) { return; } if (rsc_lh->children) { GListPtr gIter = rsc_lh->children; pe_rsc_trace(rsc_lh, "Processing ticket dependencies from %s", rsc_lh->id); for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; rsc_ticket_constraint(child_rsc, rsc_ticket, data_set); } return; } pe_rsc_trace(rsc_lh, "%s: Processing ticket dependency on %s (%s, %s)", rsc_lh->id, rsc_ticket->ticket->id, rsc_ticket->id, role2text(rsc_ticket->role_lh)); if ((rsc_ticket->ticket->granted == FALSE) && (rsc_lh->running_on != NULL)) { GListPtr gIter = NULL; switch (rsc_ticket->loss_policy) { case loss_ticket_stop: resource_location(rsc_lh, NULL, -INFINITY, "__loss_of_ticket__", data_set); break; case loss_ticket_demote: // Promotion score will be set to -INFINITY in promotion_order() if (rsc_ticket->role_lh != RSC_ROLE_MASTER) { resource_location(rsc_lh, NULL, -INFINITY, "__loss_of_ticket__", data_set); } break; case loss_ticket_fence: if (filter_rsc_ticket(rsc_lh, rsc_ticket) == FALSE) { return; } resource_location(rsc_lh, NULL, -INFINITY, "__loss_of_ticket__", data_set); for (gIter = rsc_lh->running_on; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; pe_fence_node(data_set, node, "deadman ticket was lost", FALSE); } break; case loss_ticket_freeze: if (filter_rsc_ticket(rsc_lh, rsc_ticket) == FALSE) { return; } if (rsc_lh->running_on != NULL) { pe__clear_resource_flags(rsc_lh, pe_rsc_managed); pe__set_resource_flags(rsc_lh, pe_rsc_block); } break; } } else if (rsc_ticket->ticket->granted == FALSE) { if (rsc_ticket->role_lh != RSC_ROLE_MASTER || rsc_ticket->loss_policy == loss_ticket_stop) { resource_location(rsc_lh, NULL, -INFINITY, "__no_ticket__", data_set); } } else if (rsc_ticket->ticket->standby) { if (rsc_ticket->role_lh != RSC_ROLE_MASTER || rsc_ticket->loss_policy == loss_ticket_stop) { resource_location(rsc_lh, NULL, -INFINITY, "__ticket_standby__", data_set); } } } enum pe_action_flags native_action_flags(pe_action_t * action, pe_node_t * node) { return action->flags; } static inline bool is_primitive_action(pe_action_t *action) { return action && action->rsc && (action->rsc->variant == pe_native); } /*! * \internal * \brief Set action bits appropriately when pe_restart_order is used * * \param[in] first 'First' action in an ordering with pe_restart_order * \param[in] then 'Then' action in an ordering with pe_restart_order * \param[in] filter What ordering flags to care about * * \note pe_restart_order is set for "stop resource before starting it" and * "stop later group member before stopping earlier group member" */ static void handle_restart_ordering(pe_action_t *first, pe_action_t *then, enum pe_action_flags filter) { const char *reason = NULL; CRM_ASSERT(is_primitive_action(first)); CRM_ASSERT(is_primitive_action(then)); // We need to update the action in two cases: // ... if 'then' is required if (pcmk_is_set(filter, pe_action_optional) && !pcmk_is_set(then->flags, pe_action_optional)) { reason = "restart"; } /* ... if 'then' is unrunnable start of managed resource (if a resource * should restart but can't start, we still want to stop) */ if (pcmk_is_set(filter, pe_action_runnable) && !pcmk_is_set(then->flags, pe_action_runnable) && pcmk_is_set(then->rsc->flags, pe_rsc_managed) && pcmk__str_eq(then->task, RSC_START, pcmk__str_casei)) { reason = "stop"; } if (reason == NULL) { return; } pe_rsc_trace(first->rsc, "Handling %s -> %s for %s", first->uuid, then->uuid, reason); // Make 'first' required if it is runnable if (pcmk_is_set(first->flags, pe_action_runnable)) { pe_action_implies(first, then, pe_action_optional); } // Make 'first' required if 'then' is required if (!pcmk_is_set(then->flags, pe_action_optional)) { pe_action_implies(first, then, pe_action_optional); } // Make 'first' unmigratable if 'then' is unmigratable if (!pcmk_is_set(then->flags, pe_action_migrate_runnable)) { pe_action_implies(first, then, pe_action_migrate_runnable); } // Make 'then' unrunnable if 'first' is required but unrunnable if (!pcmk_is_set(first->flags, pe_action_optional) && !pcmk_is_set(first->flags, pe_action_runnable)) { pe_action_implies(then, first, pe_action_runnable); } } enum pe_graph_flags native_update_actions(pe_action_t *first, pe_action_t *then, pe_node_t *node, enum pe_action_flags flags, enum pe_action_flags filter, enum pe_ordering type, pe_working_set_t *data_set) { /* flags == get_action_flags(first, then_node) called from update_action() */ enum pe_graph_flags changed = pe_graph_none; enum pe_action_flags then_flags = then->flags; enum pe_action_flags first_flags = first->flags; crm_trace( "Testing %s on %s (0x%.6x) with %s 0x%.6x", first->uuid, first->node ? first->node->details->uname : "[none]", first->flags, then->uuid, then->flags); if (type & pe_order_asymmetrical) { pe_resource_t *then_rsc = then->rsc; enum rsc_role_e then_rsc_role = then_rsc ? then_rsc->fns->state(then_rsc, TRUE) : 0; if (!then_rsc) { /* ignore */ } else if ((then_rsc_role == RSC_ROLE_STOPPED) && pcmk__str_eq(then->task, RSC_STOP, pcmk__str_casei)) { /* ignore... if 'then' is supposed to be stopped after 'first', but * then is already stopped, there is nothing to be done when non-symmetrical. */ } else if ((then_rsc_role >= RSC_ROLE_STARTED) && pcmk__str_eq(then->task, RSC_START, pcmk__str_casei) && pcmk_is_set(then->flags, pe_action_optional) && then->node && pcmk__list_of_1(then_rsc->running_on) && then->node->details == ((pe_node_t *) then_rsc->running_on->data)->details) { /* Ignore. If 'then' is supposed to be started after 'first', but * 'then' is already started, there is nothing to be done when * asymmetrical -- unless the start is mandatory, which indicates * the resource is restarting, and the ordering is still needed. */ } else if (!(first->flags & pe_action_runnable)) { /* prevent 'then' action from happening if 'first' is not runnable and * 'then' has not yet occurred. */ pe_action_implies(then, first, pe_action_optional); pe_action_implies(then, first, pe_action_runnable); pe_rsc_trace(then->rsc, "Unset optional and runnable on %s", then->uuid); } else { /* ignore... then is allowed to start/stop if it wants to. */ } } if (type & pe_order_implies_first) { if (pcmk_is_set(filter, pe_action_optional) && !pcmk_is_set(flags /* Should be then_flags? */, pe_action_optional)) { // Needs pcmk_is_set(first_flags, pe_action_optional) too? pe_rsc_trace(first->rsc, "Unset optional on %s because of %s", first->uuid, then->uuid); pe_action_implies(first, then, pe_action_optional); } if (pcmk_is_set(flags, pe_action_migrate_runnable) && !pcmk_is_set(then->flags, pe_action_migrate_runnable) && !pcmk_is_set(then->flags, pe_action_optional)) { pe_rsc_trace(first->rsc, "Unset migrate runnable on %s because of %s", first->uuid, then->uuid); pe_action_implies(first, then, pe_action_migrate_runnable); } } if (type & pe_order_implies_first_master) { if ((filter & pe_action_optional) && ((then->flags & pe_action_optional) == FALSE) && then->rsc && (then->rsc->role == RSC_ROLE_MASTER)) { pe_action_implies(first, then, pe_action_optional); if (pcmk_is_set(first->flags, pe_action_migrate_runnable) && !pcmk_is_set(then->flags, pe_action_migrate_runnable)) { pe_rsc_trace(first->rsc, "Unset migrate runnable on %s because of %s", first->uuid, then->uuid); pe_action_implies(first, then, pe_action_migrate_runnable); } pe_rsc_trace(then->rsc, "Unset optional on %s because of %s", first->uuid, then->uuid); } } if ((type & pe_order_implies_first_migratable) && pcmk_is_set(filter, pe_action_optional)) { if (((then->flags & pe_action_migrate_runnable) == FALSE) || ((then->flags & pe_action_runnable) == FALSE)) { pe_rsc_trace(then->rsc, "Unset runnable on %s because %s is neither runnable or migratable", first->uuid, then->uuid); pe_action_implies(first, then, pe_action_runnable); } if ((then->flags & pe_action_optional) == 0) { pe_rsc_trace(then->rsc, "Unset optional on %s because %s is not optional", first->uuid, then->uuid); pe_action_implies(first, then, pe_action_optional); } } if ((type & pe_order_pseudo_left) && pcmk_is_set(filter, pe_action_optional)) { if ((first->flags & pe_action_runnable) == FALSE) { pe_action_implies(then, first, pe_action_migrate_runnable); pe__clear_action_flags(then, pe_action_pseudo); pe_rsc_trace(then->rsc, "Unset pseudo on %s because %s is not runnable", then->uuid, first->uuid); } } if (pcmk_is_set(type, pe_order_runnable_left) && pcmk_is_set(filter, pe_action_runnable) && pcmk_is_set(then->flags, pe_action_runnable) && !pcmk_is_set(flags, pe_action_runnable)) { pe_rsc_trace(then->rsc, "Unset runnable on %s because of %s", then->uuid, first->uuid); pe_action_implies(then, first, pe_action_runnable); pe_action_implies(then, first, pe_action_migrate_runnable); } if (pcmk_is_set(type, pe_order_implies_then) && pcmk_is_set(filter, pe_action_optional) && pcmk_is_set(then->flags, pe_action_optional) && !pcmk_is_set(flags, pe_action_optional)) { /* in this case, treat migrate_runnable as if first is optional */ if (!pcmk_is_set(first->flags, pe_action_migrate_runnable)) { pe_rsc_trace(then->rsc, "Unset optional on %s because of %s", then->uuid, first->uuid); pe_action_implies(then, first, pe_action_optional); } } if (pcmk_is_set(type, pe_order_restart)) { handle_restart_ordering(first, then, filter); } if (then_flags != then->flags) { pe__set_graph_flags(changed, first, pe_graph_updated_then); pe_rsc_trace(then->rsc, "Then: Flags for %s on %s are now 0x%.6x (was 0x%.6x) because of %s 0x%.6x", then->uuid, then->node ? then->node->details->uname : "[none]", then->flags, then_flags, first->uuid, first->flags); if(then->rsc && then->rsc->parent) { /* "X_stop then X_start" doesn't get handled for cloned groups unless we do this */ update_action(then, data_set); } } if (first_flags != first->flags) { pe__set_graph_flags(changed, first, pe_graph_updated_first); pe_rsc_trace(first->rsc, "First: Flags for %s on %s are now 0x%.6x (was 0x%.6x) because of %s 0x%.6x", first->uuid, first->node ? first->node->details->uname : "[none]", first->flags, first_flags, then->uuid, then->flags); } return changed; } void native_rsc_location(pe_resource_t *rsc, pe__location_t *constraint) { GListPtr gIter = NULL; bool need_role = false; CRM_CHECK((constraint != NULL) && (rsc != NULL), return); // If a role was specified, ensure constraint is applicable need_role = (constraint->role_filter > RSC_ROLE_UNKNOWN); if (need_role && (constraint->role_filter != rsc->next_role)) { pe_rsc_trace(rsc, "Not applying %s to %s because role will be %s not %s", constraint->id, rsc->id, role2text(rsc->next_role), role2text(constraint->role_filter)); return; } if (constraint->node_list_rh == NULL) { pe_rsc_trace(rsc, "Not applying %s to %s because no nodes match", constraint->id, rsc->id); return; } pe_rsc_trace(rsc, "Applying %s%s%s to %s", constraint->id, (need_role? " for role " : ""), (need_role? role2text(constraint->role_filter) : ""), rsc->id); for (gIter = constraint->node_list_rh; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; pe_node_t *other_node = NULL; other_node = (pe_node_t *) pe_hash_table_lookup(rsc->allowed_nodes, node->details->id); if (other_node != NULL) { pe_rsc_trace(rsc, "* + %d on %s", node->weight, node->details->uname); other_node->weight = pe__add_scores(other_node->weight, node->weight); } else { pe_rsc_trace(rsc, "* = %d on %s", node->weight, node->details->uname); other_node = pe__copy_node(node); g_hash_table_insert(rsc->allowed_nodes, (gpointer) other_node->details->id, other_node); } if (other_node->rsc_discover_mode < constraint->discover_mode) { if (constraint->discover_mode == pe_discover_exclusive) { rsc->exclusive_discover = TRUE; } /* exclusive > never > always... always is default */ other_node->rsc_discover_mode = constraint->discover_mode; } } } void native_expand(pe_resource_t * rsc, pe_working_set_t * data_set) { GListPtr gIter = NULL; CRM_ASSERT(rsc); pe_rsc_trace(rsc, "Processing actions from %s", rsc->id); for (gIter = rsc->actions; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; crm_trace("processing action %d for rsc=%s", action->id, rsc->id); graph_element_from_action(action, data_set); } for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; child_rsc->cmds->expand(child_rsc, data_set); } } #define log_change(a, fmt, args...) do { \ if(a && a->reason && terminal) { \ printf(" * "fmt" \tdue to %s\n", ##args, a->reason); \ } else if(a && a->reason) { \ crm_notice(fmt" \tdue to %s", ##args, a->reason); \ } else if(terminal) { \ printf(" * "fmt"\n", ##args); \ } else { \ crm_notice(fmt, ##args); \ } \ } while(0) #define STOP_SANITY_ASSERT(lineno) do { \ if(current && current->details->unclean) { \ /* It will be a pseudo op */ \ } else if(stop == NULL) { \ crm_err("%s:%d: No stop action exists for %s", \ __func__, lineno, rsc->id); \ CRM_ASSERT(stop != NULL); \ } else if (pcmk_is_set(stop->flags, pe_action_optional)) { \ crm_err("%s:%d: Action %s is still optional", \ __func__, lineno, stop->uuid); \ CRM_ASSERT(!pcmk_is_set(stop->flags, pe_action_optional)); \ } \ } while(0) static void LogAction(const char *change, pe_resource_t *rsc, pe_node_t *origin, pe_node_t *destination, pe_action_t *action, pe_action_t *source, gboolean terminal) { int len = 0; char *reason = NULL; char *details = NULL; bool same_host = FALSE; bool same_role = FALSE; bool need_role = FALSE; static int rsc_width = 5; static int detail_width = 5; CRM_ASSERT(action); CRM_ASSERT(destination != NULL || origin != NULL); if(source == NULL) { source = action; } len = strlen(rsc->id); if(len > rsc_width) { rsc_width = len + 2; } if(rsc->role > RSC_ROLE_STARTED || rsc->next_role > RSC_ROLE_SLAVE) { need_role = TRUE; } if(origin != NULL && destination != NULL && origin->details == destination->details) { same_host = TRUE; } if(rsc->role == rsc->next_role) { same_role = TRUE; } if (need_role && (origin == NULL)) { /* Starting and promoting a promotable clone instance */ details = crm_strdup_printf("%s -> %s %s", role2text(rsc->role), role2text(rsc->next_role), destination->details->uname); } else if (origin == NULL) { /* Starting a resource */ details = crm_strdup_printf("%s", destination->details->uname); } else if (need_role && (destination == NULL)) { /* Stopping a promotable clone instance */ details = crm_strdup_printf("%s %s", role2text(rsc->role), origin->details->uname); } else if (destination == NULL) { /* Stopping a resource */ details = crm_strdup_printf("%s", origin->details->uname); } else if (need_role && same_role && same_host) { /* Recovering, restarting or re-promoting a promotable clone instance */ details = crm_strdup_printf("%s %s", role2text(rsc->role), origin->details->uname); } else if (same_role && same_host) { /* Recovering or Restarting a normal resource */ details = crm_strdup_printf("%s", origin->details->uname); } else if (need_role && same_role) { /* Moving a promotable clone instance */ details = crm_strdup_printf("%s -> %s %s", origin->details->uname, destination->details->uname, role2text(rsc->role)); } else if (same_role) { /* Moving a normal resource */ details = crm_strdup_printf("%s -> %s", origin->details->uname, destination->details->uname); } else if (same_host) { /* Promoting or demoting a promotable clone instance */ details = crm_strdup_printf("%s -> %s %s", role2text(rsc->role), role2text(rsc->next_role), origin->details->uname); } else { /* Moving and promoting/demoting */ details = crm_strdup_printf("%s %s -> %s %s", role2text(rsc->role), origin->details->uname, role2text(rsc->next_role), destination->details->uname); } len = strlen(details); if(len > detail_width) { detail_width = len; } if(source->reason && !pcmk_is_set(action->flags, pe_action_runnable)) { reason = crm_strdup_printf(" due to %s (blocked)", source->reason); } else if(source->reason) { reason = crm_strdup_printf(" due to %s", source->reason); } else if (!pcmk_is_set(action->flags, pe_action_runnable)) { reason = strdup(" blocked"); } else { reason = strdup(""); } if(terminal) { printf(" * %-8s %-*s ( %*s ) %s\n", change, rsc_width, rsc->id, detail_width, details, reason); } else { crm_notice(" * %-8s %-*s ( %*s ) %s", change, rsc_width, rsc->id, detail_width, details, reason); } free(details); free(reason); } void LogActions(pe_resource_t * rsc, pe_working_set_t * data_set, gboolean terminal) { pe_node_t *next = NULL; pe_node_t *current = NULL; pe_node_t *start_node = NULL; pe_action_t *stop = NULL; pe_action_t *start = NULL; pe_action_t *demote = NULL; pe_action_t *promote = NULL; char *key = NULL; gboolean moving = FALSE; GListPtr possible_matches = NULL; if(rsc->variant == pe_container) { pcmk__bundle_log_actions(rsc, data_set, terminal); return; } if (rsc->children) { GListPtr gIter = NULL; for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; LogActions(child_rsc, data_set, terminal); } return; } next = rsc->allocated_to; if (rsc->running_on) { current = pe__current_node(rsc); if (rsc->role == RSC_ROLE_STOPPED) { /* * This can occur when resources are being recovered * We fiddle with the current role in native_create_actions() */ rsc->role = RSC_ROLE_STARTED; } } if ((current == NULL) && pcmk_is_set(rsc->flags, pe_rsc_orphan)) { /* Don't log stopped orphans */ return; } if (!pcmk_is_set(rsc->flags, pe_rsc_managed) || (current == NULL && next == NULL)) { pe_rsc_info(rsc, "Leave %s\t(%s%s)", rsc->id, role2text(rsc->role), !pcmk_is_set(rsc->flags, pe_rsc_managed)? " unmanaged" : ""); return; } if (current != NULL && next != NULL && !pcmk__str_eq(current->details->id, next->details->id, pcmk__str_casei)) { moving = TRUE; } possible_matches = pe__resource_actions(rsc, next, RSC_START, FALSE); if (possible_matches) { start = possible_matches->data; g_list_free(possible_matches); } if ((start == NULL) || !pcmk_is_set(start->flags, pe_action_runnable)) { start_node = NULL; } else { start_node = current; } possible_matches = pe__resource_actions(rsc, start_node, RSC_STOP, FALSE); if (possible_matches) { stop = possible_matches->data; g_list_free(possible_matches); } possible_matches = pe__resource_actions(rsc, next, RSC_PROMOTE, FALSE); if (possible_matches) { promote = possible_matches->data; g_list_free(possible_matches); } possible_matches = pe__resource_actions(rsc, next, RSC_DEMOTE, FALSE); if (possible_matches) { demote = possible_matches->data; g_list_free(possible_matches); } if (rsc->role == rsc->next_role) { pe_action_t *migrate_op = NULL; possible_matches = pe__resource_actions(rsc, next, RSC_MIGRATED, FALSE); if (possible_matches) { migrate_op = possible_matches->data; } CRM_CHECK(next != NULL,); if (next == NULL) { } else if ((migrate_op != NULL) && (current != NULL) && pcmk_is_set(migrate_op->flags, pe_action_runnable)) { LogAction("Migrate", rsc, current, next, start, NULL, terminal); } else if (pcmk_is_set(rsc->flags, pe_rsc_reload)) { LogAction("Reload", rsc, current, next, start, NULL, terminal); } else if (start == NULL || pcmk_is_set(start->flags, pe_action_optional)) { if ((demote != NULL) && (promote != NULL) && !pcmk_is_set(demote->flags, pe_action_optional) && !pcmk_is_set(promote->flags, pe_action_optional)) { LogAction("Re-promote", rsc, current, next, promote, demote, terminal); } else { pe_rsc_info(rsc, "Leave %s\t(%s %s)", rsc->id, role2text(rsc->role), next->details->uname); } } else if (!pcmk_is_set(start->flags, pe_action_runnable)) { LogAction("Stop", rsc, current, NULL, stop, (stop && stop->reason)? stop : start, terminal); STOP_SANITY_ASSERT(__LINE__); } else if (moving && current) { LogAction(pcmk_is_set(rsc->flags, pe_rsc_failed)? "Recover" : "Move", rsc, current, next, stop, NULL, terminal); } else if (pcmk_is_set(rsc->flags, pe_rsc_failed)) { LogAction("Recover", rsc, current, NULL, stop, NULL, terminal); STOP_SANITY_ASSERT(__LINE__); } else { LogAction("Restart", rsc, current, next, start, NULL, terminal); /* STOP_SANITY_ASSERT(__LINE__); False positive for migrate-fail-7 */ } g_list_free(possible_matches); return; } if(stop && (rsc->next_role == RSC_ROLE_STOPPED || (start && !pcmk_is_set(start->flags, pe_action_runnable)))) { GListPtr gIter = NULL; key = stop_key(rsc); for (gIter = rsc->running_on; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; pe_action_t *stop_op = NULL; possible_matches = find_actions(rsc->actions, key, node); if (possible_matches) { stop_op = possible_matches->data; g_list_free(possible_matches); } if (stop_op && (stop_op->flags & pe_action_runnable)) { STOP_SANITY_ASSERT(__LINE__); } LogAction("Stop", rsc, node, NULL, stop_op, (stop_op && stop_op->reason)? stop_op : start, terminal); } free(key); } else if ((stop != NULL) && pcmk_all_flags_set(rsc->flags, pe_rsc_failed|pe_rsc_stop)) { /* 'stop' may be NULL if the failure was ignored */ LogAction("Recover", rsc, current, next, stop, start, terminal); STOP_SANITY_ASSERT(__LINE__); } else if (moving) { LogAction("Move", rsc, current, next, stop, NULL, terminal); STOP_SANITY_ASSERT(__LINE__); } else if (pcmk_is_set(rsc->flags, pe_rsc_reload)) { LogAction("Reload", rsc, current, next, start, NULL, terminal); } else if (stop != NULL && !pcmk_is_set(stop->flags, pe_action_optional)) { LogAction("Restart", rsc, current, next, start, NULL, terminal); STOP_SANITY_ASSERT(__LINE__); } else if (rsc->role == RSC_ROLE_MASTER) { CRM_LOG_ASSERT(current != NULL); LogAction("Demote", rsc, current, next, demote, NULL, terminal); } else if(rsc->next_role == RSC_ROLE_MASTER) { CRM_LOG_ASSERT(next); LogAction("Promote", rsc, current, next, promote, NULL, terminal); } else if (rsc->role == RSC_ROLE_STOPPED && rsc->next_role > RSC_ROLE_STOPPED) { LogAction("Start", rsc, current, next, start, NULL, terminal); } } gboolean StopRsc(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set) { GListPtr gIter = NULL; CRM_ASSERT(rsc); pe_rsc_trace(rsc, "%s", rsc->id); for (gIter = rsc->running_on; gIter != NULL; gIter = gIter->next) { pe_node_t *current = (pe_node_t *) gIter->data; pe_action_t *stop; if (rsc->partial_migration_target) { if (rsc->partial_migration_target->details == current->details) { pe_rsc_trace(rsc, "Filtered %s -> %s %s", current->details->uname, next->details->uname, rsc->id); continue; } else { pe_rsc_trace(rsc, "Forced on %s %s", current->details->uname, rsc->id); optional = FALSE; } } pe_rsc_trace(rsc, "%s on %s", rsc->id, current->details->uname); stop = stop_action(rsc, current, optional); if(rsc->allocated_to == NULL) { pe_action_set_reason(stop, "node availability", TRUE); } if (!pcmk_is_set(rsc->flags, pe_rsc_managed)) { update_action_flags(stop, pe_action_runnable | pe_action_clear, __func__, __LINE__); } if (pcmk_is_set(data_set->flags, pe_flag_remove_after_stop)) { DeleteRsc(rsc, current, optional, data_set); } if (pcmk_is_set(rsc->flags, pe_rsc_needs_unfencing)) { pe_action_t *unfence = pe_fence_op(current, "on", TRUE, NULL, FALSE, data_set); order_actions(stop, unfence, pe_order_implies_first); if (!node_has_been_unfenced(current)) { pe_proc_err("Stopping %s until %s can be unfenced", rsc->id, current->details->uname); } } } return TRUE; } static void order_after_unfencing(pe_resource_t *rsc, pe_node_t *node, pe_action_t *action, enum pe_ordering order, pe_working_set_t *data_set) { /* When unfencing is in use, we order unfence actions before any probe or * start of resources that require unfencing, and also of fence devices. * * This might seem to violate the principle that fence devices require * only quorum. However, fence agents that unfence often don't have enough * information to even probe or start unless the node is first unfenced. */ if (is_unfence_device(rsc, data_set) || pcmk_is_set(rsc->flags, pe_rsc_needs_unfencing)) { /* Start with an optional ordering. Requiring unfencing would result in * the node being unfenced, and all its resources being stopped, * whenever a new resource is added -- which would be highly suboptimal. */ pe_action_t *unfence = pe_fence_op(node, "on", TRUE, NULL, FALSE, data_set); order_actions(unfence, action, order); if (!node_has_been_unfenced(node)) { // But unfencing is required if it has never been done char *reason = crm_strdup_printf("required by %s %s", rsc->id, action->task); trigger_unfencing(NULL, node, reason, NULL, data_set); free(reason); } } } gboolean StartRsc(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set) { pe_action_t *start = NULL; CRM_ASSERT(rsc); pe_rsc_trace(rsc, "%s on %s %d %d", rsc->id, next ? next->details->uname : "N/A", optional, next ? next->weight : 0); start = start_action(rsc, next, TRUE); order_after_unfencing(rsc, next, start, pe_order_implies_then, data_set); if (pcmk_is_set(start->flags, pe_action_runnable) && !optional) { update_action_flags(start, pe_action_optional | pe_action_clear, __func__, __LINE__); } return TRUE; } gboolean PromoteRsc(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set) { GListPtr gIter = NULL; gboolean runnable = TRUE; GListPtr action_list = NULL; CRM_ASSERT(rsc); CRM_CHECK(next != NULL, return FALSE); pe_rsc_trace(rsc, "%s on %s", rsc->id, next->details->uname); action_list = pe__resource_actions(rsc, next, RSC_START, TRUE); for (gIter = action_list; gIter != NULL; gIter = gIter->next) { pe_action_t *start = (pe_action_t *) gIter->data; if (!pcmk_is_set(start->flags, pe_action_runnable)) { runnable = FALSE; } } g_list_free(action_list); if (runnable) { promote_action(rsc, next, optional); return TRUE; } pe_rsc_debug(rsc, "%s\tPromote %s (canceled)", next->details->uname, rsc->id); action_list = pe__resource_actions(rsc, next, RSC_PROMOTE, TRUE); for (gIter = action_list; gIter != NULL; gIter = gIter->next) { pe_action_t *promote = (pe_action_t *) gIter->data; update_action_flags(promote, pe_action_runnable | pe_action_clear, __func__, __LINE__); } g_list_free(action_list); return TRUE; } gboolean DemoteRsc(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set) { GListPtr gIter = NULL; CRM_ASSERT(rsc); pe_rsc_trace(rsc, "%s", rsc->id); /* CRM_CHECK(rsc->next_role == RSC_ROLE_SLAVE, return FALSE); */ for (gIter = rsc->running_on; gIter != NULL; gIter = gIter->next) { pe_node_t *current = (pe_node_t *) gIter->data; pe_rsc_trace(rsc, "%s on %s", rsc->id, next ? next->details->uname : "N/A"); demote_action(rsc, current, optional); } return TRUE; } gboolean RoleError(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set) { CRM_ASSERT(rsc); crm_err("%s on %s", rsc->id, next ? next->details->uname : "N/A"); CRM_CHECK(FALSE, return FALSE); return FALSE; } gboolean NullOp(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set) { CRM_ASSERT(rsc); pe_rsc_trace(rsc, "%s", rsc->id); return FALSE; } gboolean DeleteRsc(pe_resource_t * rsc, pe_node_t * node, gboolean optional, pe_working_set_t * data_set) { if (pcmk_is_set(rsc->flags, pe_rsc_failed)) { pe_rsc_trace(rsc, "Resource %s not deleted from %s: failed", rsc->id, node->details->uname); return FALSE; } else if (node == NULL) { pe_rsc_trace(rsc, "Resource %s not deleted: NULL node", rsc->id); return FALSE; } else if (node->details->unclean || node->details->online == FALSE) { pe_rsc_trace(rsc, "Resource %s not deleted from %s: unrunnable", rsc->id, node->details->uname); return FALSE; } crm_notice("Removing %s from %s", rsc->id, node->details->uname); delete_action(rsc, node, optional); new_rsc_order(rsc, RSC_STOP, rsc, RSC_DELETE, optional ? pe_order_implies_then : pe_order_optional, data_set); new_rsc_order(rsc, RSC_DELETE, rsc, RSC_START, optional ? pe_order_implies_then : pe_order_optional, data_set); return TRUE; } gboolean native_create_probe(pe_resource_t * rsc, pe_node_t * node, pe_action_t * complete, gboolean force, pe_working_set_t * data_set) { enum pe_ordering flags = pe_order_optional; char *key = NULL; pe_action_t *probe = NULL; pe_node_t *running = NULL; pe_node_t *allowed = NULL; pe_resource_t *top = uber_parent(rsc); static const char *rc_master = NULL; static const char *rc_inactive = NULL; if (rc_inactive == NULL) { rc_inactive = crm_itoa(PCMK_OCF_NOT_RUNNING); rc_master = crm_itoa(PCMK_OCF_RUNNING_MASTER); } CRM_CHECK(node != NULL, return FALSE); if (!force && !pcmk_is_set(data_set->flags, pe_flag_startup_probes)) { pe_rsc_trace(rsc, "Skipping active resource detection for %s", rsc->id); return FALSE; } if (pe__is_guest_or_remote_node(node)) { const char *class = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS); if (pcmk__str_eq(class, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { pe_rsc_trace(rsc, "Skipping probe for %s on %s because Pacemaker Remote nodes cannot run stonith agents", rsc->id, node->details->id); return FALSE; } else if (pe__is_guest_node(node) && pe__resource_contains_guest_node(data_set, rsc)) { pe_rsc_trace(rsc, "Skipping probe for %s on %s because guest nodes cannot run resources containing guest nodes", rsc->id, node->details->id); return FALSE; } else if (rsc->is_remote_node) { pe_rsc_trace(rsc, "Skipping probe for %s on %s because Pacemaker Remote nodes cannot host remote connections", rsc->id, node->details->id); return FALSE; } } if (rsc->children) { GListPtr gIter = NULL; gboolean any_created = FALSE; for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; any_created = child_rsc->cmds->create_probe(child_rsc, node, complete, force, data_set) || any_created; } return any_created; } else if ((rsc->container) && (!rsc->is_remote_node)) { pe_rsc_trace(rsc, "Skipping %s: it is within container %s", rsc->id, rsc->container->id); return FALSE; } if (pcmk_is_set(rsc->flags, pe_rsc_orphan)) { pe_rsc_trace(rsc, "Skipping orphan: %s", rsc->id); return FALSE; } // Check whether resource is already known on node if (!force && g_hash_table_lookup(rsc->known_on, node->details->id)) { pe_rsc_trace(rsc, "Skipping known: %s on %s", rsc->id, node->details->uname); return FALSE; } allowed = g_hash_table_lookup(rsc->allowed_nodes, node->details->id); if (rsc->exclusive_discover || top->exclusive_discover) { if (allowed == NULL) { /* exclusive discover is enabled and this node is not in the allowed list. */ pe_rsc_trace(rsc, "Skipping probe for %s on node %s, A", rsc->id, node->details->id); return FALSE; } else if (allowed->rsc_discover_mode != pe_discover_exclusive) { /* exclusive discover is enabled and this node is not marked * as a node this resource should be discovered on */ pe_rsc_trace(rsc, "Skipping probe for %s on node %s, B", rsc->id, node->details->id); return FALSE; } } if(allowed == NULL && node->rsc_discover_mode == pe_discover_never) { /* If this node was allowed to host this resource it would * have been explicitly added to the 'allowed_nodes' list. * However it wasn't and the node has discovery disabled, so * no need to probe for this resource. */ pe_rsc_trace(rsc, "Skipping probe for %s on node %s, C", rsc->id, node->details->id); return FALSE; } if (allowed && allowed->rsc_discover_mode == pe_discover_never) { /* this resource is marked as not needing to be discovered on this node */ pe_rsc_trace(rsc, "Skipping probe for %s on node %s, discovery mode", rsc->id, node->details->id); return FALSE; } if (pe__is_guest_node(node)) { pe_resource_t *remote = node->details->remote_rsc->container; if(remote->role == RSC_ROLE_STOPPED) { /* If the container is stopped, then we know anything that * might have been inside it is also stopped and there is * no need to probe. * * If we don't know the container's state on the target * either: * * - the container is running, the transition will abort * and we'll end up in a different case next time, or * * - the container is stopped * * Either way there is no need to probe. * */ if(remote->allocated_to && g_hash_table_lookup(remote->known_on, remote->allocated_to->details->id) == NULL) { /* For safety, we order the 'rsc' start after 'remote' * has been probed. * * Using 'top' helps for groups, but we may need to * follow the start's ordering chain backwards. */ custom_action_order(remote, pcmk__op_key(remote->id, RSC_STATUS, 0), NULL, top, pcmk__op_key(top->id, RSC_START, 0), NULL, pe_order_optional, data_set); } pe_rsc_trace(rsc, "Skipping probe for %s on node %s, %s is stopped", rsc->id, node->details->id, remote->id); return FALSE; /* Here we really we want to check if remote->stop is required, * but that information doesn't exist yet */ } else if(node->details->remote_requires_reset || node->details->unclean || pcmk_is_set(remote->flags, pe_rsc_failed) || remote->next_role == RSC_ROLE_STOPPED || (remote->allocated_to && pe_find_node(remote->running_on, remote->allocated_to->details->uname) == NULL) ) { /* The container is stopping or restarting, don't start * 'rsc' until 'remote' stops as this also implies that * 'rsc' is stopped - avoiding the need to probe */ custom_action_order(remote, pcmk__op_key(remote->id, RSC_STOP, 0), NULL, top, pcmk__op_key(top->id, RSC_START, 0), NULL, pe_order_optional, data_set); pe_rsc_trace(rsc, "Skipping probe for %s on node %s, %s is stopping, restarting or moving", rsc->id, node->details->id, remote->id); return FALSE; /* } else { * The container is running so there is no problem probing it */ } } key = pcmk__op_key(rsc->id, RSC_STATUS, 0); probe = custom_action(rsc, key, RSC_STATUS, node, FALSE, TRUE, data_set); update_action_flags(probe, pe_action_optional | pe_action_clear, __func__, __LINE__); order_after_unfencing(rsc, node, probe, pe_order_optional, data_set); /* * We need to know if it's running_on (not just known_on) this node * to correctly determine the target rc. */ running = pe_find_node_id(rsc->running_on, node->details->id); if (running == NULL) { add_hash_param(probe->meta, XML_ATTR_TE_TARGET_RC, rc_inactive); } else if (rsc->role == RSC_ROLE_MASTER) { add_hash_param(probe->meta, XML_ATTR_TE_TARGET_RC, rc_master); } crm_debug("Probing %s on %s (%s) %d %p", rsc->id, node->details->uname, role2text(rsc->role), pcmk_is_set(probe->flags, pe_action_runnable), rsc->running_on); if (is_unfence_device(rsc, data_set) || !pe_rsc_is_clone(top)) { top = rsc; } else { crm_trace("Probing %s on %s (%s) as %s", rsc->id, node->details->uname, role2text(rsc->role), top->id); } if (!pcmk_is_set(probe->flags, pe_action_runnable) && (rsc->running_on == NULL)) { /* Prevent the start from occurring if rsc isn't active, but * don't cause it to stop if it was active already */ pe__set_order_flags(flags, pe_order_runnable_left); } custom_action_order(rsc, NULL, probe, top, pcmk__op_key(top->id, RSC_START, 0), NULL, flags, data_set); /* Before any reloads, if they exist */ custom_action_order(rsc, NULL, probe, top, reload_key(rsc), NULL, pe_order_optional, data_set); #if 0 // complete is always null currently if (!is_unfence_device(rsc, data_set)) { /* Normally rsc.start depends on probe complete which depends * on rsc.probe. But this can't be the case for fence devices * with unfencing, as it would create graph loops. * * So instead we explicitly order 'rsc.probe then rsc.start' */ order_actions(probe, complete, pe_order_implies_then); } #endif return TRUE; } /*! * \internal * \brief Check whether a resource is known on a particular node * * \param[in] rsc Resource to check * \param[in] node Node to check * * \return TRUE if resource (or parent if an anonymous clone) is known */ static bool rsc_is_known_on(pe_resource_t *rsc, const pe_node_t *node) { if (pe_hash_table_lookup(rsc->known_on, node->details->id)) { return TRUE; } else if ((rsc->variant == pe_native) && pe_rsc_is_anon_clone(rsc->parent) && pe_hash_table_lookup(rsc->parent->known_on, node->details->id)) { /* We check only the parent, not the uber-parent, because we cannot * assume that the resource is known if it is in an anonymously cloned * group (which may be only partially known). */ return TRUE; } return FALSE; } /*! * \internal * \brief Order a resource's start and promote actions relative to fencing * * \param[in] rsc Resource to be ordered * \param[in] stonith_op Fence action * \param[in] data_set Cluster information */ static void native_start_constraints(pe_resource_t * rsc, pe_action_t * stonith_op, pe_working_set_t * data_set) { pe_node_t *target; GListPtr gIter = NULL; CRM_CHECK(stonith_op && stonith_op->node, return); target = stonith_op->node; for (gIter = rsc->actions; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; switch (action->needs) { case rsc_req_nothing: // Anything other than start or promote requires nothing break; case rsc_req_stonith: order_actions(stonith_op, action, pe_order_optional); break; case rsc_req_quorum: if (pcmk__str_eq(action->task, RSC_START, pcmk__str_casei) && pe_hash_table_lookup(rsc->allowed_nodes, target->details->id) && !rsc_is_known_on(rsc, target)) { /* If we don't know the status of the resource on the node * we're about to shoot, we have to assume it may be active * there. Order the resource start after the fencing. This * is analogous to waiting for all the probes for a resource * to complete before starting it. * * The most likely explanation is that the DC died and took * its status with it. */ pe_rsc_debug(rsc, "Ordering %s after %s recovery", action->uuid, target->details->uname); order_actions(stonith_op, action, pe_order_optional | pe_order_runnable_left); } break; } } } static void native_stop_constraints(pe_resource_t * rsc, pe_action_t * stonith_op, pe_working_set_t * data_set) { GListPtr gIter = NULL; GListPtr action_list = NULL; bool order_implicit = false; pe_resource_t *top = uber_parent(rsc); pe_action_t *parent_stop = NULL; pe_node_t *target; CRM_CHECK(stonith_op && stonith_op->node, return); target = stonith_op->node; /* Get a list of stop actions potentially implied by the fencing */ action_list = pe__resource_actions(rsc, target, RSC_STOP, FALSE); /* If resource requires fencing, implicit actions must occur after fencing. * * Implied stops and demotes of resources running on guest nodes are always * ordered after fencing, even if the resource does not require fencing, * because guest node "fencing" is actually just a resource stop. */ if (pcmk_is_set(rsc->flags, pe_rsc_needs_fencing) || pe__is_guest_node(target)) { order_implicit = true; } if (action_list && order_implicit) { parent_stop = find_first_action(top->actions, NULL, RSC_STOP, NULL); } for (gIter = action_list; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; // The stop would never complete, so convert it into a pseudo-action. update_action_flags(action, pe_action_pseudo|pe_action_runnable, __func__, __LINE__); if (order_implicit) { update_action_flags(action, pe_action_implied_by_stonith, __func__, __LINE__); /* Order the stonith before the parent stop (if any). * * Also order the stonith before the resource stop, unless the * resource is inside a bundle -- that would cause a graph loop. * We can rely on the parent stop's ordering instead. * * User constraints must not order a resource in a guest node * relative to the guest node container resource. The * pe_order_preserve flag marks constraints as generated by the * cluster and thus immune to that check (and is irrelevant if * target is not a guest). */ if (!pe_rsc_is_bundled(rsc)) { order_actions(stonith_op, action, pe_order_preserve); } order_actions(stonith_op, parent_stop, pe_order_preserve); } if (pcmk_is_set(rsc->flags, pe_rsc_failed)) { crm_notice("Stop of failed resource %s is implicit %s %s is fenced", rsc->id, (order_implicit? "after" : "because"), target->details->uname); } else { crm_info("%s is implicit %s %s is fenced", action->uuid, (order_implicit? "after" : "because"), target->details->uname); } if (pcmk_is_set(rsc->flags, pe_rsc_notify)) { /* Create a second notification that will be delivered * immediately after the node is fenced * * Basic problem: * - C is a clone active on the node to be shot and stopping on another * - R is a resource that depends on C * * + C.stop depends on R.stop * + C.stopped depends on STONITH * + C.notify depends on C.stopped * + C.healthy depends on C.notify * + R.stop depends on C.healthy * * The extra notification here changes * + C.healthy depends on C.notify * into: * + C.healthy depends on C.notify' * + C.notify' depends on STONITH' * thus breaking the loop */ create_secondary_notification(action, rsc, stonith_op, data_set); } /* From Bug #1601, successful fencing must be an input to a failed resources stop action. However given group(rA, rB) running on nodeX and B.stop has failed, A := stop healthy resource (rA.stop) B := stop failed resource (pseudo operation B.stop) C := stonith nodeX A requires B, B requires C, C requires A This loop would prevent the cluster from making progress. This block creates the "C requires A" dependency and therefore must (at least for now) be disabled. Instead, run the block above and treat all resources on nodeX as B would be (marked as a pseudo op depending on the STONITH). TODO: Break the "A requires B" dependency in update_action() and re-enable this block } else if(is_stonith == FALSE) { crm_info("Moving healthy resource %s" " off %s before fencing", rsc->id, node->details->uname); * stop healthy resources before the * stonith op * custom_action_order( rsc, stop_key(rsc), NULL, NULL,strdup(CRM_OP_FENCE),stonith_op, pe_order_optional, data_set); */ } g_list_free(action_list); /* Get a list of demote actions potentially implied by the fencing */ action_list = pe__resource_actions(rsc, target, RSC_DEMOTE, FALSE); for (gIter = action_list; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; if (action->node->details->online == FALSE || action->node->details->unclean == TRUE || pcmk_is_set(rsc->flags, pe_rsc_failed)) { if (pcmk_is_set(rsc->flags, pe_rsc_failed)) { pe_rsc_info(rsc, "Demote of failed resource %s is implicit after %s is fenced", rsc->id, target->details->uname); } else { pe_rsc_info(rsc, "%s is implicit after %s is fenced", action->uuid, target->details->uname); } /* The demote would never complete and is now implied by the * fencing, so convert it into a pseudo-action. */ update_action_flags(action, pe_action_pseudo|pe_action_runnable, __func__, __LINE__); if (pe_rsc_is_bundled(rsc)) { /* Do nothing, let the recovery be ordered after the parent's implied stop */ } else if (order_implicit) { order_actions(stonith_op, action, pe_order_preserve|pe_order_optional); } } } g_list_free(action_list); } void rsc_stonith_ordering(pe_resource_t * rsc, pe_action_t * stonith_op, pe_working_set_t * data_set) { if (rsc->children) { GListPtr gIter = NULL; for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; rsc_stonith_ordering(child_rsc, stonith_op, data_set); } } else if (!pcmk_is_set(rsc->flags, pe_rsc_managed)) { pe_rsc_trace(rsc, "Skipping fencing constraints for unmanaged resource: %s", rsc->id); } else { native_start_constraints(rsc, stonith_op, data_set); native_stop_constraints(rsc, stonith_op, data_set); } } void ReloadRsc(pe_resource_t * rsc, pe_node_t *node, pe_working_set_t * data_set) { GListPtr gIter = NULL; pe_action_t *reload = NULL; if (rsc->children) { for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; ReloadRsc(child_rsc, node, data_set); } return; } else if (rsc->variant > pe_native) { /* Complex resource with no children */ return; } else if (!pcmk_is_set(rsc->flags, pe_rsc_managed)) { pe_rsc_trace(rsc, "%s: unmanaged", rsc->id); return; } else if (pcmk_is_set(rsc->flags, pe_rsc_failed)) { /* We don't need to specify any particular actions here, normal failure * recovery will apply. */ pe_rsc_trace(rsc, "%s: preventing reload because failed", rsc->id); return; } else if (pcmk_is_set(rsc->flags, pe_rsc_start_pending)) { /* If a resource's configuration changed while a start was pending, * force a full restart. */ pe_rsc_trace(rsc, "%s: preventing reload because start pending", rsc->id); stop_action(rsc, node, FALSE); return; } else if (node == NULL) { pe_rsc_trace(rsc, "%s: not active", rsc->id); return; } pe_rsc_trace(rsc, "Processing %s", rsc->id); pe__set_resource_flags(rsc, pe_rsc_reload); reload = custom_action( rsc, reload_key(rsc), CRMD_ACTION_RELOAD, node, FALSE, TRUE, data_set); pe_action_set_reason(reload, "resource definition change", FALSE); custom_action_order(NULL, NULL, reload, rsc, stop_key(rsc), NULL, pe_order_optional|pe_order_then_cancels_first, data_set); custom_action_order(NULL, NULL, reload, rsc, demote_key(rsc), NULL, pe_order_optional|pe_order_then_cancels_first, data_set); } void native_append_meta(pe_resource_t * rsc, xmlNode * xml) { char *value = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_INCARNATION); pe_resource_t *parent; if (value) { char *name = NULL; name = crm_meta_name(XML_RSC_ATTR_INCARNATION); crm_xml_add(xml, name, value); free(name); } value = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_REMOTE_NODE); if (value) { char *name = NULL; name = crm_meta_name(XML_RSC_ATTR_REMOTE_NODE); crm_xml_add(xml, name, value); free(name); } for (parent = rsc; parent != NULL; parent = parent->parent) { if (parent->container) { crm_xml_add(xml, CRM_META"_"XML_RSC_ATTR_CONTAINER, parent->container->id); } } } diff --git a/lib/pacemaker/pcmk_sched_promotable.c b/lib/pacemaker/pcmk_sched_promotable.c index b976344c53..a0eeaadd26 100644 --- a/lib/pacemaker/pcmk_sched_promotable.c +++ b/lib/pacemaker/pcmk_sched_promotable.c @@ -1,1042 +1,1033 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #define VARIANT_CLONE 1 #include extern gint sort_clone_instance(gconstpointer a, gconstpointer b, gpointer data_set); static void child_promoting_constraints(clone_variant_data_t * clone_data, enum pe_ordering type, pe_resource_t * rsc, pe_resource_t * child, pe_resource_t * last, pe_working_set_t * data_set) { if (child == NULL) { if (clone_data->ordered && last != NULL) { pe_rsc_trace(rsc, "Ordered version (last node)"); /* last child promote before promoted started */ new_rsc_order(last, RSC_PROMOTE, rsc, RSC_PROMOTED, type, data_set); } return; } /* child promote before global promoted */ new_rsc_order(child, RSC_PROMOTE, rsc, RSC_PROMOTED, type, data_set); /* global promote before child promote */ new_rsc_order(rsc, RSC_PROMOTE, child, RSC_PROMOTE, type, data_set); if (clone_data->ordered) { pe_rsc_trace(rsc, "Ordered version"); if (last == NULL) { /* global promote before first child promote */ last = rsc; } /* else: child/child relative promote */ order_start_start(last, child, type); new_rsc_order(last, RSC_PROMOTE, child, RSC_PROMOTE, type, data_set); } else { pe_rsc_trace(rsc, "Un-ordered version"); } } static void child_demoting_constraints(clone_variant_data_t * clone_data, enum pe_ordering type, pe_resource_t * rsc, pe_resource_t * child, pe_resource_t * last, pe_working_set_t * data_set) { if (child == NULL) { if (clone_data->ordered && last != NULL) { pe_rsc_trace(rsc, "Ordered version (last node)"); /* global demote before first child demote */ new_rsc_order(rsc, RSC_DEMOTE, last, RSC_DEMOTE, pe_order_optional, data_set); } return; } /* child demote before global demoted */ new_rsc_order(child, RSC_DEMOTE, rsc, RSC_DEMOTED, pe_order_implies_then_printed, data_set); /* global demote before child demote */ new_rsc_order(rsc, RSC_DEMOTE, child, RSC_DEMOTE, pe_order_implies_first_printed, data_set); if (clone_data->ordered && last != NULL) { pe_rsc_trace(rsc, "Ordered version"); /* child/child relative demote */ new_rsc_order(child, RSC_DEMOTE, last, RSC_DEMOTE, type, data_set); } else if (clone_data->ordered) { pe_rsc_trace(rsc, "Ordered version (1st node)"); /* first child stop before global stopped */ new_rsc_order(child, RSC_DEMOTE, rsc, RSC_DEMOTED, type, data_set); } else { pe_rsc_trace(rsc, "Un-ordered version"); } } static void check_promotable_actions(pe_resource_t *rsc, gboolean *demoting, gboolean *promoting) { GListPtr gIter = NULL; if (rsc->children) { gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child = (pe_resource_t *) gIter->data; check_promotable_actions(child, demoting, promoting); } return; } CRM_ASSERT(demoting != NULL); CRM_ASSERT(promoting != NULL); gIter = rsc->actions; for (; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; if (*promoting && *demoting) { return; } else if (pcmk_is_set(action->flags, pe_action_optional)) { continue; } else if (pcmk__str_eq(RSC_DEMOTE, action->task, pcmk__str_casei)) { *demoting = TRUE; } else if (pcmk__str_eq(RSC_PROMOTE, action->task, pcmk__str_casei)) { *promoting = TRUE; } } } static void apply_master_location(pe_resource_t *child, GListPtr location_constraints, pe_node_t *chosen) { CRM_CHECK(child && chosen, return); for (GListPtr gIter = location_constraints; gIter; gIter = gIter->next) { pe_node_t *cons_node = NULL; pe__location_t *cons = gIter->data; if (cons->role_filter == RSC_ROLE_MASTER) { pe_rsc_trace(child, "Applying %s to %s", cons->id, child->id); cons_node = pe_find_node_id(cons->node_list_rh, chosen->details->id); } if (cons_node != NULL) { int new_priority = pe__add_scores(child->priority, cons_node->weight); pe_rsc_trace(child, "\t%s[%s]: %d -> %d (%d)", child->id, cons_node->details->uname, child->priority, new_priority, cons_node->weight); child->priority = new_priority; } } } static pe_node_t * guest_location(pe_node_t *guest_node) { pe_resource_t *guest = guest_node->details->remote_rsc->container; return guest->fns->location(guest, NULL, FALSE); } static pe_node_t * can_be_master(pe_resource_t * rsc) { pe_node_t *node = NULL; pe_node_t *local_node = NULL; pe_resource_t *parent = uber_parent(rsc); clone_variant_data_t *clone_data = NULL; #if 0 enum rsc_role_e role = RSC_ROLE_UNKNOWN; role = rsc->fns->state(rsc, FALSE); crm_info("%s role: %s", rsc->id, role2text(role)); #endif if (rsc->children) { GListPtr gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child = (pe_resource_t *) gIter->data; if (can_be_master(child) == NULL) { pe_rsc_trace(rsc, "Child %s of %s can't be promoted", child->id, rsc->id); return NULL; } } } node = rsc->fns->location(rsc, NULL, FALSE); if (node == NULL) { pe_rsc_trace(rsc, "%s cannot be master: not allocated", rsc->id); return NULL; } else if (!pcmk_is_set(rsc->flags, pe_rsc_managed)) { if (rsc->fns->state(rsc, TRUE) == RSC_ROLE_MASTER) { crm_notice("Forcing unmanaged master %s to remain promoted on %s", rsc->id, node->details->uname); } else { return NULL; } } else if (rsc->priority < 0) { pe_rsc_trace(rsc, "%s cannot be master: preference: %d", rsc->id, rsc->priority); return NULL; } else if (can_run_resources(node) == FALSE) { crm_trace("Node can't run any resources: %s", node->details->uname); return NULL; /* @TODO It's possible this check should be done in can_run_resources() * instead. We should investigate all its callers to figure out whether that * would be a good idea. */ } else if (pe__is_guest_node(node) && (guest_location(node) == NULL)) { pe_rsc_trace(rsc, "%s cannot be promoted: guest %s not allocated", rsc->id, node->details->remote_rsc->container->id); return NULL; } get_clone_variant_data(clone_data, parent); local_node = pe_hash_table_lookup(parent->allowed_nodes, node->details->id); if (local_node == NULL) { crm_err("%s cannot run on %s: node not allowed", rsc->id, node->details->uname); return NULL; } else if ((local_node->count < clone_data->promoted_node_max) || !pcmk_is_set(rsc->flags, pe_rsc_managed)) { return local_node; } else { pe_rsc_trace(rsc, "%s cannot be master on %s: node full", rsc->id, node->details->uname); } return NULL; } static gint sort_promotable_instance(gconstpointer a, gconstpointer b, gpointer data_set) { int rc; enum rsc_role_e role1 = RSC_ROLE_UNKNOWN; enum rsc_role_e role2 = RSC_ROLE_UNKNOWN; const pe_resource_t *resource1 = (const pe_resource_t *)a; const pe_resource_t *resource2 = (const pe_resource_t *)b; CRM_ASSERT(resource1 != NULL); CRM_ASSERT(resource2 != NULL); role1 = resource1->fns->state(resource1, TRUE); role2 = resource2->fns->state(resource2, TRUE); rc = sort_rsc_index(a, b); if (rc != 0) { crm_trace("%s %c %s (index)", resource1->id, rc < 0 ? '<' : '>', resource2->id); return rc; } if (role1 > role2) { crm_trace("%s %c %s (role)", resource1->id, '<', resource2->id); return -1; } else if (role1 < role2) { crm_trace("%s %c %s (role)", resource1->id, '>', resource2->id); return 1; } return sort_clone_instance(a, b, data_set); } static void promotion_order(pe_resource_t *rsc, pe_working_set_t *data_set) { GListPtr gIter = NULL; pe_node_t *node = NULL; pe_node_t *chosen = NULL; clone_variant_data_t *clone_data = NULL; char score[33]; size_t len = sizeof(score); get_clone_variant_data(clone_data, rsc); if (clone_data->merged_master_weights) { return; } clone_data->merged_master_weights = TRUE; pe_rsc_trace(rsc, "Merging weights for %s", rsc->id); pe__set_resource_flags(rsc, pe_rsc_merging); for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child = (pe_resource_t *) gIter->data; pe_rsc_trace(rsc, "Sort index: %s = %d", child->id, child->sort_index); } pe__show_node_weights(true, rsc, "Before", rsc->allowed_nodes); gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child = (pe_resource_t *) gIter->data; chosen = child->fns->location(child, NULL, FALSE); if (chosen == NULL || child->sort_index < 0) { pe_rsc_trace(rsc, "Skipping %s", child->id); continue; } node = (pe_node_t *) pe_hash_table_lookup(rsc->allowed_nodes, chosen->details->id); CRM_ASSERT(node != NULL); /* adds in master preferences and rsc_location.role=Master */ score2char_stack(child->sort_index, score, len); pe_rsc_trace(rsc, "Adding %s to %s from %s", score, node->details->uname, child->id); node->weight = pe__add_scores(child->sort_index, node->weight); } pe__show_node_weights(true, rsc, "Middle", rsc->allowed_nodes); gIter = rsc->rsc_cons; for (; gIter != NULL; gIter = gIter->next) { - rsc_colocation_t *constraint = (rsc_colocation_t *) gIter->data; + pcmk__colocation_t *constraint = (pcmk__colocation_t *) gIter->data; if (constraint->score == 0) { continue; } /* (re-)adds location preferences of resources that the * master instance should/must be colocated with */ if (constraint->role_lh == RSC_ROLE_MASTER) { enum pe_weights flags = constraint->score == INFINITY ? 0 : pe_weights_rollback; pe_rsc_trace(rsc, "RHS: %s with %s: %d", constraint->rsc_lh->id, constraint->rsc_rh->id, constraint->score); rsc->allowed_nodes = constraint->rsc_rh->cmds->merge_weights(constraint->rsc_rh, rsc->id, rsc->allowed_nodes, constraint->node_attribute, (float)constraint->score / INFINITY, flags); } } gIter = rsc->rsc_cons_lhs; for (; gIter != NULL; gIter = gIter->next) { - rsc_colocation_t *constraint = (rsc_colocation_t *) gIter->data; + pcmk__colocation_t *constraint = (pcmk__colocation_t *) gIter->data; - if (constraint->score == 0) { - continue; - } - - /* (re-)adds location preferences of resource that wish to be - * colocated with the master instance - */ - if (constraint->role_rh == RSC_ROLE_MASTER) { - pe_rsc_trace(rsc, "LHS: %s with %s: %d", constraint->rsc_lh->id, constraint->rsc_rh->id, - constraint->score); - rsc->allowed_nodes = - constraint->rsc_lh->cmds->merge_weights(constraint->rsc_lh, rsc->id, - rsc->allowed_nodes, - constraint->node_attribute, - (float)constraint->score / INFINITY, - (pe_weights_rollback | - pe_weights_positive)); + if (pcmk__colocation_applies(rsc, constraint, true)) { + /* (Re-)add location preferences of resource that wishes to be + * colocated with the promoted instance. + */ + rsc->allowed_nodes = constraint->rsc_lh->cmds->merge_weights(constraint->rsc_lh, + rsc->id, rsc->allowed_nodes, constraint->node_attribute, + constraint->score / (float) INFINITY, + pe_weights_rollback|pe_weights_positive); } } gIter = rsc->rsc_tickets; for (; gIter != NULL; gIter = gIter->next) { rsc_ticket_t *rsc_ticket = (rsc_ticket_t *) gIter->data; if (rsc_ticket->role_lh == RSC_ROLE_MASTER && (rsc_ticket->ticket->granted == FALSE || rsc_ticket->ticket->standby)) { resource_location(rsc, NULL, -INFINITY, "__stateful_without_ticket__", data_set); } } pe__show_node_weights(true, rsc, "After", rsc->allowed_nodes); /* write them back and sort */ gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child = (pe_resource_t *) gIter->data; chosen = child->fns->location(child, NULL, FALSE); if (!pcmk_is_set(child->flags, pe_rsc_managed) && (child->next_role == RSC_ROLE_MASTER)) { child->sort_index = INFINITY; } else if (chosen == NULL || child->sort_index < 0) { pe_rsc_trace(rsc, "%s: %d", child->id, child->sort_index); } else { node = (pe_node_t *) pe_hash_table_lookup(rsc->allowed_nodes, chosen->details->id); CRM_ASSERT(node != NULL); child->sort_index = node->weight; } pe_rsc_trace(rsc, "Set sort index: %s = %d", child->id, child->sort_index); } rsc->children = g_list_sort_with_data(rsc->children, sort_promotable_instance, data_set); pe__clear_resource_flags(rsc, pe_rsc_merging); } static gboolean filter_anonymous_instance(pe_resource_t *rsc, const pe_node_t *node) { GListPtr rIter = NULL; char *key = clone_strip(rsc->id); pe_resource_t *parent = uber_parent(rsc); for (rIter = parent->children; rIter; rIter = rIter->next) { /* If there is an active instance on the node, only it receives the * promotion score. Use ->find_rsc() in case this is a cloned group. */ pe_resource_t *child = rIter->data; pe_resource_t *active = parent->fns->find_rsc(child, key, node, pe_find_clone|pe_find_current); if(rsc == active) { pe_rsc_trace(rsc, "Found %s for %s active on %s: done", active->id, key, node->details->uname); free(key); return TRUE; } else if(active) { pe_rsc_trace(rsc, "Found %s for %s on %s: not %s", active->id, key, node->details->uname, rsc->id); free(key); return FALSE; } else { pe_rsc_trace(rsc, "%s on %s: not active", key, node->details->uname); } } for (rIter = parent->children; rIter; rIter = rIter->next) { pe_resource_t *child = rIter->data; /* * We know it's not running, but any score will still count if * the instance has been probed on $node * * Again use ->find_rsc() because we might be a cloned group * and knowing that other members of the group are known here * implies nothing */ rsc = parent->fns->find_rsc(child, key, NULL, pe_find_clone); CRM_LOG_ASSERT(rsc); if(rsc) { pe_rsc_trace(rsc, "Checking %s for %s on %s", rsc->id, key, node->details->uname); if (g_hash_table_lookup(rsc->known_on, node->details->id)) { free(key); return TRUE; } } } free(key); return FALSE; } static const char * lookup_promotion_score(pe_resource_t *rsc, const pe_node_t *node, const char *name) { const char *attr_value = NULL; if (node && name) { char *attr_name = crm_strdup_printf("master-%s", name); attr_value = pe_node_attribute_calculated(node, attr_name, rsc); free(attr_name); } return attr_value; } static int promotion_score(pe_resource_t *rsc, const pe_node_t *node, int not_set_value) { char *name = rsc->id; const char *attr_value = NULL; int score = not_set_value; pe_node_t *match = NULL; CRM_CHECK(node != NULL, return not_set_value); if (rsc->children) { GListPtr gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child = (pe_resource_t *) gIter->data; int c_score = promotion_score(child, node, not_set_value); if (score == not_set_value) { score = c_score; } else { score += c_score; } } return score; } if (!pcmk_is_set(rsc->flags, pe_rsc_unique) && filter_anonymous_instance(rsc, node)) { pe_rsc_trace(rsc, "Anonymous clone %s is allowed on %s", rsc->id, node->details->uname); } else if (rsc->running_on || g_hash_table_size(rsc->known_on)) { /* If we've probed and/or started the resource anywhere, consider * promotion scores only from nodes where we know the status. However, * if the status of all nodes is unknown (e.g. cluster startup), * skip this code, to make sure we take into account any permanent * promotion scores set previously. */ pe_node_t *known = pe_hash_table_lookup(rsc->known_on, node->details->id); match = pe_find_node_id(rsc->running_on, node->details->id); if ((match == NULL) && (known == NULL)) { pe_rsc_trace(rsc, "skipping %s (aka. %s) promotion score on %s because inactive", rsc->id, rsc->clone_name, node->details->uname); return score; } } match = pe_hash_table_lookup(rsc->allowed_nodes, node->details->id); if (match == NULL) { return score; } else if (match->weight < 0) { pe_rsc_trace(rsc, "%s on %s has score: %d - ignoring", rsc->id, match->details->uname, match->weight); return score; } if (rsc->clone_name) { /* Use the name the lrm knows this resource as, * since that's what crm_master would have used too */ name = rsc->clone_name; } attr_value = lookup_promotion_score(rsc, node, name); pe_rsc_trace(rsc, "promotion score for %s on %s = %s", name, node->details->uname, crm_str(attr_value)); if ((attr_value == NULL) && !pcmk_is_set(rsc->flags, pe_rsc_unique)) { /* If we don't have any LRM history yet, we won't have clone_name -- in * that case, for anonymous clones, try the resource name without any * instance number. */ name = clone_strip(rsc->id); if (strcmp(rsc->id, name)) { attr_value = lookup_promotion_score(rsc, node, name); pe_rsc_trace(rsc, "stripped promotion score for %s on %s = %s", name, node->details->uname, crm_str(attr_value)); } free(name); } if (attr_value != NULL) { score = char2score(attr_value); } return score; } void apply_master_prefs(pe_resource_t *rsc) { int score, new_score; GListPtr gIter = rsc->children; clone_variant_data_t *clone_data = NULL; get_clone_variant_data(clone_data, rsc); if (clone_data->applied_master_prefs) { /* Make sure we only do this once */ return; } clone_data->applied_master_prefs = TRUE; for (; gIter != NULL; gIter = gIter->next) { GHashTableIter iter; pe_node_t *node = NULL; pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; g_hash_table_iter_init(&iter, child_rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { if (can_run_resources(node) == FALSE) { /* This node will never be promoted to master, * so don't apply the promotion score as that may * lead to clone shuffling */ continue; } score = promotion_score(child_rsc, node, 0); if (score > 0) { new_score = pe__add_scores(node->weight, score); if (new_score != node->weight) { pe_rsc_trace(rsc, "\t%s: Updating preference for %s (%d->%d)", child_rsc->id, node->details->uname, node->weight, new_score); node->weight = new_score; } } new_score = QB_MAX(child_rsc->priority, score); if (new_score != child_rsc->priority) { pe_rsc_trace(rsc, "\t%s: Updating priority (%d->%d)", child_rsc->id, child_rsc->priority, new_score); child_rsc->priority = new_score; } } } } static void set_role_slave(pe_resource_t * rsc, gboolean current) { GListPtr gIter = rsc->children; if (current) { if (rsc->role == RSC_ROLE_STARTED) { rsc->role = RSC_ROLE_SLAVE; } } else { GListPtr allocated = NULL; rsc->fns->location(rsc, &allocated, FALSE); if (allocated) { rsc->next_role = RSC_ROLE_SLAVE; } else { rsc->next_role = RSC_ROLE_STOPPED; } g_list_free(allocated); } for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; set_role_slave(child_rsc, current); } } static void set_role_master(pe_resource_t * rsc) { GListPtr gIter = rsc->children; if (rsc->next_role == RSC_ROLE_UNKNOWN) { rsc->next_role = RSC_ROLE_MASTER; } for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; set_role_master(child_rsc); } } pe_node_t * pcmk__set_instance_roles(pe_resource_t *rsc, pe_working_set_t *data_set) { int promoted = 0; GListPtr gIter = NULL; GListPtr gIter2 = NULL; GHashTableIter iter; pe_node_t *node = NULL; pe_node_t *chosen = NULL; enum rsc_role_e next_role = RSC_ROLE_UNKNOWN; char score[33]; size_t len = sizeof(score); clone_variant_data_t *clone_data = NULL; get_clone_variant_data(clone_data, rsc); /* count now tracks the number of masters allocated */ g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { node->count = 0; } /* * assign priority */ for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) { GListPtr list = NULL; pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; pe_rsc_trace(rsc, "Assigning priority for %s: %s", child_rsc->id, role2text(child_rsc->next_role)); if (child_rsc->fns->state(child_rsc, TRUE) == RSC_ROLE_STARTED) { set_role_slave(child_rsc, TRUE); } chosen = child_rsc->fns->location(child_rsc, &list, FALSE); if (pcmk__list_of_multiple(list)) { pcmk__config_err("Cannot promote non-colocated child %s", child_rsc->id); } g_list_free(list); if (chosen == NULL) { continue; } next_role = child_rsc->fns->state(child_rsc, FALSE); switch (next_role) { case RSC_ROLE_STARTED: case RSC_ROLE_UNKNOWN: /* * Default to -1 if no value is set * * This allows master locations to be specified * based solely on rsc_location constraints, * but prevents anyone from being promoted if * neither a constraint nor a promotion score is present */ child_rsc->priority = promotion_score(child_rsc, chosen, -1); break; case RSC_ROLE_SLAVE: case RSC_ROLE_STOPPED: child_rsc->priority = -INFINITY; break; case RSC_ROLE_MASTER: /* We will arrive here if we're re-creating actions after a stonith */ break; default: CRM_CHECK(FALSE /* unhandled */ , crm_err("Unknown resource role: %d for %s", next_role, child_rsc->id)); } apply_master_location(child_rsc, child_rsc->rsc_location, chosen); apply_master_location(child_rsc, rsc->rsc_location, chosen); for (gIter2 = child_rsc->rsc_cons; gIter2 != NULL; gIter2 = gIter2->next) { - rsc_colocation_t *cons = (rsc_colocation_t *) gIter2->data; + pcmk__colocation_t *cons = (pcmk__colocation_t *) gIter2->data; if (cons->score == 0) { continue; } child_rsc->cmds->rsc_colocation_lh(child_rsc, cons->rsc_rh, cons, data_set); } child_rsc->sort_index = child_rsc->priority; pe_rsc_trace(rsc, "Assigning priority for %s: %d", child_rsc->id, child_rsc->priority); if (next_role == RSC_ROLE_MASTER) { child_rsc->sort_index = INFINITY; } } pe__show_node_weights(true, rsc, "Pre merge", rsc->allowed_nodes); promotion_order(rsc, data_set); /* mark the first N as masters */ for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; score2char_stack(child_rsc->sort_index, score, len); chosen = child_rsc->fns->location(child_rsc, NULL, FALSE); if (show_scores) { if (pcmk_is_set(data_set->flags, pe_flag_stdout)) { printf("%s promotion score on %s: %s\n", child_rsc->id, (chosen? chosen->details->uname : "none"), score); } } else { pe_rsc_trace(rsc, "%s promotion score on %s: %s", child_rsc->id, (chosen? chosen->details->uname : "none"), score); } chosen = NULL; /* nuke 'chosen' so that we don't promote more than the * required number of instances */ if (child_rsc->sort_index < 0) { pe_rsc_trace(rsc, "Not supposed to promote child: %s", child_rsc->id); } else if ((promoted < clone_data->promoted_max) || !pcmk_is_set(rsc->flags, pe_rsc_managed)) { chosen = can_be_master(child_rsc); } pe_rsc_debug(rsc, "%s promotion score: %d", child_rsc->id, child_rsc->priority); if (chosen == NULL) { set_role_slave(child_rsc, FALSE); continue; } else if(child_rsc->role < RSC_ROLE_MASTER && !pcmk_is_set(data_set->flags, pe_flag_have_quorum) && data_set->no_quorum_policy == no_quorum_freeze) { crm_notice("Resource %s cannot be elevated from %s to %s: no-quorum-policy=freeze", child_rsc->id, role2text(child_rsc->role), role2text(child_rsc->next_role)); set_role_slave(child_rsc, FALSE); continue; } chosen->count++; pe_rsc_info(rsc, "Promoting %s (%s %s)", child_rsc->id, role2text(child_rsc->role), chosen->details->uname); set_role_master(child_rsc); promoted++; } pe_rsc_info(rsc, "%s: Promoted %d instances of a possible %d to master", rsc->id, promoted, clone_data->promoted_max); return NULL; } void create_promotable_actions(pe_resource_t * rsc, pe_working_set_t * data_set) { pe_action_t *action = NULL; GListPtr gIter = rsc->children; pe_action_t *action_complete = NULL; gboolean any_promoting = FALSE; gboolean any_demoting = FALSE; pe_resource_t *last_promote_rsc = NULL; pe_resource_t *last_demote_rsc = NULL; clone_variant_data_t *clone_data = NULL; get_clone_variant_data(clone_data, rsc); pe_rsc_debug(rsc, "Creating actions for %s", rsc->id); for (; gIter != NULL; gIter = gIter->next) { gboolean child_promoting = FALSE; gboolean child_demoting = FALSE; pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; pe_rsc_trace(rsc, "Creating actions for %s", child_rsc->id); child_rsc->cmds->create_actions(child_rsc, data_set); check_promotable_actions(child_rsc, &child_demoting, &child_promoting); any_demoting = any_demoting || child_demoting; any_promoting = any_promoting || child_promoting; pe_rsc_trace(rsc, "Created actions for %s: %d %d", child_rsc->id, child_promoting, child_demoting); } /* promote */ action = create_pseudo_resource_op(rsc, RSC_PROMOTE, !any_promoting, TRUE, data_set); action_complete = create_pseudo_resource_op(rsc, RSC_PROMOTED, !any_promoting, TRUE, data_set); action_complete->priority = INFINITY; child_promoting_constraints(clone_data, pe_order_optional, rsc, NULL, last_promote_rsc, data_set); if (clone_data->promote_notify == NULL) { clone_data->promote_notify = create_notification_boundaries(rsc, RSC_PROMOTE, action, action_complete, data_set); } /* demote */ action = create_pseudo_resource_op(rsc, RSC_DEMOTE, !any_demoting, TRUE, data_set); action_complete = create_pseudo_resource_op(rsc, RSC_DEMOTED, !any_demoting, TRUE, data_set); action_complete->priority = INFINITY; child_demoting_constraints(clone_data, pe_order_optional, rsc, NULL, last_demote_rsc, data_set); if (clone_data->demote_notify == NULL) { clone_data->demote_notify = create_notification_boundaries(rsc, RSC_DEMOTE, action, action_complete, data_set); if (clone_data->promote_notify) { /* If we ever wanted groups to have notifications we'd need to move this to native_internal_constraints() one day * Requires exposing *_notify */ order_actions(clone_data->stop_notify->post_done, clone_data->promote_notify->pre, pe_order_optional); order_actions(clone_data->start_notify->post_done, clone_data->promote_notify->pre, pe_order_optional); order_actions(clone_data->demote_notify->post_done, clone_data->promote_notify->pre, pe_order_optional); order_actions(clone_data->demote_notify->post_done, clone_data->start_notify->pre, pe_order_optional); order_actions(clone_data->demote_notify->post_done, clone_data->stop_notify->pre, pe_order_optional); } } /* restore the correct priority */ gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; child_rsc->priority = rsc->priority; } } void promote_demote_constraints(pe_resource_t *rsc, pe_working_set_t *data_set) { /* global stopped before start */ new_rsc_order(rsc, RSC_STOPPED, rsc, RSC_START, pe_order_optional, data_set); /* global stopped before promote */ new_rsc_order(rsc, RSC_STOPPED, rsc, RSC_PROMOTE, pe_order_optional, data_set); /* global demoted before start */ new_rsc_order(rsc, RSC_DEMOTED, rsc, RSC_START, pe_order_optional, data_set); /* global started before promote */ new_rsc_order(rsc, RSC_STARTED, rsc, RSC_PROMOTE, pe_order_optional, data_set); /* global demoted before stop */ new_rsc_order(rsc, RSC_DEMOTED, rsc, RSC_STOP, pe_order_optional, data_set); /* global demote before demoted */ new_rsc_order(rsc, RSC_DEMOTE, rsc, RSC_DEMOTED, pe_order_optional, data_set); /* global demoted before promote */ new_rsc_order(rsc, RSC_DEMOTED, rsc, RSC_PROMOTE, pe_order_optional, data_set); } void promotable_constraints(pe_resource_t * rsc, pe_working_set_t * data_set) { GListPtr gIter = rsc->children; pe_resource_t *last_rsc = NULL; clone_variant_data_t *clone_data = NULL; get_clone_variant_data(clone_data, rsc); promote_demote_constraints(rsc, data_set); for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; /* child demote before promote */ new_rsc_order(child_rsc, RSC_DEMOTE, child_rsc, RSC_PROMOTE, pe_order_optional, data_set); child_promoting_constraints(clone_data, pe_order_optional, rsc, child_rsc, last_rsc, data_set); child_demoting_constraints(clone_data, pe_order_optional, rsc, child_rsc, last_rsc, data_set); last_rsc = child_rsc; } } static void node_hash_update_one(GHashTable * hash, pe_node_t * other, const char *attr, int score) { GHashTableIter iter; pe_node_t *node = NULL; const char *value = NULL; if (other == NULL) { return; } else if (attr == NULL) { attr = CRM_ATTR_UNAME; } value = pe_node_attribute_raw(other, attr); g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { const char *tmp = pe_node_attribute_raw(node, attr); if (pcmk__str_eq(value, tmp, pcmk__str_casei)) { crm_trace("%s: %d + %d", node->details->uname, node->weight, other->weight); node->weight = pe__add_scores(node->weight, score); } } } void promotable_colocation_rh(pe_resource_t *rsc_lh, pe_resource_t *rsc_rh, - rsc_colocation_t *constraint, + pcmk__colocation_t *constraint, pe_working_set_t *data_set) { GListPtr gIter = NULL; if (constraint->score == 0) { return; } if (pcmk_is_set(rsc_lh->flags, pe_rsc_provisional)) { GListPtr rhs = NULL; for (gIter = rsc_rh->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; pe_node_t *chosen = child_rsc->fns->location(child_rsc, NULL, FALSE); enum rsc_role_e next_role = child_rsc->fns->state(child_rsc, FALSE); pe_rsc_trace(rsc_rh, "Processing: %s", child_rsc->id); if (chosen != NULL && next_role == constraint->role_rh) { pe_rsc_trace(rsc_rh, "Applying: %s %s %s %d", child_rsc->id, role2text(next_role), chosen->details->uname, constraint->score); if (constraint->score < INFINITY) { node_hash_update_one(rsc_lh->allowed_nodes, chosen, constraint->node_attribute, constraint->score); } rhs = g_list_prepend(rhs, chosen); } } /* Only do this if it's not a master-master colocation * Doing this unconditionally would prevent the slaves from being started */ if (constraint->role_lh != RSC_ROLE_MASTER || constraint->role_rh != RSC_ROLE_MASTER) { if (constraint->score >= INFINITY) { node_list_exclude(rsc_lh->allowed_nodes, rhs, TRUE); } } g_list_free(rhs); } else if (constraint->role_lh == RSC_ROLE_MASTER) { pe_resource_t *rh_child = find_compatible_child(rsc_lh, rsc_rh, constraint->role_rh, FALSE, data_set); if (rh_child == NULL && constraint->score >= INFINITY) { pe_rsc_trace(rsc_lh, "%s can't be promoted %s", rsc_lh->id, constraint->id); rsc_lh->priority = -INFINITY; } else if (rh_child != NULL) { int new_priority = pe__add_scores(rsc_lh->priority, constraint->score); pe_rsc_debug(rsc_lh, "Applying %s to %s", constraint->id, rsc_lh->id); pe_rsc_debug(rsc_lh, "\t%s: %d->%d", rsc_lh->id, rsc_lh->priority, new_priority); rsc_lh->priority = new_priority; } } return; } diff --git a/lib/pacemaker/pcmk_sched_utilization.c b/lib/pacemaker/pcmk_sched_utilization.c index b46b2fac14..0df3a17770 100644 --- a/lib/pacemaker/pcmk_sched_utilization.c +++ b/lib/pacemaker/pcmk_sched_utilization.c @@ -1,471 +1,471 @@ /* * Copyright 2014-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include static GListPtr find_colocated_rscs(GListPtr colocated_rscs, pe_resource_t * rsc, pe_resource_t * orig_rsc); static GListPtr group_find_colocated_rscs(GListPtr colocated_rscs, pe_resource_t * rsc, pe_resource_t * orig_rsc); static void group_add_unallocated_utilization(GHashTable * all_utilization, pe_resource_t * rsc, GListPtr all_rscs); struct compare_data { const pe_node_t *node1; const pe_node_t *node2; int result; }; static void do_compare_capacity1(gpointer key, gpointer value, gpointer user_data) { int node1_capacity = 0; int node2_capacity = 0; struct compare_data *data = user_data; node1_capacity = crm_parse_int(value, "0"); node2_capacity = crm_parse_int(g_hash_table_lookup(data->node2->details->utilization, key), "0"); if (node1_capacity > node2_capacity) { data->result--; } else if (node1_capacity < node2_capacity) { data->result++; } } static void do_compare_capacity2(gpointer key, gpointer value, gpointer user_data) { int node1_capacity = 0; int node2_capacity = 0; struct compare_data *data = user_data; if (g_hash_table_lookup_extended(data->node1->details->utilization, key, NULL, NULL)) { return; } node1_capacity = 0; node2_capacity = crm_parse_int(value, "0"); if (node1_capacity > node2_capacity) { data->result--; } else if (node1_capacity < node2_capacity) { data->result++; } } /* rc < 0 if 'node1' has more capacity remaining * rc > 0 if 'node1' has less capacity remaining */ int compare_capacity(const pe_node_t * node1, const pe_node_t * node2) { struct compare_data data; data.node1 = node1; data.node2 = node2; data.result = 0; g_hash_table_foreach(node1->details->utilization, do_compare_capacity1, &data); g_hash_table_foreach(node2->details->utilization, do_compare_capacity2, &data); return data.result; } struct calculate_data { GHashTable *current_utilization; gboolean plus; }; static void do_calculate_utilization(gpointer key, gpointer value, gpointer user_data) { const char *current = NULL; char *result = NULL; struct calculate_data *data = user_data; current = g_hash_table_lookup(data->current_utilization, key); if (data->plus) { result = crm_itoa(crm_parse_int(current, "0") + crm_parse_int(value, "0")); g_hash_table_replace(data->current_utilization, strdup(key), result); } else if (current) { result = crm_itoa(crm_parse_int(current, "0") - crm_parse_int(value, "0")); g_hash_table_replace(data->current_utilization, strdup(key), result); } } /* Specify 'plus' to FALSE when allocating * Otherwise to TRUE when deallocating */ void calculate_utilization(GHashTable * current_utilization, GHashTable * utilization, gboolean plus) { struct calculate_data data; data.current_utilization = current_utilization; data.plus = plus; g_hash_table_foreach(utilization, do_calculate_utilization, &data); } struct capacity_data { pe_node_t *node; const char *rsc_id; gboolean is_enough; }; static void check_capacity(gpointer key, gpointer value, gpointer user_data) { int required = 0; int remaining = 0; struct capacity_data *data = user_data; required = crm_parse_int(value, "0"); remaining = crm_parse_int(g_hash_table_lookup(data->node->details->utilization, key), "0"); if (required > remaining) { CRM_ASSERT(data->rsc_id); CRM_ASSERT(data->node); crm_debug("Node %s does not have enough %s for %s: required=%d remaining=%d", data->node->details->uname, (char *)key, data->rsc_id, required, remaining); data->is_enough = FALSE; } } static gboolean have_enough_capacity(pe_node_t * node, const char * rsc_id, GHashTable * utilization) { struct capacity_data data; data.node = node; data.rsc_id = rsc_id; data.is_enough = TRUE; g_hash_table_foreach(utilization, check_capacity, &data); return data.is_enough; } static void native_add_unallocated_utilization(GHashTable * all_utilization, pe_resource_t * rsc) { if (!pcmk_is_set(rsc->flags, pe_rsc_provisional)) { return; } calculate_utilization(all_utilization, rsc->utilization, TRUE); } static void add_unallocated_utilization(GHashTable * all_utilization, pe_resource_t * rsc, GListPtr all_rscs, pe_resource_t * orig_rsc) { if (!pcmk_is_set(rsc->flags, pe_rsc_provisional)) { return; } if (rsc->variant == pe_native) { pe_rsc_trace(orig_rsc, "%s: Adding %s as colocated utilization", orig_rsc->id, rsc->id); native_add_unallocated_utilization(all_utilization, rsc); } else if (rsc->variant == pe_group) { pe_rsc_trace(orig_rsc, "%s: Adding %s as colocated utilization", orig_rsc->id, rsc->id); group_add_unallocated_utilization(all_utilization, rsc, all_rscs); } else if (pe_rsc_is_clone(rsc)) { GListPtr gIter1 = NULL; gboolean existing = FALSE; /* Check if there's any child already existing in the list */ gIter1 = rsc->children; for (; gIter1 != NULL; gIter1 = gIter1->next) { pe_resource_t *child = (pe_resource_t *) gIter1->data; GListPtr gIter2 = NULL; if (g_list_find(all_rscs, child)) { existing = TRUE; } else { /* Check if there's any child of another cloned group already existing in the list */ gIter2 = child->children; for (; gIter2 != NULL; gIter2 = gIter2->next) { pe_resource_t *grandchild = (pe_resource_t *) gIter2->data; if (g_list_find(all_rscs, grandchild)) { pe_rsc_trace(orig_rsc, "%s: Adding %s as colocated utilization", orig_rsc->id, child->id); add_unallocated_utilization(all_utilization, child, all_rscs, orig_rsc); existing = TRUE; break; } } } } // rsc->children is always non-NULL but this makes static analysis happy if (!existing && (rsc->children != NULL)) { pe_resource_t *first_child = (pe_resource_t *) rsc->children->data; pe_rsc_trace(orig_rsc, "%s: Adding %s as colocated utilization", orig_rsc->id, ID(first_child->xml)); add_unallocated_utilization(all_utilization, first_child, all_rscs, orig_rsc); } } } static GHashTable * sum_unallocated_utilization(pe_resource_t * rsc, GListPtr colocated_rscs) { GListPtr gIter = NULL; GListPtr all_rscs = NULL; GHashTable *all_utilization = crm_str_table_new(); all_rscs = g_list_copy(colocated_rscs); if (g_list_find(all_rscs, rsc) == FALSE) { all_rscs = g_list_append(all_rscs, rsc); } for (gIter = all_rscs; gIter != NULL; gIter = gIter->next) { pe_resource_t *listed_rsc = (pe_resource_t *) gIter->data; if (!pcmk_is_set(listed_rsc->flags, pe_rsc_provisional)) { continue; } pe_rsc_trace(rsc, "%s: Processing unallocated colocated %s", rsc->id, listed_rsc->id); add_unallocated_utilization(all_utilization, listed_rsc, all_rscs, rsc); } g_list_free(all_rscs); return all_utilization; } static GListPtr find_colocated_rscs(GListPtr colocated_rscs, pe_resource_t * rsc, pe_resource_t * orig_rsc) { GListPtr gIter = NULL; if (rsc == NULL) { return colocated_rscs; } else if (g_list_find(colocated_rscs, rsc)) { return colocated_rscs; } crm_trace("%s: %s is supposed to be colocated with %s", orig_rsc->id, rsc->id, orig_rsc->id); colocated_rscs = g_list_append(colocated_rscs, rsc); for (gIter = rsc->rsc_cons; gIter != NULL; gIter = gIter->next) { - rsc_colocation_t *constraint = (rsc_colocation_t *) gIter->data; + pcmk__colocation_t *constraint = (pcmk__colocation_t *) gIter->data; pe_resource_t *rsc_rh = constraint->rsc_rh; /* Break colocation loop */ if (rsc_rh == orig_rsc) { continue; } if (constraint->score == INFINITY && filter_colocation_constraint(rsc, rsc_rh, constraint, TRUE) == influence_rsc_location) { if (rsc_rh->variant == pe_group) { /* Need to use group_variant_data */ colocated_rscs = group_find_colocated_rscs(colocated_rscs, rsc_rh, orig_rsc); } else { colocated_rscs = find_colocated_rscs(colocated_rscs, rsc_rh, orig_rsc); } } } for (gIter = rsc->rsc_cons_lhs; gIter != NULL; gIter = gIter->next) { - rsc_colocation_t *constraint = (rsc_colocation_t *) gIter->data; + pcmk__colocation_t *constraint = (pcmk__colocation_t *) gIter->data; pe_resource_t *rsc_lh = constraint->rsc_lh; /* Break colocation loop */ if (rsc_lh == orig_rsc) { continue; } if (pe_rsc_is_clone(rsc_lh) == FALSE && pe_rsc_is_clone(rsc)) { /* We do not know if rsc_lh will be colocated with orig_rsc in this case */ continue; } if (constraint->score == INFINITY && filter_colocation_constraint(rsc_lh, rsc, constraint, TRUE) == influence_rsc_location) { if (rsc_lh->variant == pe_group) { /* Need to use group_variant_data */ colocated_rscs = group_find_colocated_rscs(colocated_rscs, rsc_lh, orig_rsc); } else { colocated_rscs = find_colocated_rscs(colocated_rscs, rsc_lh, orig_rsc); } } } return colocated_rscs; } void process_utilization(pe_resource_t * rsc, pe_node_t ** prefer, pe_working_set_t * data_set) { CRM_CHECK(rsc && prefer && data_set, return); if (!pcmk__str_eq(data_set->placement_strategy, "default", pcmk__str_casei)) { GHashTableIter iter; GListPtr colocated_rscs = NULL; gboolean any_capable = FALSE; pe_node_t *node = NULL; colocated_rscs = find_colocated_rscs(colocated_rscs, rsc, rsc); if (colocated_rscs) { GHashTable *unallocated_utilization = NULL; char *rscs_id = crm_strdup_printf("%s and its colocated resources", rsc->id); pe_node_t *most_capable_node = NULL; unallocated_utilization = sum_unallocated_utilization(rsc, colocated_rscs); g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { if (can_run_resources(node) == FALSE || node->weight < 0) { continue; } if (have_enough_capacity(node, rscs_id, unallocated_utilization)) { any_capable = TRUE; } if (most_capable_node == NULL || compare_capacity(node, most_capable_node) < 0) { /* < 0 means 'node' is more capable */ most_capable_node = node; } } if (any_capable) { g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { if (can_run_resources(node) == FALSE || node->weight < 0) { continue; } if (have_enough_capacity(node, rscs_id, unallocated_utilization) == FALSE) { pe_rsc_debug(rsc, "Resource %s and its colocated resources" " cannot be allocated to node %s: not enough capacity", rsc->id, node->details->uname); resource_location(rsc, node, -INFINITY, "__limit_utilization__", data_set); } } } else if (*prefer == NULL) { *prefer = most_capable_node; } if (unallocated_utilization) { g_hash_table_destroy(unallocated_utilization); } g_list_free(colocated_rscs); free(rscs_id); } if (any_capable == FALSE) { g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { if (can_run_resources(node) == FALSE || node->weight < 0) { continue; } if (have_enough_capacity(node, rsc->id, rsc->utilization) == FALSE) { pe_rsc_debug(rsc, "Resource %s cannot be allocated to node %s:" " not enough capacity", rsc->id, node->details->uname); resource_location(rsc, node, -INFINITY, "__limit_utilization__", data_set); } } } pe__show_node_weights(true, rsc, "Post-utilization", rsc->allowed_nodes); } } #define VARIANT_GROUP 1 #include GListPtr group_find_colocated_rscs(GListPtr colocated_rscs, pe_resource_t * rsc, pe_resource_t * orig_rsc) { group_variant_data_t *group_data = NULL; get_group_variant_data(group_data, rsc); if (group_data->colocated || pe_rsc_is_clone(rsc->parent)) { GListPtr gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; colocated_rscs = find_colocated_rscs(colocated_rscs, child_rsc, orig_rsc); } } else { if (group_data->first_child) { colocated_rscs = find_colocated_rscs(colocated_rscs, group_data->first_child, orig_rsc); } } colocated_rscs = find_colocated_rscs(colocated_rscs, rsc, orig_rsc); return colocated_rscs; } static void group_add_unallocated_utilization(GHashTable * all_utilization, pe_resource_t * rsc, GListPtr all_rscs) { group_variant_data_t *group_data = NULL; get_group_variant_data(group_data, rsc); if (group_data->colocated || pe_rsc_is_clone(rsc->parent)) { GListPtr gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; if (pcmk_is_set(child_rsc->flags, pe_rsc_provisional) && g_list_find(all_rscs, child_rsc) == FALSE) { native_add_unallocated_utilization(all_utilization, child_rsc); } } } else { if (group_data->first_child && pcmk_is_set(group_data->first_child->flags, pe_rsc_provisional) && g_list_find(all_rscs, group_data->first_child) == FALSE) { native_add_unallocated_utilization(all_utilization, group_data->first_child); } } } diff --git a/lib/pacemaker/pcmk_sched_utils.c b/lib/pacemaker/pcmk_sched_utils.c index eaaf5261a3..aba417a557 100644 --- a/lib/pacemaker/pcmk_sched_utils.c +++ b/lib/pacemaker/pcmk_sched_utils.c @@ -1,767 +1,798 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include // lrmd_event_data_t #include #include pe__location_t * rsc2node_new(const char *id, pe_resource_t *rsc, int node_weight, const char *discover_mode, pe_node_t *foo_node, pe_working_set_t *data_set) { pe__location_t *new_con = NULL; if (rsc == NULL || id == NULL) { pe_err("Invalid constraint %s for rsc=%p", crm_str(id), rsc); return NULL; } else if (foo_node == NULL) { CRM_CHECK(node_weight == 0, return NULL); } new_con = calloc(1, sizeof(pe__location_t)); if (new_con != NULL) { new_con->id = strdup(id); new_con->rsc_lh = rsc; new_con->node_list_rh = NULL; new_con->role_filter = RSC_ROLE_UNKNOWN; if (pcmk__str_eq(discover_mode, "always", pcmk__str_null_matches | pcmk__str_casei)) { new_con->discover_mode = pe_discover_always; } else if (pcmk__str_eq(discover_mode, "never", pcmk__str_casei)) { new_con->discover_mode = pe_discover_never; } else if (pcmk__str_eq(discover_mode, "exclusive", pcmk__str_casei)) { new_con->discover_mode = pe_discover_exclusive; rsc->exclusive_discover = TRUE; } else { pe_err("Invalid %s value %s in location constraint", XML_LOCATION_ATTR_DISCOVERY, discover_mode); } if (foo_node != NULL) { pe_node_t *copy = pe__copy_node(foo_node); copy->weight = node_weight; new_con->node_list_rh = g_list_prepend(NULL, copy); } data_set->placement_constraints = g_list_prepend(data_set->placement_constraints, new_con); rsc->rsc_location = g_list_prepend(rsc->rsc_location, new_con); } return new_con; } gboolean can_run_resources(const pe_node_t * node) { if (node == NULL) { return FALSE; } #if 0 if (node->weight < 0) { return FALSE; } #endif if (node->details->online == FALSE || node->details->shutdown || node->details->unclean || node->details->standby || node->details->maintenance) { crm_trace("%s: online=%d, unclean=%d, standby=%d, maintenance=%d", node->details->uname, node->details->online, node->details->unclean, node->details->standby, node->details->maintenance); return FALSE; } return TRUE; } /*! * \internal * \brief Copy a hash table of node objects * * \param[in] nodes Hash table to copy * * \return New copy of nodes (or NULL if nodes is NULL) */ GHashTable * pcmk__copy_node_table(GHashTable *nodes) { GHashTable *new_table = NULL; GHashTableIter iter; pe_node_t *node = NULL; if (nodes == NULL) { return NULL; } new_table = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, free); g_hash_table_iter_init(&iter, nodes); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { pe_node_t *new_node = pe__copy_node(node); g_hash_table_insert(new_table, (gpointer) new_node->details->id, new_node); } return new_table; } /*! * \internal * \brief Copy a list of node objects * * \param[in] list List to copy * \param[in] reset Set copies' scores to 0 * * \return New list of shallow copies of nodes in original list */ GList * pcmk__copy_node_list(const GList *list, bool reset) { GList *result = NULL; for (const GList *gIter = list; gIter != NULL; gIter = gIter->next) { pe_node_t *new_node = NULL; pe_node_t *this_node = (pe_node_t *) gIter->data; new_node = pe__copy_node(this_node); if (reset) { new_node->weight = 0; } result = g_list_prepend(result, new_node); } return result; } struct node_weight_s { pe_node_t *active; pe_working_set_t *data_set; }; /* return -1 if 'a' is more preferred * return 1 if 'b' is more preferred */ static gint sort_node_weight(gconstpointer a, gconstpointer b, gpointer data) { const pe_node_t *node1 = (const pe_node_t *)a; const pe_node_t *node2 = (const pe_node_t *)b; struct node_weight_s *nw = data; int node1_weight = 0; int node2_weight = 0; int result = 0; if (a == NULL) { return 1; } if (b == NULL) { return -1; } node1_weight = node1->weight; node2_weight = node2->weight; if (can_run_resources(node1) == FALSE) { node1_weight = -INFINITY; } if (can_run_resources(node2) == FALSE) { node2_weight = -INFINITY; } if (node1_weight > node2_weight) { crm_trace("%s (%d) > %s (%d) : weight", node1->details->uname, node1_weight, node2->details->uname, node2_weight); return -1; } if (node1_weight < node2_weight) { crm_trace("%s (%d) < %s (%d) : weight", node1->details->uname, node1_weight, node2->details->uname, node2_weight); return 1; } crm_trace("%s (%d) == %s (%d) : weight", node1->details->uname, node1_weight, node2->details->uname, node2_weight); if (pcmk__str_eq(nw->data_set->placement_strategy, "minimal", pcmk__str_casei)) { goto equal; } if (pcmk__str_eq(nw->data_set->placement_strategy, "balanced", pcmk__str_casei)) { result = compare_capacity(node1, node2); if (result < 0) { crm_trace("%s > %s : capacity (%d)", node1->details->uname, node2->details->uname, result); return -1; } else if (result > 0) { crm_trace("%s < %s : capacity (%d)", node1->details->uname, node2->details->uname, result); return 1; } } /* now try to balance resources across the cluster */ if (node1->details->num_resources < node2->details->num_resources) { crm_trace("%s (%d) > %s (%d) : resources", node1->details->uname, node1->details->num_resources, node2->details->uname, node2->details->num_resources); return -1; } else if (node1->details->num_resources > node2->details->num_resources) { crm_trace("%s (%d) < %s (%d) : resources", node1->details->uname, node1->details->num_resources, node2->details->uname, node2->details->num_resources); return 1; } if (nw->active && nw->active->details == node1->details) { crm_trace("%s (%d) > %s (%d) : active", node1->details->uname, node1->details->num_resources, node2->details->uname, node2->details->num_resources); return -1; } else if (nw->active && nw->active->details == node2->details) { crm_trace("%s (%d) < %s (%d) : active", node1->details->uname, node1->details->num_resources, node2->details->uname, node2->details->num_resources); return 1; } equal: crm_trace("%s = %s", node1->details->uname, node2->details->uname); return strcmp(node1->details->uname, node2->details->uname); } GList * sort_nodes_by_weight(GList *nodes, pe_node_t *active_node, pe_working_set_t *data_set) { struct node_weight_s nw = { active_node, data_set }; return g_list_sort_with_data(nodes, sort_node_weight, &nw); } void native_deallocate(pe_resource_t * rsc) { if (rsc->allocated_to) { pe_node_t *old = rsc->allocated_to; crm_info("Deallocating %s from %s", rsc->id, old->details->uname); pe__set_resource_flags(rsc, pe_rsc_provisional); rsc->allocated_to = NULL; old->details->allocated_rsc = g_list_remove(old->details->allocated_rsc, rsc); old->details->num_resources--; /* old->count--; */ calculate_utilization(old->details->utilization, rsc->utilization, TRUE); free(old); } } gboolean native_assign_node(pe_resource_t * rsc, GListPtr nodes, pe_node_t * chosen, gboolean force) { CRM_ASSERT(rsc->variant == pe_native); if (force == FALSE && chosen != NULL) { bool unset = FALSE; if(chosen->weight < 0) { unset = TRUE; // Allow the graph to assume that the remote resource will come up } else if (!can_run_resources(chosen) && !pe__is_guest_node(chosen)) { unset = TRUE; } if(unset) { crm_debug("All nodes for resource %s are unavailable" ", unclean or shutting down (%s: %d, %d)", rsc->id, chosen->details->uname, can_run_resources(chosen), chosen->weight); rsc->next_role = RSC_ROLE_STOPPED; chosen = NULL; } } /* todo: update the old node for each resource to reflect its * new resource count */ native_deallocate(rsc); pe__clear_resource_flags(rsc, pe_rsc_provisional); if (chosen == NULL) { GListPtr gIter = NULL; char *rc_inactive = crm_itoa(PCMK_OCF_NOT_RUNNING); crm_debug("Could not allocate a node for %s", rsc->id); rsc->next_role = RSC_ROLE_STOPPED; for (gIter = rsc->actions; gIter != NULL; gIter = gIter->next) { pe_action_t *op = (pe_action_t *) gIter->data; const char *interval_ms_s = g_hash_table_lookup(op->meta, XML_LRM_ATTR_INTERVAL_MS); crm_debug("Processing %s", op->uuid); if(pcmk__str_eq(RSC_STOP, op->task, pcmk__str_casei)) { update_action_flags(op, pe_action_optional | pe_action_clear, __func__, __LINE__); } else if(pcmk__str_eq(RSC_START, op->task, pcmk__str_casei)) { update_action_flags(op, pe_action_runnable | pe_action_clear, __func__, __LINE__); //pe__set_resource_flags(rsc, pe_rsc_block); } else if (interval_ms_s && !pcmk__str_eq(interval_ms_s, "0", pcmk__str_casei)) { if(pcmk__str_eq(rc_inactive, g_hash_table_lookup(op->meta, XML_ATTR_TE_TARGET_RC), pcmk__str_casei)) { /* This is a recurring monitor for the stopped state, leave it alone */ } else { /* Normal monitor operation, cancel it */ update_action_flags(op, pe_action_runnable | pe_action_clear, __func__, __LINE__); } } } free(rc_inactive); return FALSE; } crm_debug("Assigning %s to %s", chosen->details->uname, rsc->id); rsc->allocated_to = pe__copy_node(chosen); chosen->details->allocated_rsc = g_list_prepend(chosen->details->allocated_rsc, rsc); chosen->details->num_resources++; chosen->count++; calculate_utilization(chosen->details->utilization, rsc->utilization, FALSE); dump_rsc_utilization((show_utilization? LOG_STDOUT : LOG_TRACE), __func__, rsc, chosen); return TRUE; } void log_action(unsigned int log_level, const char *pre_text, pe_action_t * action, gboolean details) { const char *node_uname = NULL; const char *node_uuid = NULL; const char *desc = NULL; if (action == NULL) { crm_trace("%s%s: ", pre_text == NULL ? "" : pre_text, pre_text == NULL ? "" : ": "); return; } if (pcmk_is_set(action->flags, pe_action_pseudo)) { node_uname = NULL; node_uuid = NULL; } else if (action->node != NULL) { node_uname = action->node->details->uname; node_uuid = action->node->details->id; } else { node_uname = ""; node_uuid = NULL; } switch (text2task(action->task)) { case stonith_node: case shutdown_crm: if (pcmk_is_set(action->flags, pe_action_pseudo)) { desc = "Pseudo "; } else if (pcmk_is_set(action->flags, pe_action_optional)) { desc = "Optional "; } else if (!pcmk_is_set(action->flags, pe_action_runnable)) { desc = "!!Non-Startable!! "; } else if (pcmk_is_set(action->flags, pe_action_processed)) { desc = ""; } else { desc = "(Provisional) "; } crm_trace("%s%s%sAction %d: %s%s%s%s%s%s", ((pre_text == NULL)? "" : pre_text), ((pre_text == NULL)? "" : ": "), desc, action->id, action->uuid, (node_uname? "\ton " : ""), (node_uname? node_uname : ""), (node_uuid? "\t\t(" : ""), (node_uuid? node_uuid : ""), (node_uuid? ")" : "")); break; default: if (pcmk_is_set(action->flags, pe_action_optional)) { desc = "Optional "; } else if (pcmk_is_set(action->flags, pe_action_pseudo)) { desc = "Pseudo "; } else if (!pcmk_is_set(action->flags, pe_action_runnable)) { desc = "!!Non-Startable!! "; } else if (pcmk_is_set(action->flags, pe_action_processed)) { desc = ""; } else { desc = "(Provisional) "; } crm_trace("%s%s%sAction %d: %s %s%s%s%s%s%s", ((pre_text == NULL)? "" : pre_text), ((pre_text == NULL)? "" : ": "), desc, action->id, action->uuid, (action->rsc? action->rsc->id : ""), (node_uname? "\ton " : ""), (node_uname? node_uname : ""), (node_uuid? "\t\t(" : ""), (node_uuid? node_uuid : ""), (node_uuid? ")" : "")); break; } if (details) { GListPtr gIter = NULL; crm_trace("\t\t====== Preceding Actions"); gIter = action->actions_before; for (; gIter != NULL; gIter = gIter->next) { pe_action_wrapper_t *other = (pe_action_wrapper_t *) gIter->data; log_action(log_level + 1, "\t\t", other->action, FALSE); } crm_trace("\t\t====== Subsequent Actions"); gIter = action->actions_after; for (; gIter != NULL; gIter = gIter->next) { pe_action_wrapper_t *other = (pe_action_wrapper_t *) gIter->data; log_action(log_level + 1, "\t\t", other->action, FALSE); } crm_trace("\t\t====== End"); } else { crm_trace("\t\t(before=%d, after=%d)", g_list_length(action->actions_before), g_list_length(action->actions_after)); } } gboolean can_run_any(GHashTable * nodes) { GHashTableIter iter; pe_node_t *node = NULL; if (nodes == NULL) { return FALSE; } g_hash_table_iter_init(&iter, nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { if (can_run_resources(node) && node->weight >= 0) { return TRUE; } } return FALSE; } pe_action_t * create_pseudo_resource_op(pe_resource_t * rsc, const char *task, bool optional, bool runnable, pe_working_set_t *data_set) { pe_action_t *action = custom_action(rsc, pcmk__op_key(rsc->id, task, 0), task, NULL, optional, TRUE, data_set); update_action_flags(action, pe_action_pseudo, __func__, __LINE__); update_action_flags(action, pe_action_runnable, __func__, __LINE__); if(runnable) { update_action_flags(action, pe_action_runnable, __func__, __LINE__); } return action; } /*! * \internal * \brief Create an executor cancel op * * \param[in] rsc Resource of action to cancel * \param[in] task Name of action to cancel * \param[in] interval_ms Interval of action to cancel * \param[in] node Node of action to cancel * \param[in] data_set Working set of cluster * * \return Created op */ pe_action_t * pe_cancel_op(pe_resource_t *rsc, const char *task, guint interval_ms, pe_node_t *node, pe_working_set_t *data_set) { pe_action_t *cancel_op; char *interval_ms_s = crm_strdup_printf("%u", interval_ms); // @TODO dangerous if possible to schedule another action with this key char *key = pcmk__op_key(rsc->id, task, interval_ms); cancel_op = custom_action(rsc, key, RSC_CANCEL, node, FALSE, TRUE, data_set); free(cancel_op->task); cancel_op->task = strdup(RSC_CANCEL); free(cancel_op->cancel_task); cancel_op->cancel_task = strdup(task); add_hash_param(cancel_op->meta, XML_LRM_ATTR_TASK, task); add_hash_param(cancel_op->meta, XML_LRM_ATTR_INTERVAL_MS, interval_ms_s); free(interval_ms_s); return cancel_op; } /*! * \internal * \brief Create a shutdown op for a scheduler transition * * \param[in] node Node being shut down * \param[in] data_set Working set of cluster * * \return Created op */ pe_action_t * sched_shutdown_op(pe_node_t *node, pe_working_set_t *data_set) { char *shutdown_id = crm_strdup_printf("%s-%s", CRM_OP_SHUTDOWN, node->details->uname); pe_action_t *shutdown_op = custom_action(NULL, shutdown_id, CRM_OP_SHUTDOWN, node, FALSE, TRUE, data_set); crm_notice("Scheduling shutdown of node %s", node->details->uname); shutdown_constraints(node, shutdown_op, data_set); add_hash_param(shutdown_op->meta, XML_ATTR_TE_NOWAIT, XML_BOOLEAN_TRUE); return shutdown_op; } static char * generate_transition_magic(const char *transition_key, int op_status, int op_rc) { CRM_CHECK(transition_key != NULL, return NULL); return crm_strdup_printf("%d:%d;%s", op_status, op_rc, transition_key); } static void append_digest(lrmd_event_data_t *op, xmlNode *update, const char *version, const char *magic, int level) { /* this will enable us to later determine that the * resource's parameters have changed and we should force * a restart */ char *digest = NULL; xmlNode *args_xml = NULL; if (op->params == NULL) { return; } args_xml = create_xml_node(NULL, XML_TAG_PARAMS); g_hash_table_foreach(op->params, hash2field, args_xml); pcmk__filter_op_for_digest(args_xml); digest = calculate_operation_digest(args_xml, version); #if 0 if (level < get_crm_log_level() && op->interval_ms == 0 && pcmk__str_eq(op->op_type, CRMD_ACTION_START, pcmk__str_none)) { char *digest_source = dump_xml_unformatted(args_xml); do_crm_log(level, "Calculated digest %s for %s (%s). Source: %s\n", digest, ID(update), magic, digest_source); free(digest_source); } #endif crm_xml_add(update, XML_LRM_ATTR_OP_DIGEST, digest); free_xml(args_xml); free(digest); } #define FAKE_TE_ID "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx" /*! * \internal * \brief Create XML for resource operation history update * * \param[in,out] parent Parent XML node to add to * \param[in,out] op Operation event data * \param[in] caller_version DC feature set * \param[in] target_rc Expected result of operation * \param[in] node Name of node on which operation was performed * \param[in] origin Arbitrary description of update source * \param[in] level A log message will be logged at this level * * \return Newly created XML node for history update */ xmlNode * pcmk__create_history_xml(xmlNode *parent, lrmd_event_data_t *op, const char *caller_version, int target_rc, const char *node, const char *origin, int level) { char *key = NULL; char *magic = NULL; char *op_id = NULL; char *op_id_additional = NULL; char *local_user_data = NULL; const char *exit_reason = NULL; xmlNode *xml_op = NULL; const char *task = NULL; CRM_CHECK(op != NULL, return NULL); do_crm_log(level, "%s: Updating resource %s after %s op %s (interval=%u)", origin, op->rsc_id, op->op_type, services_lrm_status_str(op->op_status), op->interval_ms); crm_trace("DC version: %s", caller_version); task = op->op_type; /* Record a successful reload as a start, and a failed reload as a monitor, * to make life easier for the scheduler when determining the current state. */ if (pcmk__str_eq(task, "reload", pcmk__str_none)) { if (op->op_status == PCMK_LRM_OP_DONE) { task = CRMD_ACTION_START; } else { task = CRMD_ACTION_STATUS; } } key = pcmk__op_key(op->rsc_id, task, op->interval_ms); if (pcmk__str_eq(task, CRMD_ACTION_NOTIFY, pcmk__str_none)) { const char *n_type = crm_meta_value(op->params, "notify_type"); const char *n_task = crm_meta_value(op->params, "notify_operation"); CRM_LOG_ASSERT(n_type != NULL); CRM_LOG_ASSERT(n_task != NULL); op_id = pcmk__notify_key(op->rsc_id, n_type, n_task); if (op->op_status != PCMK_LRM_OP_PENDING) { /* Ignore notify errors. * * @TODO It might be better to keep the correct result here, and * ignore it in process_graph_event(). */ op->op_status = PCMK_LRM_OP_DONE; op->rc = 0; } } else if (did_rsc_op_fail(op, target_rc)) { op_id = pcmk__op_key(op->rsc_id, "last_failure", 0); if (op->interval_ms == 0) { // Ensure 'last' gets updated, in case record-pending is true op_id_additional = pcmk__op_key(op->rsc_id, "last", 0); } exit_reason = op->exit_reason; } else if (op->interval_ms > 0) { op_id = strdup(key); } else { op_id = pcmk__op_key(op->rsc_id, "last", 0); } again: xml_op = pcmk__xe_match(parent, XML_LRM_TAG_RSC_OP, XML_ATTR_ID, op_id); if (xml_op == NULL) { xml_op = create_xml_node(parent, XML_LRM_TAG_RSC_OP); } if (op->user_data == NULL) { crm_debug("Generating fake transition key for: " PCMK__OP_FMT " %d from %s", op->rsc_id, op->op_type, op->interval_ms, op->call_id, origin); local_user_data = pcmk__transition_key(-1, op->call_id, target_rc, FAKE_TE_ID); op->user_data = local_user_data; } if(magic == NULL) { magic = generate_transition_magic(op->user_data, op->op_status, op->rc); } crm_xml_add(xml_op, XML_ATTR_ID, op_id); crm_xml_add(xml_op, XML_LRM_ATTR_TASK_KEY, key); crm_xml_add(xml_op, XML_LRM_ATTR_TASK, task); crm_xml_add(xml_op, XML_ATTR_ORIGIN, origin); crm_xml_add(xml_op, XML_ATTR_CRM_VERSION, caller_version); crm_xml_add(xml_op, XML_ATTR_TRANSITION_KEY, op->user_data); crm_xml_add(xml_op, XML_ATTR_TRANSITION_MAGIC, magic); crm_xml_add(xml_op, XML_LRM_ATTR_EXIT_REASON, exit_reason == NULL ? "" : exit_reason); crm_xml_add(xml_op, XML_LRM_ATTR_TARGET, node); /* For context during triage */ crm_xml_add_int(xml_op, XML_LRM_ATTR_CALLID, op->call_id); crm_xml_add_int(xml_op, XML_LRM_ATTR_RC, op->rc); crm_xml_add_int(xml_op, XML_LRM_ATTR_OPSTATUS, op->op_status); crm_xml_add_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, op->interval_ms); if (compare_version("2.1", caller_version) <= 0) { if (op->t_run || op->t_rcchange || op->exec_time || op->queue_time) { crm_trace("Timing data (" PCMK__OP_FMT "): last=%u change=%u exec=%u queue=%u", op->rsc_id, op->op_type, op->interval_ms, op->t_run, op->t_rcchange, op->exec_time, op->queue_time); if (op->interval_ms == 0) { crm_xml_add_ll(xml_op, XML_RSC_OP_LAST_CHANGE, (long long) op->t_run); // @COMPAT last-run is deprecated crm_xml_add_ll(xml_op, XML_RSC_OP_LAST_RUN, (long long) op->t_run); } else if(op->t_rcchange) { /* last-run is not accurate for recurring ops */ crm_xml_add_ll(xml_op, XML_RSC_OP_LAST_CHANGE, (long long) op->t_rcchange); } else { /* ...but is better than nothing otherwise */ crm_xml_add_ll(xml_op, XML_RSC_OP_LAST_CHANGE, (long long) op->t_run); } crm_xml_add_int(xml_op, XML_RSC_OP_T_EXEC, op->exec_time); crm_xml_add_int(xml_op, XML_RSC_OP_T_QUEUE, op->queue_time); } } if (pcmk__str_any_of(op->op_type, CRMD_ACTION_MIGRATE, CRMD_ACTION_MIGRATED, NULL)) { /* * Record migrate_source and migrate_target always for migrate ops. */ const char *name = XML_LRM_ATTR_MIGRATE_SOURCE; crm_xml_add(xml_op, name, crm_meta_value(op->params, name)); name = XML_LRM_ATTR_MIGRATE_TARGET; crm_xml_add(xml_op, name, crm_meta_value(op->params, name)); } append_digest(op, xml_op, caller_version, magic, LOG_DEBUG); if (op_id_additional) { free(op_id); op_id = op_id_additional; op_id_additional = NULL; goto again; } if (local_user_data) { free(local_user_data); op->user_data = NULL; } free(magic); free(op_id); free(key); return xml_op; } + +/*! + * \internal + * \brief Check whether a colocation constraint should apply + * + * \param[in] rsc Resource of interest (for logging) + * \param[in] colocation Colocation constraint to check + * \param[in] promoted_only If true, constraint applies if right-hand is promoted + */ +bool +pcmk__colocation_applies(pe_resource_t *rsc, pcmk__colocation_t *colocation, + bool promoted_only) +{ + CRM_CHECK((rsc != NULL) && (colocation != NULL), return false); + + if (colocation->score == 0) { + pe_rsc_trace(rsc, "Ignoring colocation constraint %s: 0 score", + colocation->id); + return false; + } + if (promoted_only && (colocation->role_rh != RSC_ROLE_MASTER)) { + pe_rsc_trace(rsc, "Ignoring colocation constraint %s: role", + colocation->id); + return false; + } + pe_rsc_trace(rsc, "Applying colocation constraint %s: %s with %s%s (%d)", + colocation->id, colocation->rsc_lh->id, + (promoted_only? "promoted " : ""), + colocation->rsc_rh->id, colocation->score); + return true; +} diff --git a/tools/crm_resource_runtime.c b/tools/crm_resource_runtime.c index f4500db9f7..e0804fc4b6 100644 --- a/tools/crm_resource_runtime.c +++ b/tools/crm_resource_runtime.c @@ -1,1955 +1,1955 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include resource_checks_t * cli_check_resource(pe_resource_t *rsc, char *role_s, char *managed) { pe_resource_t *parent = uber_parent(rsc); resource_checks_t *rc = calloc(1, sizeof(resource_checks_t)); if (role_s) { enum rsc_role_e role = text2role(role_s); if (role == RSC_ROLE_STOPPED) { rc->flags |= rsc_remain_stopped; } else if (pcmk_is_set(parent->flags, pe_rsc_promotable) && role == RSC_ROLE_SLAVE) { rc->flags |= rsc_unpromotable; } } if (managed && !crm_is_true(managed)) { rc->flags |= rsc_unmanaged; } if (rsc->lock_node) { rc->lock_node = rsc->lock_node->details->uname; } rc->rsc = rsc; return rc; } GListPtr cli_resource_search(pcmk__output_t *out, pe_resource_t *rsc, const char *requested_name, pe_working_set_t *data_set) { GListPtr found = NULL; pe_resource_t *parent = uber_parent(rsc); if (pe_rsc_is_clone(rsc)) { for (GListPtr iter = rsc->children; iter != NULL; iter = iter->next) { GListPtr extra = ((pe_resource_t *) iter->data)->running_on; if (extra != NULL) { found = g_list_concat(found, extra); } } /* The anonymous clone children's common ID is supplied */ } else if (pe_rsc_is_clone(parent) && !pcmk_is_set(rsc->flags, pe_rsc_unique) && rsc->clone_name && pcmk__str_eq(requested_name, rsc->clone_name, pcmk__str_casei) && !pcmk__str_eq(requested_name, rsc->id, pcmk__str_casei)) { for (GListPtr iter = parent->children; iter; iter = iter->next) { GListPtr extra = ((pe_resource_t *) iter->data)->running_on; if (extra != NULL) { found = g_list_concat(found, extra); } } } else if (rsc->running_on != NULL) { found = g_list_concat(found, rsc->running_on); } return found; } #define XPATH_MAX 1024 // \return Standard Pacemaker return code static int find_resource_attr(pcmk__output_t *out, cib_t * the_cib, const char *attr, const char *rsc, const char *attr_set_type, const char *set_name, const char *attr_id, const char *attr_name, char **value) { int offset = 0; int rc = pcmk_rc_ok; xmlNode *xml_search = NULL; char *xpath_string = NULL; if(value) { *value = NULL; } if(the_cib == NULL) { return ENOTCONN; } xpath_string = calloc(1, XPATH_MAX); offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "%s", get_object_path("resources")); offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "//*[@id=\"%s\"]", rsc); if (attr_set_type) { offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "/%s", attr_set_type); if (set_name) { offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "[@id=\"%s\"]", set_name); } } offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "//nvpair["); if (attr_id) { offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "@id=\"%s\"", attr_id); } if (attr_name) { if (attr_id) { offset += snprintf(xpath_string + offset, XPATH_MAX - offset, " and "); } offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "@name=\"%s\"", attr_name); } offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "]"); CRM_LOG_ASSERT(offset > 0); rc = the_cib->cmds->query(the_cib, xpath_string, &xml_search, cib_sync_call | cib_scope_local | cib_xpath); rc = pcmk_legacy2rc(rc); if (rc != pcmk_rc_ok) { goto bail; } crm_log_xml_debug(xml_search, "Match"); if (xml_has_children(xml_search)) { xmlNode *child = NULL; rc = EINVAL; out->info(out, "Multiple attributes match name=%s", attr_name); for (child = pcmk__xml_first_child(xml_search); child != NULL; child = pcmk__xml_next(child)) { out->info(out, " Value: %s \t(id=%s)", crm_element_value(child, XML_NVPAIR_ATTR_VALUE), ID(child)); } out->spacer(out); } else if(value) { const char *tmp = crm_element_value(xml_search, attr); if (tmp) { *value = strdup(tmp); } } bail: free(xpath_string); free_xml(xml_search); return rc; } /* PRIVATE. Use the find_matching_attr_resources instead. */ static void find_matching_attr_resources_recursive(pcmk__output_t *out, GList/* */ ** result, pe_resource_t * rsc, const char * rsc_id, const char * attr_set, const char * attr_set_type, const char * attr_id, const char * attr_name, cib_t * cib, const char * cmd, int depth) { int rc = pcmk_rc_ok; char *lookup_id = clone_strip(rsc->id); char *local_attr_id = NULL; /* visit the children */ for(GList *gIter = rsc->children; gIter; gIter = gIter->next) { find_matching_attr_resources_recursive(out, result, (pe_resource_t*)gIter->data, rsc_id, attr_set, attr_set_type, attr_id, attr_name, cib, cmd, depth+1); /* do it only once for clones */ if(pe_clone == rsc->variant) { break; } } rc = find_resource_attr(out, cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); /* Post-order traversal. * The root is always on the list and it is the last item. */ if((0 == depth) || (pcmk_rc_ok == rc)) { /* push the head */ *result = g_list_append(*result, rsc); } free(local_attr_id); free(lookup_id); } /* The result is a linearized pre-ordered tree of resources. */ static GList/**/ * find_matching_attr_resources(pcmk__output_t *out, pe_resource_t * rsc, const char * rsc_id, const char * attr_set, const char * attr_set_type, const char * attr_id, const char * attr_name, cib_t * cib, const char * cmd, gboolean force) { int rc = pcmk_rc_ok; char *lookup_id = NULL; char *local_attr_id = NULL; GList * result = NULL; /* If --force is used, update only the requested resource (clone or primitive). * Otherwise, if the primitive has the attribute, use that. * Otherwise use the clone. */ if(force == TRUE) { return g_list_append(result, rsc); } if(rsc->parent && pe_clone == rsc->parent->variant) { int rc = pcmk_rc_ok; char *local_attr_id = NULL; rc = find_resource_attr(out, cib, XML_ATTR_ID, rsc_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); free(local_attr_id); if(rc != pcmk_rc_ok) { rsc = rsc->parent; if (!out->is_quiet(out)) { out->info(out, "Performing %s of '%s' on '%s', the parent of '%s'", cmd, attr_name, rsc->id, rsc_id); } } return g_list_append(result, rsc); } else if(rsc->parent == NULL && rsc->children && pe_clone == rsc->variant) { pe_resource_t *child = rsc->children->data; if(child->variant == pe_native) { lookup_id = clone_strip(child->id); /* Could be a cloned group! */ rc = find_resource_attr(out, cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); if(rc == pcmk_rc_ok) { rsc = child; if (!out->is_quiet(out)) { out->info(out, "A value for '%s' already exists in child '%s', performing %s on that instead of '%s'", attr_name, lookup_id, cmd, rsc_id); } } free(local_attr_id); free(lookup_id); } return g_list_append(result, rsc); } /* If the resource is a group ==> children inherit the attribute if defined. */ find_matching_attr_resources_recursive(out, &result, rsc, rsc_id, attr_set, attr_set_type, attr_id, attr_name, cib, cmd, 0); return result; } // \return Standard Pacemaker return code int cli_resource_update_attribute(pcmk__output_t *out, pe_resource_t *rsc, const char *requested_name, const char *attr_set, const char *attr_set_type, const char *attr_id, const char *attr_name, const char *attr_value, gboolean recursive, cib_t *cib, int cib_options, pe_working_set_t *data_set, gboolean force) { int rc = pcmk_rc_ok; static bool need_init = TRUE; char *local_attr_id = NULL; char *local_attr_set = NULL; GList/**/ *resources = NULL; const char *common_attr_id = attr_id; if (attr_id == NULL && force == FALSE) { find_resource_attr (out, cib, XML_ATTR_ID, uber_parent(rsc)->id, NULL, NULL, NULL, attr_name, NULL); } if (pcmk__str_eq(attr_set_type, XML_TAG_ATTR_SETS, pcmk__str_casei)) { if (force == FALSE) { rc = find_resource_attr(out, cib, XML_ATTR_ID, uber_parent(rsc)->id, XML_TAG_META_SETS, attr_set, attr_id, attr_name, &local_attr_id); if (rc == pcmk_rc_ok && !out->is_quiet(out)) { out->err(out, "WARNING: There is already a meta attribute for '%s' called '%s' (id=%s)", uber_parent(rsc)->id, attr_name, local_attr_id); out->err(out, " Delete '%s' first or use the force option to override", local_attr_id); } free(local_attr_id); if (rc == pcmk_rc_ok) { return ENOTUNIQ; } } resources = g_list_append(resources, rsc); } else { resources = find_matching_attr_resources(out, rsc, requested_name, attr_set, attr_set_type, attr_id, attr_name, cib, "update", force); } /* If either attr_set or attr_id is specified, * one clearly intends to modify a single resource. * It is the last item on the resource list.*/ for(GList *gIter = (attr_set||attr_id) ? g_list_last(resources) : resources ; gIter; gIter = gIter->next) { char *lookup_id = NULL; xmlNode *xml_top = NULL; xmlNode *xml_obj = NULL; local_attr_id = NULL; local_attr_set = NULL; rsc = (pe_resource_t*)gIter->data; attr_id = common_attr_id; lookup_id = clone_strip(rsc->id); /* Could be a cloned group! */ rc = find_resource_attr(out, cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); if (rc == pcmk_rc_ok) { crm_debug("Found a match for name=%s: id=%s", attr_name, local_attr_id); attr_id = local_attr_id; } else if (rc != ENXIO) { free(lookup_id); free(local_attr_id); g_list_free(resources); return rc; } else { const char *tag = crm_element_name(rsc->xml); if (attr_set == NULL) { local_attr_set = crm_strdup_printf("%s-%s", lookup_id, attr_set_type); attr_set = local_attr_set; } if (attr_id == NULL) { local_attr_id = crm_strdup_printf("%s-%s", attr_set, attr_name); attr_id = local_attr_id; } xml_top = create_xml_node(NULL, tag); crm_xml_add(xml_top, XML_ATTR_ID, lookup_id); xml_obj = create_xml_node(xml_top, attr_set_type); crm_xml_add(xml_obj, XML_ATTR_ID, attr_set); } xml_obj = crm_create_nvpair_xml(xml_obj, attr_id, attr_name, attr_value); if (xml_top == NULL) { xml_top = xml_obj; } crm_log_xml_debug(xml_top, "Update"); rc = cib->cmds->modify(cib, XML_CIB_TAG_RESOURCES, xml_top, cib_options); rc = pcmk_legacy2rc(rc); if (rc == pcmk_rc_ok && !out->is_quiet(out)) { out->info(out, "Set '%s' option: id=%s%s%s%s%s value=%s", lookup_id, local_attr_id, attr_set ? " set=" : "", attr_set ? attr_set : "", attr_name ? " name=" : "", attr_name ? attr_name : "", attr_value); } free_xml(xml_top); free(lookup_id); free(local_attr_id); free(local_attr_set); if(recursive && pcmk__str_eq(attr_set_type, XML_TAG_META_SETS, pcmk__str_casei)) { GListPtr lpc = NULL; if(need_init) { xmlNode *cib_constraints = get_object_root(XML_CIB_TAG_CONSTRAINTS, data_set->input); need_init = FALSE; unpack_constraints(cib_constraints, data_set); pe__clear_resource_flags_on_all(data_set, pe_rsc_allocating); } crm_debug("Looking for dependencies %p", rsc->rsc_cons_lhs); pe__set_resource_flags(rsc, pe_rsc_allocating); for (lpc = rsc->rsc_cons_lhs; lpc != NULL; lpc = lpc->next) { - rsc_colocation_t *cons = (rsc_colocation_t *) lpc->data; + pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data; pe_resource_t *peer = cons->rsc_lh; crm_debug("Checking %s %d", cons->id, cons->score); if (cons->score > 0 && !pcmk_is_set(peer->flags, pe_rsc_allocating)) { /* Don't get into colocation loops */ crm_debug("Setting %s=%s for dependent resource %s", attr_name, attr_value, peer->id); cli_resource_update_attribute(out, peer, peer->id, NULL, attr_set_type, NULL, attr_name, attr_value, recursive, cib, cib_options, data_set, force); } } } } g_list_free(resources); return rc; } // \return Standard Pacemaker return code int cli_resource_delete_attribute(pcmk__output_t *out, pe_resource_t *rsc, const char *requested_name, const char *attr_set, const char *attr_set_type, const char *attr_id, const char *attr_name, cib_t *cib, int cib_options, pe_working_set_t *data_set, gboolean force) { int rc = pcmk_rc_ok; GList/**/ *resources = NULL; if (attr_id == NULL && force == FALSE) { find_resource_attr(out, cib, XML_ATTR_ID, uber_parent(rsc)->id, NULL, NULL, NULL, attr_name, NULL); } if(pcmk__str_eq(attr_set_type, XML_TAG_META_SETS, pcmk__str_casei)) { resources = find_matching_attr_resources(out, rsc, requested_name, attr_set, attr_set_type, attr_id, attr_name, cib, "delete", force); } else { resources = g_list_append(resources, rsc); } for(GList *gIter = resources; gIter; gIter = gIter->next) { char *lookup_id = NULL; xmlNode *xml_obj = NULL; char *local_attr_id = NULL; rsc = (pe_resource_t*)gIter->data; lookup_id = clone_strip(rsc->id); rc = find_resource_attr(out, cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); if (rc == ENXIO) { free(lookup_id); rc = pcmk_rc_ok; continue; } else if (rc != pcmk_rc_ok) { free(lookup_id); g_list_free(resources); return rc; } if (attr_id == NULL) { attr_id = local_attr_id; } xml_obj = crm_create_nvpair_xml(NULL, attr_id, attr_name, NULL); crm_log_xml_debug(xml_obj, "Delete"); CRM_ASSERT(cib); rc = cib->cmds->remove(cib, XML_CIB_TAG_RESOURCES, xml_obj, cib_options); rc = pcmk_legacy2rc(rc); if (rc == pcmk_rc_ok && !out->is_quiet(out)) { out->info(out, "Deleted '%s' option: id=%s%s%s%s%s", lookup_id, local_attr_id, attr_set ? " set=" : "", attr_set ? attr_set : "", attr_name ? " name=" : "", attr_name ? attr_name : ""); } free(lookup_id); free_xml(xml_obj); free(local_attr_id); } g_list_free(resources); return rc; } // \return Standard Pacemaker return code static int send_lrm_rsc_op(pcmk__output_t *out, pcmk_ipc_api_t *controld_api, bool do_fail_resource, const char *host_uname, const char *rsc_id, pe_working_set_t *data_set) { const char *router_node = host_uname; const char *rsc_api_id = NULL; const char *rsc_long_id = NULL; const char *rsc_class = NULL; const char *rsc_provider = NULL; const char *rsc_type = NULL; bool cib_only = false; pe_resource_t *rsc = pe_find_resource(data_set->resources, rsc_id); if (rsc == NULL) { out->err(out, "Resource %s not found", rsc_id); return ENXIO; } else if (rsc->variant != pe_native) { out->err(out, "We can only process primitive resources, not %s", rsc_id); return EINVAL; } rsc_class = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS); rsc_provider = crm_element_value(rsc->xml, XML_AGENT_ATTR_PROVIDER), rsc_type = crm_element_value(rsc->xml, XML_ATTR_TYPE); if ((rsc_class == NULL) || (rsc_type == NULL)) { out->err(out, "Resource %s does not have a class and type", rsc_id); return EINVAL; } if (host_uname == NULL) { out->err(out, "Please specify a node name"); return EINVAL; } else { pe_node_t *node = pe_find_node(data_set->nodes, host_uname); if (node == NULL) { out->err(out, "Node %s not found", host_uname); return pcmk_rc_node_unknown; } if (!(node->details->online)) { if (do_fail_resource) { out->err(out, "Node %s is not online", host_uname); return ENOTCONN; } else { cib_only = true; } } if (!cib_only && pe__is_guest_or_remote_node(node)) { node = pe__current_node(node->details->remote_rsc); if (node == NULL) { out->err(out, "No cluster connection to Pacemaker Remote node %s detected", host_uname); return ENOTCONN; } router_node = node->details->uname; } } if (rsc->clone_name) { rsc_api_id = rsc->clone_name; rsc_long_id = rsc->id; } else { rsc_api_id = rsc->id; } if (do_fail_resource) { return pcmk_controld_api_fail(controld_api, host_uname, router_node, rsc_api_id, rsc_long_id, rsc_class, rsc_provider, rsc_type); } else { return pcmk_controld_api_refresh(controld_api, host_uname, router_node, rsc_api_id, rsc_long_id, rsc_class, rsc_provider, rsc_type, cib_only); } } /*! * \internal * \brief Get resource name as used in failure-related node attributes * * \param[in] rsc Resource to check * * \return Newly allocated string containing resource's fail name * \note The caller is responsible for freeing the result. */ static inline char * rsc_fail_name(pe_resource_t *rsc) { const char *name = (rsc->clone_name? rsc->clone_name : rsc->id); return pcmk_is_set(rsc->flags, pe_rsc_unique)? strdup(name) : clone_strip(name); } // \return Standard Pacemaker return code static int clear_rsc_history(pcmk__output_t *out, pcmk_ipc_api_t *controld_api, const char *host_uname, const char *rsc_id, pe_working_set_t *data_set) { int rc = pcmk_rc_ok; /* Erase the resource's entire LRM history in the CIB, even if we're only * clearing a single operation's fail count. If we erased only entries for a * single operation, we might wind up with a wrong idea of the current * resource state, and we might not re-probe the resource. */ rc = send_lrm_rsc_op(out, controld_api, false, host_uname, rsc_id, data_set); if (rc != pcmk_rc_ok) { return rc; } crm_trace("Processing %d mainloop inputs", pcmk_controld_api_replies_expected(controld_api)); while (g_main_context_iteration(NULL, FALSE)) { crm_trace("Processed mainloop input, %d still remaining", pcmk_controld_api_replies_expected(controld_api)); } return rc; } // \return Standard Pacemaker return code static int clear_rsc_failures(pcmk__output_t *out, pcmk_ipc_api_t *controld_api, const char *node_name, const char *rsc_id, const char *operation, const char *interval_spec, pe_working_set_t *data_set) { int rc = pcmk_rc_ok; const char *failed_value = NULL; const char *failed_id = NULL; const char *interval_ms_s = NULL; GHashTable *rscs = NULL; GHashTableIter iter; /* Create a hash table to use as a set of resources to clean. This lets us * clean each resource only once (per node) regardless of how many failed * operations it has. */ rscs = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, NULL); // Normalize interval to milliseconds for comparison to history entry if (operation) { interval_ms_s = crm_strdup_printf("%u", crm_parse_interval_spec(interval_spec)); } for (xmlNode *xml_op = pcmk__xml_first_child(data_set->failed); xml_op != NULL; xml_op = pcmk__xml_next(xml_op)) { failed_id = crm_element_value(xml_op, XML_LRM_ATTR_RSCID); if (failed_id == NULL) { // Malformed history entry, should never happen continue; } // No resource specified means all resources match if (rsc_id) { pe_resource_t *fail_rsc = pe_find_resource_with_flags(data_set->resources, failed_id, pe_find_renamed|pe_find_anon); if (!fail_rsc || !pcmk__str_eq(rsc_id, fail_rsc->id, pcmk__str_casei)) { continue; } } // Host name should always have been provided by this point failed_value = crm_element_value(xml_op, XML_ATTR_UNAME); if (!pcmk__str_eq(node_name, failed_value, pcmk__str_casei)) { continue; } // No operation specified means all operations match if (operation) { failed_value = crm_element_value(xml_op, XML_LRM_ATTR_TASK); if (!pcmk__str_eq(operation, failed_value, pcmk__str_casei)) { continue; } // Interval (if operation was specified) defaults to 0 (not all) failed_value = crm_element_value(xml_op, XML_LRM_ATTR_INTERVAL_MS); if (!pcmk__str_eq(interval_ms_s, failed_value, pcmk__str_casei)) { continue; } } /* not available until glib 2.32 g_hash_table_add(rscs, (gpointer) failed_id); */ g_hash_table_insert(rscs, (gpointer) failed_id, (gpointer) failed_id); } g_hash_table_iter_init(&iter, rscs); while (g_hash_table_iter_next(&iter, (gpointer *) &failed_id, NULL)) { crm_debug("Erasing failures of %s on %s", failed_id, node_name); rc = clear_rsc_history(out, controld_api, node_name, failed_id, data_set); if (rc != pcmk_rc_ok) { return rc; } } g_hash_table_destroy(rscs); return rc; } // \return Standard Pacemaker return code static int clear_rsc_fail_attrs(pe_resource_t *rsc, const char *operation, const char *interval_spec, pe_node_t *node) { int rc = pcmk_rc_ok; int attr_options = pcmk__node_attr_none; char *rsc_name = rsc_fail_name(rsc); if (pe__is_guest_or_remote_node(node)) { attr_options |= pcmk__node_attr_remote; } rc = pcmk__node_attr_request_clear(NULL, node->details->uname, rsc_name, operation, interval_spec, NULL, attr_options); free(rsc_name); return rc; } // \return Standard Pacemaker return code int cli_resource_delete(pcmk__output_t *out, pcmk_ipc_api_t *controld_api, const char *host_uname, pe_resource_t *rsc, const char *operation, const char *interval_spec, bool just_failures, pe_working_set_t *data_set, gboolean force) { int rc = pcmk_rc_ok; pe_node_t *node = NULL; if (rsc == NULL) { return ENXIO; } else if (rsc->children) { GListPtr lpc = NULL; for (lpc = rsc->children; lpc != NULL; lpc = lpc->next) { pe_resource_t *child = (pe_resource_t *) lpc->data; rc = cli_resource_delete(out, controld_api, host_uname, child, operation, interval_spec, just_failures, data_set, force); if (rc != pcmk_rc_ok) { return rc; } } return pcmk_rc_ok; } else if (host_uname == NULL) { GListPtr lpc = NULL; GListPtr nodes = g_hash_table_get_values(rsc->known_on); if(nodes == NULL && force) { nodes = pcmk__copy_node_list(data_set->nodes, false); } else if(nodes == NULL && rsc->exclusive_discover) { GHashTableIter iter; pe_node_t *node = NULL; g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void**)&node)) { if(node->weight >= 0) { nodes = g_list_prepend(nodes, node); } } } else if(nodes == NULL) { nodes = g_hash_table_get_values(rsc->allowed_nodes); } for (lpc = nodes; lpc != NULL; lpc = lpc->next) { node = (pe_node_t *) lpc->data; if (node->details->online) { rc = cli_resource_delete(out, controld_api, node->details->uname, rsc, operation, interval_spec, just_failures, data_set, force); } if (rc != pcmk_rc_ok) { g_list_free(nodes); return rc; } } g_list_free(nodes); return pcmk_rc_ok; } node = pe_find_node(data_set->nodes, host_uname); if (node == NULL) { out->err(out, "Unable to clean up %s because node %s not found", rsc->id, host_uname); return ENODEV; } if (!node->details->rsc_discovery_enabled) { out->err(out, "Unable to clean up %s because resource discovery disabled on %s", rsc->id, host_uname); return EOPNOTSUPP; } if (controld_api == NULL) { out->err(out, "Dry run: skipping clean-up of %s on %s due to CIB_file", rsc->id, host_uname); return pcmk_rc_ok; } rc = clear_rsc_fail_attrs(rsc, operation, interval_spec, node); if (rc != pcmk_rc_ok) { out->err(out, "Unable to clean up %s failures on %s: %s", rsc->id, host_uname, pcmk_rc_str(rc)); return rc; } if (just_failures) { rc = clear_rsc_failures(out, controld_api, host_uname, rsc->id, operation, interval_spec, data_set); } else { rc = clear_rsc_history(out, controld_api, host_uname, rsc->id, data_set); } if (rc != pcmk_rc_ok) { out->err(out, "Cleaned %s failures on %s, but unable to clean history: %s", rsc->id, host_uname, pcmk_strerror(rc)); } else { out->info(out, "Cleaned up %s on %s", rsc->id, host_uname); } return rc; } // \return Standard Pacemaker return code int cli_cleanup_all(pcmk__output_t *out, pcmk_ipc_api_t *controld_api, const char *node_name, const char *operation, const char *interval_spec, pe_working_set_t *data_set) { int rc = pcmk_rc_ok; int attr_options = pcmk__node_attr_none; const char *display_name = node_name? node_name : "all nodes"; if (controld_api == NULL) { out->info(out, "Dry run: skipping clean-up of %s due to CIB_file", display_name); return rc; } if (node_name) { pe_node_t *node = pe_find_node(data_set->nodes, node_name); if (node == NULL) { out->err(out, "Unknown node: %s", node_name); return ENXIO; } if (pe__is_guest_or_remote_node(node)) { attr_options |= pcmk__node_attr_remote; } } rc = pcmk__node_attr_request_clear(NULL, node_name, NULL, operation, interval_spec, NULL, attr_options); if (rc != pcmk_rc_ok) { out->err(out, "Unable to clean up all failures on %s: %s", display_name, pcmk_rc_str(rc)); return rc; } if (node_name) { rc = clear_rsc_failures(out, controld_api, node_name, NULL, operation, interval_spec, data_set); if (rc != pcmk_rc_ok) { out->err(out, "Cleaned all resource failures on %s, but unable to clean history: %s", node_name, pcmk_strerror(rc)); return rc; } } else { for (GList *iter = data_set->nodes; iter; iter = iter->next) { pe_node_t *node = (pe_node_t *) iter->data; rc = clear_rsc_failures(out, controld_api, node->details->uname, NULL, operation, interval_spec, data_set); if (rc != pcmk_rc_ok) { out->err(out, "Cleaned all resource failures on all nodes, but unable to clean history: %s", pcmk_strerror(rc)); return rc; } } } out->info(out, "Cleaned up all resources on %s", display_name); return rc; } int cli_resource_check(pcmk__output_t *out, cib_t * cib_conn, pe_resource_t *rsc) { char *role_s = NULL; char *managed = NULL; pe_resource_t *parent = uber_parent(rsc); int rc = pcmk_rc_no_output; resource_checks_t *checks = NULL; find_resource_attr(out, cib_conn, XML_NVPAIR_ATTR_VALUE, parent->id, NULL, NULL, NULL, XML_RSC_ATTR_MANAGED, &managed); find_resource_attr(out, cib_conn, XML_NVPAIR_ATTR_VALUE, parent->id, NULL, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, &role_s); checks = cli_check_resource(rsc, role_s, managed); if (checks->flags != 0 || checks->lock_node != NULL) { rc = out->message(out, "resource-check-list", checks); } free(role_s); free(managed); free(checks); return rc; } // \return Standard Pacemaker return code int cli_resource_fail(pcmk__output_t *out, pcmk_ipc_api_t *controld_api, const char *host_uname, const char *rsc_id, pe_working_set_t *data_set) { crm_notice("Failing %s on %s", rsc_id, host_uname); return send_lrm_rsc_op(out, controld_api, true, host_uname, rsc_id, data_set); } static GHashTable * generate_resource_params(pe_resource_t * rsc, pe_working_set_t * data_set) { GHashTable *params = NULL; GHashTable *meta = NULL; GHashTable *combined = NULL; GHashTableIter iter; if (!rsc) { crm_err("Resource does not exist in config"); return NULL; } params = crm_str_table_new(); meta = crm_str_table_new(); combined = crm_str_table_new(); get_rsc_attributes(params, rsc, NULL /* TODO: Pass in local node */ , data_set); get_meta_attributes(meta, rsc, NULL /* TODO: Pass in local node */ , data_set); if (params) { char *key = NULL; char *value = NULL; g_hash_table_iter_init(&iter, params); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { g_hash_table_insert(combined, strdup(key), strdup(value)); } g_hash_table_destroy(params); } if (meta) { char *key = NULL; char *value = NULL; g_hash_table_iter_init(&iter, meta); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { char *crm_name = crm_meta_name(key); g_hash_table_insert(combined, crm_name, strdup(value)); } g_hash_table_destroy(meta); } return combined; } bool resource_is_running_on(pe_resource_t *rsc, const char *host) { bool found = TRUE; GListPtr hIter = NULL; GListPtr hosts = NULL; if(rsc == NULL) { return FALSE; } rsc->fns->location(rsc, &hosts, TRUE); for (hIter = hosts; host != NULL && hIter != NULL; hIter = hIter->next) { pe_node_t *node = (pe_node_t *) hIter->data; if(strcmp(host, node->details->uname) == 0) { crm_trace("Resource %s is running on %s\n", rsc->id, host); goto done; } else if(strcmp(host, node->details->id) == 0) { crm_trace("Resource %s is running on %s\n", rsc->id, host); goto done; } } if(host != NULL) { crm_trace("Resource %s is not running on: %s\n", rsc->id, host); found = FALSE; } else if(host == NULL && hosts == NULL) { crm_trace("Resource %s is not running\n", rsc->id); found = FALSE; } done: g_list_free(hosts); return found; } /*! * \internal * \brief Create a list of all resources active on host from a given list * * \param[in] host Name of host to check whether resources are active * \param[in] rsc_list List of resources to check * * \return New list of resources from list that are active on host */ static GList * get_active_resources(const char *host, GList *rsc_list) { GList *rIter = NULL; GList *active = NULL; for (rIter = rsc_list; rIter != NULL; rIter = rIter->next) { pe_resource_t *rsc = (pe_resource_t *) rIter->data; /* Expand groups to their members, because if we're restarting a member * other than the first, we can't otherwise tell which resources are * stopping and starting. */ if (rsc->variant == pe_group) { active = g_list_concat(active, get_active_resources(host, rsc->children)); } else if (resource_is_running_on(rsc, host)) { active = g_list_append(active, strdup(rsc->id)); } } return active; } static void dump_list(GList *items, const char *tag) { int lpc = 0; GList *item = NULL; for (item = items; item != NULL; item = item->next) { crm_trace("%s[%d]: %s", tag, lpc, (char*)item->data); lpc++; } } static void display_list(pcmk__output_t *out, GList *items, const char *tag) { GList *item = NULL; for (item = items; item != NULL; item = item->next) { out->info(out, "%s%s", tag, (const char *)item->data); } } /*! * \internal * \brief Upgrade XML to latest schema version and use it as working set input * * This also updates the working set timestamp to the current time. * * \param[in] data_set Working set instance to update * \param[in] xml XML to use as input * * \return Standard Pacemaker return code * \note On success, caller is responsible for freeing memory allocated for * data_set->now. * \todo This follows the example of other callers of cli_config_update() * and returns ENOKEY ("Required key not available") if that fails, * but perhaps pcmk_rc_schema_validation would be better in that case. */ int update_working_set_xml(pe_working_set_t *data_set, xmlNode **xml) { if (cli_config_update(xml, NULL, FALSE) == FALSE) { return ENOKEY; } data_set->input = *xml; data_set->now = crm_time_new(NULL); return pcmk_rc_ok; } /*! * \internal * \brief Update a working set's XML input based on a CIB query * * \param[in] data_set Data set instance to initialize * \param[in] cib Connection to the CIB manager * * \return Standard Pacemaker return code * \note On success, caller is responsible for freeing memory allocated for * data_set->input and data_set->now. */ static int update_working_set_from_cib(pcmk__output_t *out, pe_working_set_t * data_set, cib_t *cib) { xmlNode *cib_xml_copy = NULL; int rc = pcmk_rc_ok; rc = cib->cmds->query(cib, NULL, &cib_xml_copy, cib_scope_local | cib_sync_call); rc = pcmk_legacy2rc(rc); if (rc != pcmk_rc_ok) { out->err(out, "Could not obtain the current CIB: %s (%d)", pcmk_strerror(rc), rc); return rc; } rc = update_working_set_xml(data_set, &cib_xml_copy); if (rc != pcmk_rc_ok) { out->err(out, "Could not upgrade the current CIB XML"); free_xml(cib_xml_copy); return rc; } return rc; } // \return Standard Pacemaker return code static int update_dataset(pcmk__output_t *out, cib_t *cib, pe_working_set_t * data_set, bool simulate) { char *pid = NULL; char *shadow_file = NULL; cib_t *shadow_cib = NULL; int rc = pcmk_rc_ok; pe_reset_working_set(data_set); rc = update_working_set_from_cib(out, data_set, cib); if (rc != pcmk_rc_ok) { return rc; } if(simulate) { pid = pcmk__getpid_s(); shadow_cib = cib_shadow_new(pid); shadow_file = get_shadow_file(pid); if (shadow_cib == NULL) { out->err(out, "Could not create shadow cib: '%s'", pid); rc = ENXIO; goto cleanup; } rc = write_xml_file(data_set->input, shadow_file, FALSE); if (rc < 0) { out->err(out, "Could not populate shadow cib: %s (%d)", pcmk_strerror(rc), rc); goto cleanup; } rc = shadow_cib->cmds->signon(shadow_cib, crm_system_name, cib_command); rc = pcmk_legacy2rc(rc); if (rc != pcmk_rc_ok) { out->err(out, "Could not connect to shadow cib: %s (%d)", pcmk_strerror(rc), rc); goto cleanup; } pcmk__schedule_actions(data_set, data_set->input, NULL); run_simulation(data_set, shadow_cib, NULL, TRUE); rc = update_dataset(out, shadow_cib, data_set, FALSE); } else { cluster_status(data_set); } cleanup: /* Do not free data_set->input here, we need rsc->xml to be valid later on */ cib_delete(shadow_cib); free(pid); if(shadow_file) { unlink(shadow_file); free(shadow_file); } return rc; } static int max_delay_for_resource(pe_working_set_t * data_set, pe_resource_t *rsc) { int delay = 0; int max_delay = 0; if(rsc && rsc->children) { GList *iter = NULL; for(iter = rsc->children; iter; iter = iter->next) { pe_resource_t *child = (pe_resource_t *)iter->data; delay = max_delay_for_resource(data_set, child); if(delay > max_delay) { double seconds = delay / 1000.0; crm_trace("Calculated new delay of %.1fs due to %s", seconds, child->id); max_delay = delay; } } } else if(rsc) { char *key = crm_strdup_printf("%s_%s_0", rsc->id, RSC_STOP); pe_action_t *stop = custom_action(rsc, key, RSC_STOP, NULL, TRUE, FALSE, data_set); const char *value = g_hash_table_lookup(stop->meta, XML_ATTR_TIMEOUT); max_delay = value? (int) crm_parse_ll(value, NULL) : -1; pe_free_action(stop); } return max_delay; } static int max_delay_in(pe_working_set_t * data_set, GList *resources) { int max_delay = 0; GList *item = NULL; for (item = resources; item != NULL; item = item->next) { int delay = 0; pe_resource_t *rsc = pe_find_resource(data_set->resources, (const char *)item->data); delay = max_delay_for_resource(data_set, rsc); if(delay > max_delay) { double seconds = delay / 1000.0; crm_trace("Calculated new delay of %.1fs due to %s", seconds, rsc->id); max_delay = delay; } } return 5 + (max_delay / 1000); } #define waiting_for_starts(d, r, h) ((d != NULL) || \ (resource_is_running_on((r), (h)) == FALSE)) /*! * \internal * \brief Restart a resource (on a particular host if requested). * * \param[in] rsc The resource to restart * \param[in] host The host to restart the resource on (or NULL for all) * \param[in] timeout_ms Consider failed if actions do not complete in this time * (specified in milliseconds, but a two-second * granularity is actually used; if 0, a timeout will be * calculated based on the resource timeout) * \param[in] cib Connection to the CIB manager * * \return Standard Pacemaker return code (exits on certain failures) */ int cli_resource_restart(pcmk__output_t *out, pe_resource_t *rsc, const char *host, const char *move_lifetime, int timeout_ms, cib_t *cib, int cib_options, gboolean promoted_role_only, gboolean force) { int rc = pcmk_rc_ok; int lpc = 0; int before = 0; int step_timeout_s = 0; int sleep_interval = 2; int timeout = timeout_ms / 1000; bool stop_via_ban = FALSE; char *rsc_id = NULL; char *orig_target_role = NULL; GList *list_delta = NULL; GList *target_active = NULL; GList *current_active = NULL; GList *restart_target_active = NULL; pe_working_set_t *data_set = NULL; if(resource_is_running_on(rsc, host) == FALSE) { const char *id = rsc->clone_name?rsc->clone_name:rsc->id; if(host) { out->err(out, "%s is not running on %s and so cannot be restarted", id, host); } else { out->err(out, "%s is not running anywhere and so cannot be restarted", id); } return ENXIO; } rsc_id = strdup(rsc->id); if ((pe_rsc_is_clone(rsc) || pe_bundle_replicas(rsc)) && host) { stop_via_ban = TRUE; } /* grab full cib determine originally active resources disable or ban poll cib and watch for affected resources to get stopped without --timeout, calculate the stop timeout for each step and wait for that if we hit --timeout or the service timeout, re-enable or un-ban, report failure and indicate which resources we couldn't take down if everything stopped, re-enable or un-ban poll cib and watch for affected resources to get started without --timeout, calculate the start timeout for each step and wait for that if we hit --timeout or the service timeout, report (different) failure and indicate which resources we couldn't bring back up report success Optimizations: - use constraints to determine ordered list of affected resources - Allow a --no-deps option (aka. --force-restart) */ data_set = pe_new_working_set(); if (data_set == NULL) { crm_perror(LOG_ERR, "Could not allocate working set"); rc = ENOMEM; goto done; } pe__set_working_set_flags(data_set, pe_flag_no_counts|pe_flag_no_compat); rc = update_dataset(out, cib, data_set, FALSE); if(rc != pcmk_rc_ok) { out->err(out, "Could not get new resource list: %s (%d)", pcmk_strerror(rc), rc); goto done; } restart_target_active = get_active_resources(host, data_set->resources); current_active = get_active_resources(host, data_set->resources); dump_list(current_active, "Origin"); if (stop_via_ban) { /* Stop the clone or bundle instance by banning it from the host */ out->quiet = true; rc = cli_resource_ban(out, rsc_id, host, move_lifetime, NULL, cib, cib_options, promoted_role_only); } else { /* Stop the resource by setting target-role to Stopped. * Remember any existing target-role so we can restore it later * (though it only makes any difference if it's Slave). */ char *lookup_id = clone_strip(rsc->id); find_resource_attr(out, cib, XML_NVPAIR_ATTR_VALUE, lookup_id, NULL, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, &orig_target_role); free(lookup_id); rc = cli_resource_update_attribute(out, rsc, rsc_id, NULL, XML_TAG_META_SETS, NULL, XML_RSC_ATTR_TARGET_ROLE, RSC_STOPPED, FALSE, cib, cib_options, data_set, force); } if(rc != pcmk_rc_ok) { out->err(out, "Could not set target-role for %s: %s (%d)", rsc_id, pcmk_strerror(rc), rc); if (current_active) { g_list_free_full(current_active, free); } if (restart_target_active) { g_list_free_full(restart_target_active, free); } goto done; } rc = update_dataset(out, cib, data_set, TRUE); if(rc != pcmk_rc_ok) { out->err(out, "Could not determine which resources would be stopped"); goto failure; } target_active = get_active_resources(host, data_set->resources); dump_list(target_active, "Target"); list_delta = pcmk__subtract_lists(current_active, target_active, (GCompareFunc) strcmp); out->info(out, "Waiting for %d resources to stop:", g_list_length(list_delta)); display_list(out, list_delta, " * "); step_timeout_s = timeout / sleep_interval; while (list_delta != NULL) { before = g_list_length(list_delta); if(timeout_ms == 0) { step_timeout_s = max_delay_in(data_set, list_delta) / sleep_interval; } /* We probably don't need the entire step timeout */ for(lpc = 0; (lpc < step_timeout_s) && (list_delta != NULL); lpc++) { sleep(sleep_interval); if(timeout) { timeout -= sleep_interval; crm_trace("%ds remaining", timeout); } rc = update_dataset(out, cib, data_set, FALSE); if(rc != pcmk_rc_ok) { out->err(out, "Could not determine which resources were stopped"); goto failure; } if (current_active) { g_list_free_full(current_active, free); } current_active = get_active_resources(host, data_set->resources); g_list_free(list_delta); list_delta = pcmk__subtract_lists(current_active, target_active, (GCompareFunc) strcmp); dump_list(current_active, "Current"); dump_list(list_delta, "Delta"); } crm_trace("%d (was %d) resources remaining", g_list_length(list_delta), before); if(before == g_list_length(list_delta)) { /* aborted during stop phase, print the contents of list_delta */ out->info(out, "Could not complete shutdown of %s, %d resources remaining", rsc_id, g_list_length(list_delta)); display_list(out, list_delta, " * "); rc = ETIME; goto failure; } } if (stop_via_ban) { rc = cli_resource_clear(rsc_id, host, NULL, cib, cib_options, TRUE, force); } else if (orig_target_role) { rc = cli_resource_update_attribute(out, rsc, rsc_id, NULL, XML_TAG_META_SETS, NULL, XML_RSC_ATTR_TARGET_ROLE, orig_target_role, FALSE, cib, cib_options, data_set, force); free(orig_target_role); orig_target_role = NULL; } else { rc = cli_resource_delete_attribute(out, rsc, rsc_id, NULL, XML_TAG_META_SETS, NULL, XML_RSC_ATTR_TARGET_ROLE, cib, cib_options, data_set, force); } if(rc != pcmk_rc_ok) { out->err(out, "Could not unset target-role for %s: %s (%d)", rsc_id, pcmk_strerror(rc), rc); goto done; } if (target_active) { g_list_free_full(target_active, free); } target_active = restart_target_active; list_delta = pcmk__subtract_lists(target_active, current_active, (GCompareFunc) strcmp); out->info(out, "Waiting for %d resources to start again:", g_list_length(list_delta)); display_list(out, list_delta, " * "); step_timeout_s = timeout / sleep_interval; while (waiting_for_starts(list_delta, rsc, host)) { before = g_list_length(list_delta); if(timeout_ms == 0) { step_timeout_s = max_delay_in(data_set, list_delta) / sleep_interval; } /* We probably don't need the entire step timeout */ for (lpc = 0; (lpc < step_timeout_s) && waiting_for_starts(list_delta, rsc, host); lpc++) { sleep(sleep_interval); if(timeout) { timeout -= sleep_interval; crm_trace("%ds remaining", timeout); } rc = update_dataset(out, cib, data_set, FALSE); if(rc != pcmk_rc_ok) { out->err(out, "Could not determine which resources were started"); goto failure; } if (current_active) { g_list_free_full(current_active, free); } /* It's OK if dependent resources moved to a different node, * so we check active resources on all nodes. */ current_active = get_active_resources(NULL, data_set->resources); g_list_free(list_delta); list_delta = pcmk__subtract_lists(target_active, current_active, (GCompareFunc) strcmp); dump_list(current_active, "Current"); dump_list(list_delta, "Delta"); } if(before == g_list_length(list_delta)) { /* aborted during start phase, print the contents of list_delta */ out->info(out, "Could not complete restart of %s, %d resources remaining", rsc_id, g_list_length(list_delta)); display_list(out, list_delta, " * "); rc = ETIME; goto failure; } } rc = pcmk_rc_ok; goto done; failure: if (stop_via_ban) { cli_resource_clear(rsc_id, host, NULL, cib, cib_options, TRUE, force); } else if (orig_target_role) { cli_resource_update_attribute(out, rsc, rsc_id, NULL, XML_TAG_META_SETS, NULL, XML_RSC_ATTR_TARGET_ROLE, orig_target_role, FALSE, cib, cib_options, data_set, force); free(orig_target_role); } else { cli_resource_delete_attribute(out, rsc, rsc_id, NULL, XML_TAG_META_SETS, NULL, XML_RSC_ATTR_TARGET_ROLE, cib, cib_options, data_set, force); } done: if (list_delta) { g_list_free(list_delta); } if (current_active) { g_list_free_full(current_active, free); } if (target_active && (target_active != restart_target_active)) { g_list_free_full(target_active, free); } if (restart_target_active) { g_list_free_full(restart_target_active, free); } free(rsc_id); pe_free_working_set(data_set); return rc; } static inline bool action_is_pending(pe_action_t *action) { if (pcmk_any_flags_set(action->flags, pe_action_optional|pe_action_pseudo) || !pcmk_is_set(action->flags, pe_action_runnable) || pcmk__str_eq("notify", action->task, pcmk__str_casei)) { return false; } return true; } /*! * \internal * \brief Return TRUE if any actions in a list are pending * * \param[in] actions List of actions to check * * \return TRUE if any actions in the list are pending, FALSE otherwise */ static bool actions_are_pending(GListPtr actions) { GListPtr action; for (action = actions; action != NULL; action = action->next) { pe_action_t *a = (pe_action_t *)action->data; if (action_is_pending(a)) { crm_notice("Waiting for %s (flags=0x%.8x)", a->uuid, a->flags); return TRUE; } } return FALSE; } static void print_pending_actions(pcmk__output_t *out, GListPtr actions) { GListPtr action; out->info(out, "Pending actions:"); for (action = actions; action != NULL; action = action->next) { pe_action_t *a = (pe_action_t *) action->data; if (!action_is_pending(a)) { continue; } if (a->node) { out->info(out, "\tAction %d: %s\ton %s", a->id, a->uuid, a->node->details->uname); } else { out->info(out, "\tAction %d: %s", a->id, a->uuid); } } } /* For --wait, timeout (in seconds) to use if caller doesn't specify one */ #define WAIT_DEFAULT_TIMEOUT_S (60 * 60) /* For --wait, how long to sleep between cluster state checks */ #define WAIT_SLEEP_S (2) /*! * \internal * \brief Wait until all pending cluster actions are complete * * This waits until either the CIB's transition graph is idle or a timeout is * reached. * * \param[in] timeout_ms Consider failed if actions do not complete in this time * (specified in milliseconds, but one-second granularity * is actually used; if 0, a default will be used) * \param[in] cib Connection to the CIB manager * * \return Standard Pacemaker return code */ int wait_till_stable(pcmk__output_t *out, int timeout_ms, cib_t * cib) { pe_working_set_t *data_set = NULL; int rc = pcmk_rc_ok; int timeout_s = timeout_ms? ((timeout_ms + 999) / 1000) : WAIT_DEFAULT_TIMEOUT_S; time_t expire_time = time(NULL) + timeout_s; time_t time_diff; bool printed_version_warning = out->is_quiet(out); // i.e. don't print if quiet data_set = pe_new_working_set(); if (data_set == NULL) { return ENOMEM; } pe__set_working_set_flags(data_set, pe_flag_no_counts|pe_flag_no_compat); do { /* Abort if timeout is reached */ time_diff = expire_time - time(NULL); if (time_diff > 0) { crm_info("Waiting up to %ld seconds for cluster actions to complete", time_diff); } else { print_pending_actions(out, data_set->actions); pe_free_working_set(data_set); return ETIME; } if (rc == pcmk_rc_ok) { /* this avoids sleep on first loop iteration */ sleep(WAIT_SLEEP_S); } /* Get latest transition graph */ pe_reset_working_set(data_set); rc = update_working_set_from_cib(out, data_set, cib); if (rc != pcmk_rc_ok) { pe_free_working_set(data_set); return rc; } pcmk__schedule_actions(data_set, data_set->input, NULL); if (!printed_version_warning) { /* If the DC has a different version than the local node, the two * could come to different conclusions about what actions need to be * done. Warn the user in this case. * * @TODO A possible long-term solution would be to reimplement the * wait as a new controller operation that would be forwarded to the * DC. However, that would have potential problems of its own. */ const char *dc_version = g_hash_table_lookup(data_set->config_hash, "dc-version"); if (!pcmk__str_eq(dc_version, PACEMAKER_VERSION "-" BUILD_VERSION, pcmk__str_casei)) { out->info(out, "warning: wait option may not work properly in " "mixed-version cluster"); printed_version_warning = TRUE; } } } while (actions_are_pending(data_set->actions)); pe_free_working_set(data_set); return rc; } crm_exit_t cli_resource_execute_from_params(pcmk__output_t *out, const char *rsc_name, const char *rsc_class, const char *rsc_prov, const char *rsc_type, const char *action, GHashTable *params, GHashTable *override_hash, int timeout_ms, int resource_verbose, gboolean force) { GHashTable *params_copy = NULL; crm_exit_t exit_code = CRM_EX_OK; svc_action_t *op = NULL; if (pcmk__str_eq(rsc_class, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { out->err(out, "Sorry, the %s option doesn't support %s resources yet", action, rsc_class); crm_exit(CRM_EX_UNIMPLEMENT_FEATURE); } /* If no timeout was provided, grab the default. */ if (timeout_ms == 0) { timeout_ms = crm_get_msec(CRM_DEFAULT_OP_TIMEOUT_S); } /* add meta_timeout env needed by some resource agents */ g_hash_table_insert(params, strdup("CRM_meta_timeout"), crm_strdup_printf("%d", timeout_ms)); /* add crm_feature_set env needed by some resource agents */ g_hash_table_insert(params, strdup(XML_ATTR_CRM_VERSION), strdup(CRM_FEATURE_SET)); /* resources_action_create frees the params hash table it's passed, but we * may need to reuse it in a second call to resources_action_create. Thus * we'll make a copy here so that gets freed and the original remains for * reuse. */ params_copy = crm_str_table_dup(params); op = resources_action_create(rsc_name, rsc_class, rsc_prov, rsc_type, action, 0, timeout_ms, params_copy, 0); if (op == NULL) { /* Re-run with stderr enabled so we can display a sane error message */ crm_enable_stderr(TRUE); params_copy = crm_str_table_dup(params); op = resources_action_create(rsc_name, rsc_class, rsc_prov, rsc_type, action, 0, timeout_ms, params_copy, 0); /* Callers of cli_resource_execute expect that the params hash table will * be freed. That function uses this one, so for that reason and for * making the two act the same, we should free the hash table here too. */ g_hash_table_destroy(params); /* We know op will be NULL, but this makes static analysis happy */ services_action_free(op); crm_exit(CRM_EX_DATAERR); return exit_code; // Never reached, but helps static analysis } setenv("HA_debug", resource_verbose > 0 ? "1" : "0", 1); if(resource_verbose > 1) { setenv("OCF_TRACE_RA", "1", 1); } /* A resource agent using the standard ocf-shellfuncs library will not print * messages to stderr if it doesn't have a controlling terminal (e.g. if * crm_resource is called via script or ssh). This forces it to do so. */ setenv("OCF_TRACE_FILE", "/dev/stderr", 0); if (override_hash) { GHashTableIter iter; char *name = NULL; char *value = NULL; g_hash_table_iter_init(&iter, override_hash); while (g_hash_table_iter_next(&iter, (gpointer *) & name, (gpointer *) & value)) { out->info(out, "Overriding the cluster configuration for '%s' with '%s' = '%s'", rsc_name, name, value); g_hash_table_replace(op->params, strdup(name), strdup(value)); } } if (services_action_sync(op)) { exit_code = op->rc; if (op->status == PCMK_LRM_OP_DONE) { out->info(out, "Operation %s for %s (%s:%s:%s) returned: '%s' (%d)", action, rsc_name, rsc_class, rsc_prov ? rsc_prov : "", rsc_type, services_ocf_exitcode_str(op->rc), op->rc); } else { out->info(out, "Operation %s for %s (%s:%s:%s) failed: '%s' (%d)", action, rsc_name, rsc_class, rsc_prov ? rsc_prov : "", rsc_type, services_lrm_status_str(op->status), op->status); } /* hide output for validate-all if not in verbose */ if (resource_verbose == 0 && pcmk__str_eq(action, "validate-all", pcmk__str_casei)) goto done; if (op->stdout_data || op->stderr_data) { out->subprocess_output(out, op->rc, op->stdout_data, op->stderr_data); } } else { exit_code = op->rc == 0 ? CRM_EX_ERROR : op->rc; } done: services_action_free(op); /* See comment above about why we free params here. */ g_hash_table_destroy(params); return exit_code; } crm_exit_t cli_resource_execute(pcmk__output_t *out, pe_resource_t *rsc, const char *requested_name, const char *rsc_action, GHashTable *override_hash, int timeout_ms, cib_t * cib, pe_working_set_t *data_set, int resource_verbose, gboolean force) { crm_exit_t exit_code = CRM_EX_OK; const char *rid = NULL; const char *rtype = NULL; const char *rprov = NULL; const char *rclass = NULL; const char *action = NULL; GHashTable *params = NULL; if (pcmk__str_eq(rsc_action, "validate", pcmk__str_casei)) { action = "validate-all"; } else if (pcmk__str_eq(rsc_action, "force-check", pcmk__str_casei)) { action = "monitor"; } else if (pcmk__str_eq(rsc_action, "force-stop", pcmk__str_casei)) { action = rsc_action+6; } else if (pcmk__strcase_any_of(rsc_action, "force-start", "force-demote", "force-promote", NULL)) { action = rsc_action+6; if(pe_rsc_is_clone(rsc)) { GListPtr rscs = cli_resource_search(out, rsc, requested_name, data_set); if(rscs != NULL && force == FALSE) { out->err(out, "It is not safe to %s %s here: the cluster claims it is already active", action, rsc->id); out->err(out, "Try setting target-role=Stopped first or specifying " "the force option"); return CRM_EX_UNSAFE; } } } else { action = rsc_action; } if(pe_rsc_is_clone(rsc)) { /* Grab the first child resource in the hope it's not a group */ rsc = rsc->children->data; } if(rsc->variant == pe_group) { out->err(out, "Sorry, the %s option doesn't support group resources", rsc_action); return CRM_EX_UNIMPLEMENT_FEATURE; } rclass = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS); rprov = crm_element_value(rsc->xml, XML_AGENT_ATTR_PROVIDER); rtype = crm_element_value(rsc->xml, XML_ATTR_TYPE); params = generate_resource_params(rsc, data_set); if (timeout_ms == 0) { timeout_ms = pe_get_configured_timeout(rsc, action, data_set); } rid = pe_rsc_is_anon_clone(rsc->parent)? requested_name : rsc->id; exit_code = cli_resource_execute_from_params(out, rid, rclass, rprov, rtype, action, params, override_hash, timeout_ms, resource_verbose, force); return exit_code; } // \return Standard Pacemaker return code int cli_resource_move(pcmk__output_t *out, pe_resource_t *rsc, const char *rsc_id, const char *host_name, const char *move_lifetime, cib_t *cib, int cib_options, pe_working_set_t *data_set, gboolean promoted_role_only, gboolean force) { int rc = pcmk_rc_ok; unsigned int count = 0; pe_node_t *current = NULL; pe_node_t *dest = pe_find_node(data_set->nodes, host_name); bool cur_is_dest = FALSE; if (dest == NULL) { return pcmk_rc_node_unknown; } if (promoted_role_only && !pcmk_is_set(rsc->flags, pe_rsc_promotable)) { pe_resource_t *p = uber_parent(rsc); if (pcmk_is_set(p->flags, pe_rsc_promotable)) { out->info(out, "Using parent '%s' for move instead of '%s'.", rsc->id, rsc_id); rsc_id = p->id; rsc = p; } else { out->info(out, "Ignoring master option: %s is not promotable", rsc_id); promoted_role_only = FALSE; } } current = pe__find_active_requires(rsc, &count); if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) { GListPtr iter = NULL; unsigned int master_count = 0; pe_node_t *master_node = NULL; for(iter = rsc->children; iter; iter = iter->next) { pe_resource_t *child = (pe_resource_t *)iter->data; enum rsc_role_e child_role = child->fns->state(child, TRUE); if(child_role == RSC_ROLE_MASTER) { rsc = child; master_node = pe__current_node(child); master_count++; } } if (promoted_role_only || master_count) { count = master_count; current = master_node; } } if (count > 1) { if (pe_rsc_is_clone(rsc)) { current = NULL; } else { return pcmk_rc_multiple; } } if (current && (current->details == dest->details)) { cur_is_dest = TRUE; if (force) { crm_info("%s is already %s on %s, reinforcing placement with location constraint.", rsc_id, promoted_role_only?"promoted":"active", dest->details->uname); } else { return pcmk_rc_already; } } /* Clear any previous prefer constraints across all nodes. */ cli_resource_clear(rsc_id, NULL, data_set->nodes, cib, cib_options, FALSE, force); /* Clear any previous ban constraints on 'dest'. */ cli_resource_clear(rsc_id, dest->details->uname, data_set->nodes, cib, cib_options, TRUE, force); /* Record an explicit preference for 'dest' */ rc = cli_resource_prefer(out, rsc_id, dest->details->uname, move_lifetime, cib, cib_options, promoted_role_only); crm_trace("%s%s now prefers node %s%s", rsc->id, promoted_role_only?" (master)":"", dest->details->uname, force?"(forced)":""); /* only ban the previous location if current location != destination location. * it is possible to use -M to enforce a location without regard of where the * resource is currently located */ if(force && (cur_is_dest == FALSE)) { /* Ban the original location if possible */ if(current) { (void)cli_resource_ban(out, rsc_id, current->details->uname, move_lifetime, NULL, cib, cib_options, promoted_role_only); } else if(count > 1) { out->info(out, "Resource '%s' is currently %s in %d locations. " "One may now move to %s", rsc_id, (promoted_role_only? "promoted" : "active"), count, dest->details->uname); out->info(out, "To prevent '%s' from being %s at a specific location, " "specify a node.", rsc_id, (promoted_role_only? "promoted" : "active")); } else { crm_trace("Not banning %s from its current location: not active", rsc_id); } } return rc; }