Page MenuHomeClusterLabs Projects

No OneTemporary

diff --git a/lib/pacemaker/libpacemaker_private.h b/lib/pacemaker/libpacemaker_private.h
index dfa8ef14c1..c27c8ca518 100644
--- a/lib/pacemaker/libpacemaker_private.h
+++ b/lib/pacemaker/libpacemaker_private.h
@@ -1,1100 +1,1094 @@
/*
* Copyright 2021-2023 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 PCMK__LIBPACEMAKER_PRIVATE__H
# define PCMK__LIBPACEMAKER_PRIVATE__H
/* This header is for the sole use of libpacemaker, so that functions can be
* declared with G_GNUC_INTERNAL for efficiency.
*/
#include <crm/pengine/pe_types.h> // pe_action_t, pe_node_t, pe_working_set_t
#include <crm/pengine/internal.h> // pe__location_t
// Colocation flags
enum pcmk__coloc_flags {
pcmk__coloc_none = 0U,
// Primary is affected even if already active
pcmk__coloc_influence = (1U << 0),
// Colocation was explicitly configured in CIB
pcmk__coloc_explicit = (1U << 1),
};
// Flags to modify the behavior of add_colocated_node_scores()
enum pcmk__coloc_select {
// With no other flags, apply all "with this" colocations
pcmk__coloc_select_default = 0,
// Apply "this with" colocations instead of "with this" colocations
pcmk__coloc_select_this_with = (1 << 0),
// Apply only colocations with non-negative scores
pcmk__coloc_select_nonnegative = (1 << 1),
// Apply only colocations with at least one matching node
pcmk__coloc_select_active = (1 << 2),
};
// Flags the update_ordered_actions() method can return
enum pcmk__updated {
pcmk__updated_none = 0, // Nothing changed
pcmk__updated_first = (1 << 0), // First action was updated
pcmk__updated_then = (1 << 1), // Then action was updated
};
#define pcmk__set_updated_flags(au_flags, action, flags_to_set) do { \
au_flags = pcmk__set_flags_as(__func__, __LINE__, \
LOG_TRACE, "Action update", \
(action)->uuid, au_flags, \
(flags_to_set), #flags_to_set); \
} while (0)
#define pcmk__clear_updated_flags(au_flags, action, flags_to_clear) do { \
au_flags = pcmk__clear_flags_as(__func__, __LINE__, \
LOG_TRACE, "Action update", \
(action)->uuid, au_flags, \
(flags_to_clear), #flags_to_clear); \
} while (0)
// Resource assignment methods
struct resource_alloc_functions_s {
/*!
* \internal
* \brief Assign a resource to a node
*
* \param[in,out] rsc Resource to assign to a node
* \param[in] prefer Node to prefer, if all else is equal
* \param[in] stop_if_fail If \c true and \p rsc can't be assigned to a
* node, set next role to stopped and update
* existing actions (if \p rsc is not a
* primitive, this applies to its primitive
* descendants instead)
*
* \return Node that \p rsc is assigned to, if assigned entirely to one node
*
* \note If \p stop_if_fail is \c false, then \c pcmk__unassign_resource()
* can completely undo the assignment. A successful assignment can be
* either undone or left alone as final. A failed assignment has the
* same effect as calling pcmk__unassign_resource(); there are no side
* effects on roles or actions.
*/
pe_node_t *(*assign)(pe_resource_t *rsc, const pe_node_t *prefer,
bool stop_if_fail);
/*!
* \internal
* \brief Create all actions needed for a given resource
*
* \param[in,out] rsc Resource to create actions for
*/
void (*create_actions)(pe_resource_t *rsc);
/*!
* \internal
* \brief Schedule any probes needed for a resource on a node
*
* \param[in,out] rsc Resource to create probe for
* \param[in,out] node Node to create probe on
*
* \return true if any probe was created, otherwise false
*/
bool (*create_probe)(pe_resource_t *rsc, pe_node_t *node);
/*!
* \internal
* \brief Create implicit constraints needed for a resource
*
* \param[in,out] rsc Resource to create implicit constraints for
*/
void (*internal_constraints)(pe_resource_t *rsc);
/*!
* \internal
* \brief Apply a colocation's score to node scores or resource priority
*
* Given a colocation constraint, apply its score to the dependent's
* allowed node scores (if we are still placing resources) or priority (if
* we are choosing promotable clone instance roles).
*
* \param[in,out] dependent Dependent resource in colocation
* \param[in] primary Primary resource in colocation
* \param[in] colocation Colocation constraint to apply
* \param[in] for_dependent true if called on behalf of dependent
*/
void (*apply_coloc_score)(pe_resource_t *dependent,
const pe_resource_t *primary,
const pcmk__colocation_t *colocation,
bool for_dependent);
/*!
* \internal
* \brief Create list of all resources in colocations with a given resource
*
* Given a resource, create a list of all resources involved in mandatory
* colocations with it, whether directly or via chained colocations.
*
* \param[in] rsc Resource to add to colocated list
* \param[in] orig_rsc Resource originally requested
* \param[in,out] colocated_rscs Existing list
*
* \return List of given resource and all resources involved in colocations
*
* \note This function is recursive; top-level callers should pass NULL as
* \p colocated_rscs and \p orig_rsc, and the desired resource as
* \p rsc. The recursive calls will use other values.
*/
GList *(*colocated_resources)(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc,
GList *colocated_rscs);
/*!
* \internal
* \brief Add colocations affecting a resource as primary to a list
*
* Given a resource being assigned (\p orig_rsc) and a resource somewhere in
* its chain of ancestors (\p rsc, which may be \p orig_rsc), get
* colocations that affect the ancestor as primary and should affect the
* resource, and add them to a given list.
*
* \param[in] rsc Resource whose colocations should be added
* \param[in] orig_rsc Affected resource (\p rsc or a descendant)
* \param[in,out] list List of colocations to add to
*
* \note All arguments should be non-NULL.
* \note The pcmk__with_this_colocations() wrapper should usually be used
* instead of using this method directly.
*/
void (*with_this_colocations)(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc, GList **list);
/*!
* \internal
* \brief Add colocations affecting a resource as dependent to a list
*
* Given a resource being assigned (\p orig_rsc) and a resource somewhere in
* its chain of ancestors (\p rsc, which may be \p orig_rsc), get
* colocations that affect the ancestor as dependent and should affect the
* resource, and add them to a given list.
*
*
* \param[in] rsc Resource whose colocations should be added
* \param[in] orig_rsc Affected resource (\p rsc or a descendant)
* \param[in,out] list List of colocations to add to
*
* \note All arguments should be non-NULL.
* \note The pcmk__this_with_colocations() wrapper should usually be used
* instead of using this method directly.
*/
void (*this_with_colocations)(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc, GList **list);
/*!
* \internal
* \brief Update nodes with scores of colocated resources' nodes
*
* Given a table of nodes and a resource, update the nodes' scores with the
* scores of the best nodes matching the attribute used for each of the
* resource's relevant colocations.
*
* \param[in,out] rsc Resource to check colocations for
* \param[in] log_id Resource ID for logs (if NULL, use \p rsc ID)
* \param[in,out] nodes Nodes to update (set initial contents to NULL
* to copy \p rsc's allowed nodes)
* \param[in] colocation Original colocation constraint (used to get
* configured primary resource's stickiness, and
* to get colocation node attribute; if NULL,
* \p rsc's own matching node scores will not be
* added, and *nodes must be NULL as well)
* \param[in] factor Incorporate scores multiplied by this factor
* \param[in] flags Bitmask of enum pcmk__coloc_select values
*
* \note NULL *nodes, NULL colocation, and the pcmk__coloc_select_this_with
* flag are used together (and only by cmp_resources()).
* \note The caller remains responsible for freeing \p *nodes.
*/
void (*add_colocated_node_scores)(pe_resource_t *rsc, const char *log_id,
GHashTable **nodes,
pcmk__colocation_t *colocation,
float factor, uint32_t flags);
/*!
* \internal
* \brief Apply a location constraint to a resource's allowed node scores
*
* \param[in,out] rsc Resource to apply constraint to
* \param[in,out] location Location constraint to apply
*/
void (*apply_location)(pe_resource_t *rsc, pe__location_t *location);
/*!
* \internal
* \brief Return action flags for a given resource action
*
* \param[in,out] action Action to get flags for
* \param[in] node If not NULL, limit effects to this node
*
* \return Flags appropriate to \p action on \p node
* \note For primitives, this will be the same as action->flags regardless
* of node. For collective resources, the flags can differ due to
* multiple instances possibly being involved.
*/
uint32_t (*action_flags)(pe_action_t *action, const pe_node_t *node);
/*!
* \internal
* \brief Update two actions according to an ordering between them
*
* Given information about an ordering of two actions, update the actions'
* flags (and runnable_before members if appropriate) as appropriate for the
* ordering. Effects may cascade to other orderings involving the actions as
* well.
*
* \param[in,out] first 'First' action in an ordering
* \param[in,out] then 'Then' action in an ordering
* \param[in] node If not NULL, limit scope of ordering to this
* node (only used when interleaving instances)
* \param[in] flags Action flags for \p first for ordering purposes
* \param[in] filter Action flags to limit scope of certain updates
* (may include pe_action_optional to affect only
* mandatory actions, and pe_action_runnable to
* affect only runnable actions)
* \param[in] type Group of enum pe_ordering flags to apply
* \param[in,out] data_set Cluster working set
*
* \return Group of enum pcmk__updated flags indicating what was updated
*/
uint32_t (*update_ordered_actions)(pe_action_t *first, pe_action_t *then,
const pe_node_t *node, uint32_t flags,
uint32_t filter, uint32_t type,
pe_working_set_t *data_set);
/*!
* \internal
* \brief Output a summary of scheduled actions for a resource
*
* \param[in,out] rsc Resource to output actions for
*/
void (*output_actions)(pe_resource_t *rsc);
/*!
* \internal
* \brief Add a resource's actions to the transition graph
*
* \param[in,out] rsc Resource whose actions should be added
*/
void (*add_actions_to_graph)(pe_resource_t *rsc);
/*!
* \internal
* \brief Add meta-attributes relevant to transition graph actions to XML
*
* If a given resource supports variant-specific meta-attributes that are
* needed for transition graph actions, add them to a given XML element.
*
* \param[in] rsc Resource whose meta-attributes should be added
* \param[in,out] xml Transition graph action attributes XML to add to
*/
void (*add_graph_meta)(const pe_resource_t *rsc, xmlNode *xml);
/*!
* \internal
* \brief Add a resource's utilization to a table of utilization values
*
* This function is used when summing the utilization of a resource and all
* resources colocated with it, to determine whether a node has sufficient
* capacity. Given a resource and a table of utilization values, it will add
* the resource's utilization to the existing values, if the resource has
* not yet been assigned to a node.
*
* \param[in] rsc Resource with utilization to add
* \param[in] orig_rsc Resource being assigned (for logging only)
* \param[in] all_rscs List of all resources that will be summed
* \param[in,out] utilization Table of utilization values to add to
*/
void (*add_utilization)(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc, GList *all_rscs,
GHashTable *utilization);
/*!
* \internal
* \brief Apply a shutdown lock for a resource, if appropriate
*
* \param[in,out] rsc Resource to check for shutdown lock
*/
void (*shutdown_lock)(pe_resource_t *rsc);
};
// Actions (pcmk_sched_actions.c)
G_GNUC_INTERNAL
void pcmk__update_action_for_orderings(pe_action_t *action,
pe_working_set_t *data_set);
G_GNUC_INTERNAL
uint32_t pcmk__update_ordered_actions(pe_action_t *first, pe_action_t *then,
const pe_node_t *node, uint32_t flags,
uint32_t filter, uint32_t type,
pe_working_set_t *data_set);
G_GNUC_INTERNAL
void pcmk__log_action(const char *pre_text, const pe_action_t *action,
bool details);
G_GNUC_INTERNAL
pe_action_t *pcmk__new_cancel_action(pe_resource_t *rsc, const char *name,
guint interval_ms, const pe_node_t *node);
G_GNUC_INTERNAL
pe_action_t *pcmk__new_shutdown_action(pe_node_t *node);
G_GNUC_INTERNAL
bool pcmk__action_locks_rsc_to_node(const pe_action_t *action);
G_GNUC_INTERNAL
void pcmk__deduplicate_action_inputs(pe_action_t *action);
G_GNUC_INTERNAL
void pcmk__output_actions(pe_working_set_t *data_set);
G_GNUC_INTERNAL
bool pcmk__check_action_config(pe_resource_t *rsc, pe_node_t *node,
const xmlNode *xml_op);
G_GNUC_INTERNAL
void pcmk__handle_rsc_config_changes(pe_working_set_t *data_set);
// Recurring actions (pcmk_sched_recurring.c)
G_GNUC_INTERNAL
void pcmk__create_recurring_actions(pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__schedule_cancel(pe_resource_t *rsc, const char *call_id,
const char *task, guint interval_ms,
const pe_node_t *node, const char *reason);
G_GNUC_INTERNAL
void pcmk__reschedule_recurring(pe_resource_t *rsc, const char *task,
guint interval_ms, pe_node_t *node);
G_GNUC_INTERNAL
bool pcmk__action_is_recurring(const pe_action_t *action);
// Producing transition graphs (pcmk_graph_producer.c)
G_GNUC_INTERNAL
bool pcmk__graph_has_loop(const pe_action_t *init_action,
const pe_action_t *action,
pe_action_wrapper_t *input);
G_GNUC_INTERNAL
void pcmk__add_rsc_actions_to_graph(pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__create_graph(pe_working_set_t *data_set);
// Fencing (pcmk_sched_fencing.c)
G_GNUC_INTERNAL
void pcmk__order_vs_fence(pe_action_t *stonith_op, pe_working_set_t *data_set);
G_GNUC_INTERNAL
void pcmk__order_vs_unfence(const pe_resource_t *rsc, pe_node_t *node,
pe_action_t *action, enum pe_ordering order);
G_GNUC_INTERNAL
void pcmk__fence_guest(pe_node_t *node);
G_GNUC_INTERNAL
bool pcmk__node_unfenced(const pe_node_t *node);
G_GNUC_INTERNAL
void pcmk__order_restart_vs_unfence(gpointer data, gpointer user_data);
// Injected scheduler inputs (pcmk_sched_injections.c)
void pcmk__inject_scheduler_input(pe_working_set_t *data_set, cib_t *cib,
const pcmk_injections_t *injections);
// Constraints of any type (pcmk_sched_constraints.c)
G_GNUC_INTERNAL
pe_resource_t *pcmk__find_constraint_resource(GList *rsc_list, const char *id);
G_GNUC_INTERNAL
xmlNode *pcmk__expand_tags_in_sets(xmlNode *xml_obj,
const pe_working_set_t *data_set);
G_GNUC_INTERNAL
bool pcmk__valid_resource_or_tag(const pe_working_set_t *data_set,
const char *id, pe_resource_t **rsc,
pe_tag_t **tag);
G_GNUC_INTERNAL
bool pcmk__tag_to_set(xmlNode *xml_obj, xmlNode **rsc_set, const char *attr,
bool convert_rsc, const pe_working_set_t *data_set);
G_GNUC_INTERNAL
void pcmk__create_internal_constraints(pe_working_set_t *data_set);
// Location constraints
G_GNUC_INTERNAL
void pcmk__unpack_location(xmlNode *xml_obj, pe_working_set_t *data_set);
G_GNUC_INTERNAL
pe__location_t *pcmk__new_location(const char *id, pe_resource_t *rsc,
int node_score, const char *discover_mode,
pe_node_t *foo_node);
G_GNUC_INTERNAL
void pcmk__apply_locations(pe_working_set_t *data_set);
G_GNUC_INTERNAL
void pcmk__apply_location(pe_resource_t *rsc, pe__location_t *constraint);
// Colocation constraints (pcmk_sched_colocation.c)
enum pcmk__coloc_affects {
pcmk__coloc_affects_nothing = 0,
pcmk__coloc_affects_location,
pcmk__coloc_affects_role,
};
G_GNUC_INTERNAL
enum pcmk__coloc_affects pcmk__colocation_affects(const pe_resource_t
*dependent,
const pe_resource_t *primary,
const pcmk__colocation_t
*colocation,
bool preview);
G_GNUC_INTERNAL
void pcmk__apply_coloc_to_scores(pe_resource_t *dependent,
const pe_resource_t *primary,
const pcmk__colocation_t *colocation);
G_GNUC_INTERNAL
void pcmk__apply_coloc_to_priority(pe_resource_t *dependent,
const pe_resource_t *primary,
const pcmk__colocation_t *colocation);
G_GNUC_INTERNAL
void pcmk__add_colocated_node_scores(pe_resource_t *rsc, const char *log_id,
GHashTable **nodes,
pcmk__colocation_t *colocation,
float factor, uint32_t flags);
G_GNUC_INTERNAL
void pcmk__add_dependent_scores(gpointer data, gpointer user_data);
G_GNUC_INTERNAL
void pcmk__unpack_colocation(xmlNode *xml_obj, pe_working_set_t *data_set);
G_GNUC_INTERNAL
void pcmk__add_this_with(GList **list, const pcmk__colocation_t *colocation,
const pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__add_this_with_list(GList **list, GList *addition,
const pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__add_with_this(GList **list, const pcmk__colocation_t *colocation,
const pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__add_with_this_list(GList **list, GList *addition,
const pe_resource_t *rsc);
G_GNUC_INTERNAL
GList *pcmk__with_this_colocations(const pe_resource_t *rsc);
G_GNUC_INTERNAL
GList *pcmk__this_with_colocations(const pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__new_colocation(const char *id, const char *node_attr, int score,
pe_resource_t *dependent, pe_resource_t *primary,
const char *dependent_role, const char *primary_role,
uint32_t flags);
G_GNUC_INTERNAL
void pcmk__block_colocation_dependents(pe_action_t *action);
/*!
* \internal
* \brief Check whether colocation's dependent preferences should be considered
*
* \param[in] colocation Colocation constraint
* \param[in] rsc Primary instance (normally this will be
* colocation->primary, which NULL will be treated as,
* but for clones or bundles with multiple instances
* this can be a particular instance)
*
* \return true if colocation influence should be effective, otherwise false
*/
static inline bool
pcmk__colocation_has_influence(const pcmk__colocation_t *colocation,
const pe_resource_t *rsc)
{
if (rsc == NULL) {
rsc = colocation->primary;
}
/* A bundle replica colocates its remote connection with its container,
* using a finite score so that the container can run on Pacemaker Remote
* nodes.
*
* Moving a connection is lightweight and does not interrupt the service,
* while moving a container is heavyweight and does interrupt the service,
* so don't move a clean, active container based solely on the preferences
* of its connection.
*
* This also avoids problematic scenarios where two containers want to
* perpetually swap places.
*/
if (pcmk_is_set(colocation->dependent->flags, pe_rsc_allow_remote_remotes)
&& !pcmk_is_set(rsc->flags, pe_rsc_failed)
&& pcmk__list_of_1(rsc->running_on)) {
return false;
}
/* The dependent in a colocation influences the primary's location
* if the influence option is true or the primary is not yet active.
*/
return pcmk_is_set(colocation->flags, pcmk__coloc_influence)
|| (rsc->running_on == NULL);
}
// Ordering constraints (pcmk_sched_ordering.c)
G_GNUC_INTERNAL
void pcmk__new_ordering(pe_resource_t *first_rsc, char *first_task,
pe_action_t *first_action, pe_resource_t *then_rsc,
char *then_task, pe_action_t *then_action,
uint32_t flags, pe_working_set_t *sched);
G_GNUC_INTERNAL
void pcmk__unpack_ordering(xmlNode *xml_obj, pe_working_set_t *data_set);
G_GNUC_INTERNAL
void pcmk__disable_invalid_orderings(pe_working_set_t *data_set);
G_GNUC_INTERNAL
void pcmk__order_stops_before_shutdown(pe_node_t *node,
pe_action_t *shutdown_op);
G_GNUC_INTERNAL
void pcmk__apply_orderings(pe_working_set_t *sched);
G_GNUC_INTERNAL
void pcmk__order_after_each(pe_action_t *after, GList *list);
/*!
* \internal
* \brief Create a new ordering between two resource actions
*
* \param[in,out] first_rsc Resource for 'first' action
* \param[in,out] first_task Action key for 'first' action
* \param[in] then_rsc Resource for 'then' action
* \param[in,out] then_task Action key for 'then' action
* \param[in] flags Bitmask of enum pe_ordering flags
*/
#define pcmk__order_resource_actions(first_rsc, first_task, \
then_rsc, then_task, flags) \
pcmk__new_ordering((first_rsc), \
pcmk__op_key((first_rsc)->id, (first_task), 0), \
NULL, \
(then_rsc), \
pcmk__op_key((then_rsc)->id, (then_task), 0), \
NULL, (flags), (first_rsc)->cluster)
#define pcmk__order_starts(rsc1, rsc2, flags) \
pcmk__order_resource_actions((rsc1), CRMD_ACTION_START, \
(rsc2), CRMD_ACTION_START, (flags))
#define pcmk__order_stops(rsc1, rsc2, flags) \
pcmk__order_resource_actions((rsc1), CRMD_ACTION_STOP, \
(rsc2), CRMD_ACTION_STOP, (flags))
// Ticket constraints (pcmk_sched_tickets.c)
G_GNUC_INTERNAL
void pcmk__unpack_rsc_ticket(xmlNode *xml_obj, pe_working_set_t *data_set);
// Promotable clone resources (pcmk_sched_promotable.c)
G_GNUC_INTERNAL
void pcmk__add_promotion_scores(pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__require_promotion_tickets(pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__set_instance_roles(pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__create_promotable_actions(pe_resource_t *clone);
G_GNUC_INTERNAL
void pcmk__promotable_restart_ordering(pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__order_promotable_instances(pe_resource_t *clone);
G_GNUC_INTERNAL
void pcmk__update_dependent_with_promotable(const pe_resource_t *primary,
pe_resource_t *dependent,
const pcmk__colocation_t
*colocation);
G_GNUC_INTERNAL
void pcmk__update_promotable_dependent_priority(const pe_resource_t *primary,
pe_resource_t *dependent,
const pcmk__colocation_t
*colocation);
// Pacemaker Remote nodes (pcmk_sched_remote.c)
G_GNUC_INTERNAL
bool pcmk__is_failed_remote_node(const pe_node_t *node);
G_GNUC_INTERNAL
void pcmk__order_remote_connection_actions(pe_working_set_t *data_set);
G_GNUC_INTERNAL
bool pcmk__rsc_corresponds_to_guest(const pe_resource_t *rsc,
const pe_node_t *node);
G_GNUC_INTERNAL
pe_node_t *pcmk__connection_host_for_action(const pe_action_t *action);
G_GNUC_INTERNAL
void pcmk__substitute_remote_addr(pe_resource_t *rsc, GHashTable *params);
G_GNUC_INTERNAL
void pcmk__add_bundle_meta_to_xml(xmlNode *args_xml, const pe_action_t *action);
// Primitives (pcmk_sched_primitive.c)
G_GNUC_INTERNAL
pe_node_t *pcmk__primitive_assign(pe_resource_t *rsc, const pe_node_t *prefer,
bool stop_if_fail);
G_GNUC_INTERNAL
void pcmk__primitive_create_actions(pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__primitive_internal_constraints(pe_resource_t *rsc);
G_GNUC_INTERNAL
uint32_t pcmk__primitive_action_flags(pe_action_t *action,
const pe_node_t *node);
G_GNUC_INTERNAL
void pcmk__primitive_apply_coloc_score(pe_resource_t *dependent,
const pe_resource_t *primary,
const pcmk__colocation_t *colocation,
bool for_dependent);
G_GNUC_INTERNAL
void pcmk__with_primitive_colocations(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc,
GList **list);
G_GNUC_INTERNAL
void pcmk__primitive_with_colocations(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc,
GList **list);
G_GNUC_INTERNAL
void pcmk__schedule_cleanup(pe_resource_t *rsc, const pe_node_t *node,
bool optional);
G_GNUC_INTERNAL
void pcmk__primitive_add_graph_meta(const pe_resource_t *rsc, xmlNode *xml);
G_GNUC_INTERNAL
void pcmk__primitive_add_utilization(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc,
GList *all_rscs, GHashTable *utilization);
G_GNUC_INTERNAL
void pcmk__primitive_shutdown_lock(pe_resource_t *rsc);
// Groups (pcmk_sched_group.c)
G_GNUC_INTERNAL
pe_node_t *pcmk__group_assign(pe_resource_t *rsc, const pe_node_t *prefer,
bool stop_if_fail);
G_GNUC_INTERNAL
void pcmk__group_create_actions(pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__group_internal_constraints(pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__group_apply_coloc_score(pe_resource_t *dependent,
const pe_resource_t *primary,
const pcmk__colocation_t *colocation,
bool for_dependent);
G_GNUC_INTERNAL
void pcmk__with_group_colocations(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc, GList **list);
G_GNUC_INTERNAL
void pcmk__group_with_colocations(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc, GList **list);
G_GNUC_INTERNAL
void pcmk__group_add_colocated_node_scores(pe_resource_t *rsc,
const char *log_id,
GHashTable **nodes,
pcmk__colocation_t *colocation,
float factor, uint32_t flags);
G_GNUC_INTERNAL
void pcmk__group_apply_location(pe_resource_t *rsc, pe__location_t *location);
G_GNUC_INTERNAL
uint32_t pcmk__group_action_flags(pe_action_t *action, const pe_node_t *node);
G_GNUC_INTERNAL
uint32_t pcmk__group_update_ordered_actions(pe_action_t *first,
pe_action_t *then,
const pe_node_t *node,
uint32_t flags, uint32_t filter,
uint32_t type,
pe_working_set_t *data_set);
G_GNUC_INTERNAL
GList *pcmk__group_colocated_resources(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc,
GList *colocated_rscs);
G_GNUC_INTERNAL
void pcmk__group_add_utilization(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc, GList *all_rscs,
GHashTable *utilization);
G_GNUC_INTERNAL
void pcmk__group_shutdown_lock(pe_resource_t *rsc);
// Clones (pcmk_sched_clone.c)
G_GNUC_INTERNAL
pe_node_t *pcmk__clone_assign(pe_resource_t *rsc, const pe_node_t *prefer,
bool stop_if_fail);
G_GNUC_INTERNAL
void pcmk__clone_create_actions(pe_resource_t *rsc);
G_GNUC_INTERNAL
bool pcmk__clone_create_probe(pe_resource_t *rsc, pe_node_t *node);
G_GNUC_INTERNAL
void pcmk__clone_internal_constraints(pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__clone_apply_coloc_score(pe_resource_t *dependent,
const pe_resource_t *primary,
const pcmk__colocation_t *colocation,
bool for_dependent);
G_GNUC_INTERNAL
void pcmk__with_clone_colocations(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc, GList **list);
G_GNUC_INTERNAL
void pcmk__clone_with_colocations(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc, GList **list);
G_GNUC_INTERNAL
void pcmk__clone_apply_location(pe_resource_t *rsc, pe__location_t *constraint);
G_GNUC_INTERNAL
uint32_t pcmk__clone_action_flags(pe_action_t *action, const pe_node_t *node);
G_GNUC_INTERNAL
void pcmk__clone_add_actions_to_graph(pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__clone_add_graph_meta(const pe_resource_t *rsc, xmlNode *xml);
G_GNUC_INTERNAL
void pcmk__clone_add_utilization(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc,
GList *all_rscs, GHashTable *utilization);
G_GNUC_INTERNAL
void pcmk__clone_shutdown_lock(pe_resource_t *rsc);
// Bundles (pcmk_sched_bundle.c)
G_GNUC_INTERNAL
pe_node_t *pcmk__bundle_assign(pe_resource_t *rsc, const pe_node_t *prefer,
bool stop_if_fail);
G_GNUC_INTERNAL
void pcmk__bundle_create_actions(pe_resource_t *rsc);
G_GNUC_INTERNAL
bool pcmk__bundle_create_probe(pe_resource_t *rsc, pe_node_t *node);
G_GNUC_INTERNAL
void pcmk__bundle_internal_constraints(pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__bundle_apply_coloc_score(pe_resource_t *dependent,
const pe_resource_t *primary,
const pcmk__colocation_t *colocation,
bool for_dependent);
G_GNUC_INTERNAL
void pcmk__with_bundle_colocations(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc, GList **list);
G_GNUC_INTERNAL
void pcmk__bundle_with_colocations(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc, GList **list);
G_GNUC_INTERNAL
void pcmk__bundle_apply_location(pe_resource_t *rsc,
pe__location_t *constraint);
G_GNUC_INTERNAL
uint32_t pcmk__bundle_action_flags(pe_action_t *action, const pe_node_t *node);
G_GNUC_INTERNAL
void pcmk__output_bundle_actions(pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__bundle_add_actions_to_graph(pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__bundle_add_utilization(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc,
GList *all_rscs, GHashTable *utilization);
G_GNUC_INTERNAL
void pcmk__bundle_shutdown_lock(pe_resource_t *rsc);
// Clone instances or bundle replica containers (pcmk_sched_instances.c)
G_GNUC_INTERNAL
void pcmk__assign_instances(pe_resource_t *collective, GList *instances,
int max_total, int max_per_node);
G_GNUC_INTERNAL
void pcmk__create_instance_actions(pe_resource_t *rsc, GList *instances);
G_GNUC_INTERNAL
bool pcmk__instance_matches(const pe_resource_t *instance,
const pe_node_t *node, enum rsc_role_e role,
bool current);
G_GNUC_INTERNAL
pe_resource_t *pcmk__find_compatible_instance(const pe_resource_t *match_rsc,
const pe_resource_t *rsc,
enum rsc_role_e role,
bool current);
G_GNUC_INTERNAL
uint32_t pcmk__instance_update_ordered_actions(pe_action_t *first,
pe_action_t *then,
const pe_node_t *node,
uint32_t flags, uint32_t filter,
uint32_t type,
pe_working_set_t *data_set);
G_GNUC_INTERNAL
uint32_t pcmk__collective_action_flags(pe_action_t *action,
const GList *instances,
const pe_node_t *node);
-G_GNUC_INTERNAL
-void pcmk__add_collective_constraints(GList **list,
- const pe_resource_t *instance,
- const pe_resource_t *collective,
- bool with_this);
-
// Injections (pcmk_injections.c)
G_GNUC_INTERNAL
xmlNode *pcmk__inject_node(cib_t *cib_conn, const char *node, const char *uuid);
G_GNUC_INTERNAL
xmlNode *pcmk__inject_node_state_change(cib_t *cib_conn, const char *node,
bool up);
G_GNUC_INTERNAL
xmlNode *pcmk__inject_resource_history(pcmk__output_t *out, xmlNode *cib_node,
const char *resource,
const char *lrm_name,
const char *rclass,
const char *rtype,
const char *rprovider);
G_GNUC_INTERNAL
void pcmk__inject_failcount(pcmk__output_t *out, xmlNode *cib_node,
const char *resource, const char *task,
guint interval_ms, int rc);
G_GNUC_INTERNAL
xmlNode *pcmk__inject_action_result(xmlNode *cib_resource,
lrmd_event_data_t *op, int target_rc);
// Nodes (pcmk_sched_nodes.c)
G_GNUC_INTERNAL
bool pcmk__node_available(const pe_node_t *node, bool consider_score,
bool consider_guest);
G_GNUC_INTERNAL
bool pcmk__any_node_available(GHashTable *nodes);
G_GNUC_INTERNAL
GHashTable *pcmk__copy_node_table(GHashTable *nodes);
G_GNUC_INTERNAL
void pcmk__copy_node_tables(const pe_resource_t *rsc, GHashTable **copy);
G_GNUC_INTERNAL
void pcmk__restore_node_tables(pe_resource_t *rsc, GHashTable *backup);
G_GNUC_INTERNAL
GList *pcmk__sort_nodes(GList *nodes, pe_node_t *active_node);
G_GNUC_INTERNAL
void pcmk__apply_node_health(pe_working_set_t *data_set);
G_GNUC_INTERNAL
pe_node_t *pcmk__top_allowed_node(const pe_resource_t *rsc,
const pe_node_t *node);
// Functions applying to more than one variant (pcmk_sched_resource.c)
G_GNUC_INTERNAL
void pcmk__set_assignment_methods(pe_working_set_t *data_set);
G_GNUC_INTERNAL
bool pcmk__rsc_agent_changed(pe_resource_t *rsc, pe_node_t *node,
const xmlNode *rsc_entry, bool active_on_node);
G_GNUC_INTERNAL
GList *pcmk__rscs_matching_id(const char *id, const pe_working_set_t *data_set);
G_GNUC_INTERNAL
GList *pcmk__colocated_resources(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc,
GList *colocated_rscs);
G_GNUC_INTERNAL
void pcmk__noop_add_graph_meta(const pe_resource_t *rsc, xmlNode *xml);
G_GNUC_INTERNAL
void pcmk__output_resource_actions(pe_resource_t *rsc);
G_GNUC_INTERNAL
bool pcmk__assign_resource(pe_resource_t *rsc, pe_node_t *node, bool force,
bool stop_if_fail);
G_GNUC_INTERNAL
void pcmk__unassign_resource(pe_resource_t *rsc);
G_GNUC_INTERNAL
bool pcmk__threshold_reached(pe_resource_t *rsc, const pe_node_t *node,
pe_resource_t **failed);
G_GNUC_INTERNAL
void pcmk__sort_resources(pe_working_set_t *data_set);
G_GNUC_INTERNAL
gint pcmk__cmp_instance(gconstpointer a, gconstpointer b);
G_GNUC_INTERNAL
gint pcmk__cmp_instance_number(gconstpointer a, gconstpointer b);
// Functions related to probes (pcmk_sched_probes.c)
G_GNUC_INTERNAL
bool pcmk__probe_rsc_on_node(pe_resource_t *rsc, pe_node_t *node);
G_GNUC_INTERNAL
void pcmk__order_probes(pe_working_set_t *data_set);
G_GNUC_INTERNAL
bool pcmk__probe_resource_list(GList *rscs, pe_node_t *node);
G_GNUC_INTERNAL
void pcmk__schedule_probes(pe_working_set_t *data_set);
// Functions related to live migration (pcmk_sched_migration.c)
void pcmk__create_migration_actions(pe_resource_t *rsc,
const pe_node_t *current);
void pcmk__abort_dangling_migration(void *data, void *user_data);
bool pcmk__rsc_can_migrate(const pe_resource_t *rsc, const pe_node_t *current);
void pcmk__order_migration_equivalents(pe__ordering_t *order);
// Functions related to node utilization (pcmk_sched_utilization.c)
G_GNUC_INTERNAL
int pcmk__compare_node_capacities(const pe_node_t *node1,
const pe_node_t *node2);
G_GNUC_INTERNAL
void pcmk__consume_node_capacity(GHashTable *current_utilization,
const pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__release_node_capacity(GHashTable *current_utilization,
const pe_resource_t *rsc);
G_GNUC_INTERNAL
const pe_node_t *pcmk__ban_insufficient_capacity(pe_resource_t *rsc);
G_GNUC_INTERNAL
void pcmk__create_utilization_constraints(pe_resource_t *rsc,
const GList *allowed_nodes);
G_GNUC_INTERNAL
void pcmk__show_node_capacities(const char *desc, pe_working_set_t *data_set);
#endif // PCMK__LIBPACEMAKER_PRIVATE__H
diff --git a/lib/pacemaker/pcmk_sched_bundle.c b/lib/pacemaker/pcmk_sched_bundle.c
index 1bfbb04155..fc6ae8c487 100644
--- a/lib/pacemaker/pcmk_sched_bundle.c
+++ b/lib/pacemaker/pcmk_sched_bundle.c
@@ -1,963 +1,961 @@
/*
* Copyright 2004-2023 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 <crm_internal.h>
#include <stdbool.h>
#include <crm/msg_xml.h>
#include <pacemaker-internal.h>
#include "libpacemaker_private.h"
struct assign_data {
const pe_node_t *prefer;
bool stop_if_fail;
};
/*!
* \internal
* \brief Assign a single bundle replica's resources (other than container)
*
* \param[in,out] replica Replica to assign
* \param[in] user_data Preferred node, if any
*
* \return true (to indicate that any further replicas should be processed)
*/
static bool
assign_replica(pe__bundle_replica_t *replica, void *user_data)
{
pe_node_t *container_host = NULL;
struct assign_data *assign_data = user_data;
const pe_node_t *prefer = assign_data->prefer;
bool stop_if_fail = assign_data->stop_if_fail;
const pe_resource_t *bundle = pe__const_top_resource(replica->container,
true);
if (replica->ip != NULL) {
pe_rsc_trace(bundle, "Assigning bundle %s IP %s",
bundle->id, replica->ip->id);
replica->ip->cmds->assign(replica->ip, prefer, stop_if_fail);
}
container_host = replica->container->allocated_to;
if (replica->remote != NULL) {
if (pe__is_guest_or_remote_node(container_host)) {
/* REMOTE_CONTAINER_HACK: "Nested" connection resources must be on
* the same host because Pacemaker Remote only supports a single
* active connection.
*/
pcmk__new_colocation("#replica-remote-with-host-remote", NULL,
INFINITY, replica->remote,
container_host->details->remote_rsc, NULL,
NULL, pcmk__coloc_influence);
}
pe_rsc_trace(bundle, "Assigning bundle %s connection %s",
bundle->id, replica->remote->id);
replica->remote->cmds->assign(replica->remote, prefer, stop_if_fail);
}
if (replica->child != NULL) {
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 (!pe__same_node(node, replica->node)) {
node->weight = -INFINITY;
} else if (!pcmk__threshold_reached(replica->child, node, NULL)) {
node->weight = INFINITY;
}
}
pe__set_resource_flags(replica->child->parent, pe_rsc_allocating);
pe_rsc_trace(bundle, "Assigning bundle %s replica child %s",
bundle->id, replica->child->id);
replica->child->cmds->assign(replica->child, replica->node,
stop_if_fail);
pe__clear_resource_flags(replica->child->parent, pe_rsc_allocating);
}
return true;
}
/*!
* \internal
* \brief Assign a bundle resource to a node
*
* \param[in,out] rsc Resource to assign to a node
* \param[in] prefer Node to prefer, if all else is equal
* \param[in] stop_if_fail If \c true and a primitive descendant of \p rsc
* can't be assigned to a node, set the
* descendant's next role to stopped and update
* existing actions
*
* \return Node that \p rsc is assigned to, if assigned entirely to one node
*
* \note If \p stop_if_fail is \c false, then \c pcmk__unassign_resource() can
* completely undo the assignment. A successful assignment can be either
* undone or left alone as final. A failed assignment has the same effect
* as calling pcmk__unassign_resource(); there are no side effects on
* roles or actions.
*/
pe_node_t *
pcmk__bundle_assign(pe_resource_t *rsc, const pe_node_t *prefer,
bool stop_if_fail)
{
GList *containers = NULL;
pe_resource_t *bundled_resource = NULL;
struct assign_data assign_data = { prefer, stop_if_fail };
CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container));
pe_rsc_trace(rsc, "Assigning bundle %s", rsc->id);
pe__set_resource_flags(rsc, pe_rsc_allocating);
pe__show_node_scores(!pcmk_is_set(rsc->cluster->flags, pe_flag_show_scores),
rsc, __func__, rsc->allowed_nodes, rsc->cluster);
// Assign all containers first, so we know what nodes the bundle will be on
containers = g_list_sort(pe__bundle_containers(rsc), pcmk__cmp_instance);
pcmk__assign_instances(rsc, containers, pe__bundle_max(rsc),
rsc->fns->max_per_node(rsc));
g_list_free(containers);
// Then assign remaining replica resources
pe__foreach_bundle_replica(rsc, assign_replica, (void *) &assign_data);
// Finally, assign the bundled resources to each bundle node
bundled_resource = pe__bundled_resource(rsc);
if (bundled_resource != NULL) {
pe_node_t *node = NULL;
GHashTableIter iter;
g_hash_table_iter_init(&iter, bundled_resource->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (gpointer *) & node)) {
if (pe__node_is_bundle_instance(rsc, node)) {
node->weight = 0;
} else {
node->weight = -INFINITY;
}
}
bundled_resource->cmds->assign(bundled_resource, prefer, stop_if_fail);
}
pe__clear_resource_flags(rsc, pe_rsc_allocating|pe_rsc_provisional);
return NULL;
}
/*!
* \internal
* \brief Create actions for a bundle replica's resources (other than child)
*
* \param[in,out] replica Replica to create actions for
* \param[in] user_data Unused
*
* \return true (to indicate that any further replicas should be processed)
*/
static bool
create_replica_actions(pe__bundle_replica_t *replica, void *user_data)
{
if (replica->ip != NULL) {
replica->ip->cmds->create_actions(replica->ip);
}
if (replica->container != NULL) {
replica->container->cmds->create_actions(replica->container);
}
if (replica->remote != NULL) {
replica->remote->cmds->create_actions(replica->remote);
}
return true;
}
/*!
* \internal
* \brief Create all actions needed for a given bundle resource
*
* \param[in,out] rsc Bundle resource to create actions for
*/
void
pcmk__bundle_create_actions(pe_resource_t *rsc)
{
pe_action_t *action = NULL;
GList *containers = NULL;
pe_resource_t *bundled_resource = NULL;
CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container));
pe__foreach_bundle_replica(rsc, create_replica_actions, NULL);
containers = pe__bundle_containers(rsc);
pcmk__create_instance_actions(rsc, containers);
g_list_free(containers);
bundled_resource = pe__bundled_resource(rsc);
if (bundled_resource != NULL) {
bundled_resource->cmds->create_actions(bundled_resource);
if (pcmk_is_set(bundled_resource->flags, pe_rsc_promotable)) {
pe__new_rsc_pseudo_action(rsc, RSC_PROMOTE, true, true);
action = pe__new_rsc_pseudo_action(rsc, RSC_PROMOTED, true, true);
action->priority = INFINITY;
pe__new_rsc_pseudo_action(rsc, RSC_DEMOTE, true, true);
action = pe__new_rsc_pseudo_action(rsc, RSC_DEMOTED, true, true);
action->priority = INFINITY;
}
}
}
/*!
* \internal
* \brief Create internal constraints for a bundle replica's resources
*
* \param[in,out] replica Replica to create internal constraints for
* \param[in,out] user_data Replica's parent bundle
*
* \return true (to indicate that any further replicas should be processed)
*/
static bool
replica_internal_constraints(pe__bundle_replica_t *replica, void *user_data)
{
pe_resource_t *bundle = user_data;
replica->container->cmds->internal_constraints(replica->container);
// Start bundle -> start replica container
pcmk__order_starts(bundle, replica->container,
pe_order_runnable_left|pe_order_implies_first_printed);
// Stop bundle -> stop replica child and container
if (replica->child != NULL) {
pcmk__order_stops(bundle, replica->child,
pe_order_implies_first_printed);
}
pcmk__order_stops(bundle, replica->container,
pe_order_implies_first_printed);
// Start replica container -> bundle is started
pcmk__order_resource_actions(replica->container, RSC_START, bundle,
RSC_STARTED,
pe_order_implies_then_printed);
// Stop replica container -> bundle is stopped
pcmk__order_resource_actions(replica->container, RSC_STOP, bundle,
RSC_STOPPED,
pe_order_implies_then_printed);
if (replica->ip != NULL) {
replica->ip->cmds->internal_constraints(replica->ip);
// Replica IP address -> replica container (symmetric)
pcmk__order_starts(replica->ip, replica->container,
pe_order_runnable_left|pe_order_preserve);
pcmk__order_stops(replica->container, replica->ip,
pe_order_implies_first|pe_order_preserve);
pcmk__new_colocation("#ip-with-container", NULL, INFINITY, replica->ip,
replica->container, NULL, NULL,
pcmk__coloc_influence);
}
if (replica->remote != NULL) {
/* 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);
}
if (replica->child != NULL) {
CRM_ASSERT(replica->remote != NULL);
// "Start remote then child" is implicit in scheduler's remote logic
}
return true;
}
/*!
* \internal
* \brief Create implicit constraints needed for a bundle resource
*
* \param[in,out] rsc Bundle resource to create implicit constraints for
*/
void
pcmk__bundle_internal_constraints(pe_resource_t *rsc)
{
pe_resource_t *bundled_resource = NULL;
CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container));
pe__foreach_bundle_replica(rsc, replica_internal_constraints, rsc);
bundled_resource = pe__bundled_resource(rsc);
if (bundled_resource == NULL) {
return;
}
// Start bundle -> start bundled clone
pcmk__order_resource_actions(rsc, RSC_START, bundled_resource,
RSC_START, pe_order_implies_first_printed);
// Bundled clone is started -> bundle is started
pcmk__order_resource_actions(bundled_resource, RSC_STARTED,
rsc, RSC_STARTED,
pe_order_implies_then_printed);
// Stop bundle -> stop bundled clone
pcmk__order_resource_actions(rsc, RSC_STOP, bundled_resource, RSC_STOP,
pe_order_implies_first_printed);
// Bundled clone is stopped -> bundle is stopped
pcmk__order_resource_actions(bundled_resource, RSC_STOPPED,
rsc, RSC_STOPPED,
pe_order_implies_then_printed);
bundled_resource->cmds->internal_constraints(bundled_resource);
if (!pcmk_is_set(bundled_resource->flags, pe_rsc_promotable)) {
return;
}
pcmk__promotable_restart_ordering(rsc);
// Demote bundle -> demote bundled clone
pcmk__order_resource_actions(rsc, RSC_DEMOTE, bundled_resource, RSC_DEMOTE,
pe_order_implies_first_printed);
// Bundled clone is demoted -> bundle is demoted
pcmk__order_resource_actions(bundled_resource, RSC_DEMOTED,
rsc, RSC_DEMOTED,
pe_order_implies_then_printed);
// Promote bundle -> promote bundled clone
pcmk__order_resource_actions(rsc, RSC_PROMOTE,
bundled_resource, RSC_PROMOTE,
pe_order_implies_first_printed);
// Bundled clone is promoted -> bundle is promoted
pcmk__order_resource_actions(bundled_resource, RSC_PROMOTED,
rsc, RSC_PROMOTED,
pe_order_implies_then_printed);
}
struct match_data {
const pe_node_t *node; // Node to compare against replica
pe_resource_t *container; // Replica container corresponding to node
};
/*!
* \internal
* \brief Check whether a replica container is assigned to a given node
*
* \param[in] replica Replica to check
* \param[in,out] user_data struct match_data with node to compare against
*
* \return true if the replica does not match (to indicate further replicas
* should be processed), otherwise false
*/
static bool
match_replica_container(const pe__bundle_replica_t *replica, void *user_data)
{
struct match_data *match_data = user_data;
if (pcmk__instance_matches(replica->container, match_data->node,
RSC_ROLE_UNKNOWN, false)) {
match_data->container = replica->container;
return false; // Match found, don't bother searching further replicas
}
return true; // No match, keep searching
}
/*!
* \internal
* \brief Find a bundle container compatible with a dependent resource
*
* \param[in] dependent Dependent resource in colocation with bundle
* \param[in] bundle Bundle that \p dependent is colocated with
*
* \return A container from \p bundle assigned to the same node as \p dependent
* if assigned, otherwise assigned to any of dependent's allowed nodes,
* otherwise NULL.
*/
static pe_resource_t *
compatible_container(const pe_resource_t *dependent,
const pe_resource_t *bundle)
{
GList *scratch = NULL;
struct match_data match_data = { NULL, NULL };
// If dependent is assigned, only check there
match_data.node = dependent->fns->location(dependent, NULL, 0);
if (match_data.node != NULL) {
pe__foreach_const_bundle_replica(bundle, match_replica_container,
&match_data);
return match_data.container;
}
// Otherwise, check for any of the dependent's allowed nodes
scratch = g_hash_table_get_values(dependent->allowed_nodes);
scratch = pcmk__sort_nodes(scratch, NULL);
for (const GList *iter = scratch; iter != NULL; iter = iter->next) {
match_data.node = iter->data;
pe__foreach_const_bundle_replica(bundle, match_replica_container,
&match_data);
if (match_data.container != NULL) {
break;
}
}
g_list_free(scratch);
return match_data.container;
}
struct coloc_data {
const pcmk__colocation_t *colocation;
pe_resource_t *dependent;
GList *container_hosts;
};
/*!
* \internal
* \brief Apply a colocation score to replica node scores or resource priority
*
* \param[in] replica Replica of primary bundle resource in colocation
* \param[in,out] user_data struct coloc_data for colocation being applied
*
* \return true (to indicate that any further replicas should be processed)
*/
static bool
replica_apply_coloc_score(const pe__bundle_replica_t *replica, void *user_data)
{
struct coloc_data *coloc_data = user_data;
pe_node_t *chosen = NULL;
if (coloc_data->colocation->score < INFINITY) {
replica->container->cmds->apply_coloc_score(coloc_data->dependent,
replica->container,
coloc_data->colocation,
false);
return true;
}
chosen = replica->container->fns->location(replica->container, NULL, 0);
if ((chosen == NULL)
|| is_set_recursive(replica->container, pe_rsc_block, true)) {
return true;
}
if ((coloc_data->colocation->primary_role >= RSC_ROLE_PROMOTED)
&& ((replica->child == NULL)
|| (replica->child->next_role < RSC_ROLE_PROMOTED))) {
return true;
}
pe_rsc_trace(pe__const_top_resource(replica->container, true),
"Allowing mandatory colocation %s using %s @%d",
coloc_data->colocation->id, pe__node_name(chosen),
chosen->weight);
coloc_data->container_hosts = g_list_prepend(coloc_data->container_hosts,
chosen);
return true;
}
/*!
* \internal
* \brief Apply a colocation's score to node scores or resource priority
*
* Given a colocation constraint, apply its score to the dependent's
* allowed node scores (if we are still placing resources) or priority (if
* we are choosing promotable clone instance roles).
*
* \param[in,out] dependent Dependent resource in colocation
* \param[in] primary Primary resource in colocation
* \param[in] colocation Colocation constraint to apply
* \param[in] for_dependent true if called on behalf of dependent
*/
void
pcmk__bundle_apply_coloc_score(pe_resource_t *dependent,
const pe_resource_t *primary,
const pcmk__colocation_t *colocation,
bool for_dependent)
{
struct coloc_data coloc_data = { colocation, dependent, NULL };
/* This should never be called for the bundle itself as a dependent.
* Instead, we add its colocation constraints to its containers and call the
* apply_coloc_score() method for the containers as dependents.
*/
CRM_ASSERT((primary != NULL) && (primary->variant == pe_container)
&& (dependent != NULL) && (dependent->variant == pe_native)
&& (colocation != NULL) && !for_dependent);
if (pcmk_is_set(primary->flags, pe_rsc_provisional)) {
pe_rsc_trace(primary,
"Skipping applying colocation %s "
"because %s is still provisional",
colocation->id, primary->id);
return;
}
pe_rsc_trace(primary, "Applying colocation %s (%s with %s at %s)",
colocation->id, dependent->id, primary->id,
pcmk_readable_score(colocation->score));
/* If the constraint dependent is a clone or bundle, "dependent" here is one
* of its instances. Look for a compatible instance of this bundle.
*/
if (colocation->dependent->variant > pe_group) {
const pe_resource_t *primary_container = compatible_container(dependent,
primary);
if (primary_container != NULL) { // Success, we found one
pe_rsc_debug(primary, "Pairing %s with %s",
dependent->id, primary_container->id);
dependent->cmds->apply_coloc_score(dependent, primary_container,
colocation, true);
} else if (colocation->score >= INFINITY) { // Failure, and it's fatal
crm_notice("%s cannot run because there is no compatible "
"instance of %s to colocate with",
dependent->id, primary->id);
pcmk__assign_resource(dependent, NULL, true, true);
} else { // Failure, but we can ignore it
pe_rsc_debug(primary,
"%s cannot be colocated with any instance of %s",
dependent->id, primary->id);
}
return;
}
pe__foreach_const_bundle_replica(primary, replica_apply_coloc_score,
&coloc_data);
if (colocation->score >= INFINITY) {
node_list_exclude(dependent->allowed_nodes, coloc_data.container_hosts,
FALSE);
}
g_list_free(coloc_data.container_hosts);
}
// Bundle implementation of resource_alloc_functions_t:with_this_colocations()
void
pcmk__with_bundle_colocations(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc, GList **list)
{
CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container)
&& (orig_rsc != NULL) && (list != NULL));
- if (rsc == orig_rsc) { // Colocations are wanted for bundle itself
- pcmk__add_with_this_list(list, rsc->rsc_cons_lhs, orig_rsc);
+ // Only the bundle itself and its containers get the bundle's constraints
+ if ((orig_rsc == rsc)
+ || pcmk_is_set(orig_rsc->flags, pe_rsc_replica_container)) {
- // Only the bundle replicas' containers get the bundle's constraints
- } else if (pcmk_is_set(orig_rsc->flags, pe_rsc_replica_container)) {
- pcmk__add_collective_constraints(list, orig_rsc, rsc, true);
+ pcmk__add_with_this_list(list, rsc->rsc_cons_lhs, orig_rsc);
}
}
// Bundle implementation of resource_alloc_functions_t:this_with_colocations()
void
pcmk__bundle_with_colocations(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc, GList **list)
{
CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container)
&& (orig_rsc != NULL) && (list != NULL));
- if (rsc == orig_rsc) { // Colocations are wanted for bundle itself
- pcmk__add_this_with_list(list, rsc->rsc_cons, orig_rsc);
+ // Only the bundle itself and its containers get the bundle's constraints
+ if ((orig_rsc == rsc)
+ || pcmk_is_set(orig_rsc->flags, pe_rsc_replica_container)) {
- // Only the bundle replicas' containers get the bundle's constraints
- } else if (pcmk_is_set(orig_rsc->flags, pe_rsc_replica_container)) {
- pcmk__add_collective_constraints(list, orig_rsc, rsc, false);
+ pcmk__add_this_with_list(list, rsc->rsc_cons, orig_rsc);
}
}
/*!
* \internal
* \brief Return action flags for a given bundle resource action
*
* \param[in,out] action Bundle resource action to get flags for
* \param[in] node If not NULL, limit effects to this node
*
* \return Flags appropriate to \p action on \p node
*/
uint32_t
pcmk__bundle_action_flags(pe_action_t *action, const pe_node_t *node)
{
GList *containers = NULL;
uint32_t flags = 0;
pe_resource_t *bundled_resource = NULL;
CRM_ASSERT((action != NULL) && (action->rsc != NULL)
&& (action->rsc->variant == pe_container));
bundled_resource = pe__bundled_resource(action->rsc);
if (bundled_resource != NULL) {
// Clone actions are done on the bundled clone resource, not container
switch (get_complex_task(bundled_resource, action->task)) {
case no_action:
case action_notify:
case action_notified:
case action_promote:
case action_promoted:
case action_demote:
case action_demoted:
return pcmk__collective_action_flags(action,
bundled_resource->children,
node);
default:
break;
}
}
containers = pe__bundle_containers(action->rsc);
flags = pcmk__collective_action_flags(action, containers, node);
g_list_free(containers);
return flags;
}
/*!
* \internal
* \brief Apply a location constraint to a bundle replica
*
* \param[in,out] replica Replica to apply constraint to
* \param[in,out] user_data Location constraint to apply
*
* \return true (to indicate that any further replicas should be processed)
*/
static bool
apply_location_to_replica(pe__bundle_replica_t *replica, void *user_data)
{
pe__location_t *location = user_data;
if (replica->container != NULL) {
replica->container->cmds->apply_location(replica->container, location);
}
if (replica->ip != NULL) {
replica->ip->cmds->apply_location(replica->ip, location);
}
return true;
}
/*!
* \internal
* \brief Apply a location constraint to a bundle resource's allowed node scores
*
* \param[in,out] rsc Bundle resource to apply constraint to
* \param[in,out] location Location constraint to apply
*/
void
pcmk__bundle_apply_location(pe_resource_t *rsc, pe__location_t *location)
{
pe_resource_t *bundled_resource = NULL;
CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container)
&& (location != NULL));
pcmk__apply_location(rsc, location);
pe__foreach_bundle_replica(rsc, apply_location_to_replica, location);
bundled_resource = pe__bundled_resource(rsc);
if ((bundled_resource != NULL)
&& ((location->role_filter == RSC_ROLE_UNPROMOTED)
|| (location->role_filter == RSC_ROLE_PROMOTED))) {
bundled_resource->cmds->apply_location(bundled_resource, location);
bundled_resource->rsc_location = g_list_prepend(
bundled_resource->rsc_location, location);
}
}
#define XPATH_REMOTE "//nvpair[@name='" XML_RSC_ATTR_REMOTE_RA_ADDR "']"
/*!
* \internal
* \brief Add a bundle replica's actions to transition graph
*
* \param[in,out] replica Replica to add to graph
* \param[in] user_data Bundle that replica belongs to (for logging only)
*
* \return true (to indicate that any further replicas should be processed)
*/
static bool
add_replica_actions_to_graph(pe__bundle_replica_t *replica, void *user_data)
{
if ((replica->remote != NULL) && (replica->container != NULL)
&& 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(XPATH_REMOTE, replica->remote->xml,
LOG_ERR);
const char *calculated_addr = NULL;
// Replace the value in replica->remote->xml (if appropriate)
calculated_addr = pe__add_bundle_remote_name(replica->remote,
replica->remote->cluster,
nvpair, "value");
if (calculated_addr != NULL) {
/* Since this is for the bundle as a resource, and not any
* particular action, replace the value in the default
* parameters (not evaluated for node). create_graph_action()
* will grab it from there to replace it in node-evaluated
* parameters.
*/
GHashTable *params = pe_rsc_params(replica->remote,
NULL, replica->remote->cluster);
g_hash_table_replace(params,
strdup(XML_RSC_ATTR_REMOTE_RA_ADDR),
strdup(calculated_addr));
} else {
pe_resource_t *bundle = user_data;
/* 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).
*/
pe_rsc_info(bundle,
"Unable to determine address for bundle %s "
"remote connection", bundle->id);
}
}
if (replica->ip != NULL) {
replica->ip->cmds->add_actions_to_graph(replica->ip);
}
if (replica->container != NULL) {
replica->container->cmds->add_actions_to_graph(replica->container);
}
if (replica->remote != NULL) {
replica->remote->cmds->add_actions_to_graph(replica->remote);
}
return true;
}
/*!
* \internal
* \brief Add a bundle resource's actions to the transition graph
*
* \param[in,out] rsc Bundle resource whose actions should be added
*/
void
pcmk__bundle_add_actions_to_graph(pe_resource_t *rsc)
{
pe_resource_t *bundled_resource = NULL;
CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container));
bundled_resource = pe__bundled_resource(rsc);
if (bundled_resource != NULL) {
bundled_resource->cmds->add_actions_to_graph(bundled_resource);
}
pe__foreach_bundle_replica(rsc, add_replica_actions_to_graph, rsc);
}
struct probe_data {
pe_resource_t *bundle; // Bundle being probed
pe_node_t *node; // Node to create probes on
bool any_created; // Whether any probes have been created
};
/*!
* \internal
* \brief Order a bundle replica's start after another replica's probe
*
* \param[in,out] replica Replica to order start for
* \param[in,out] user_data Replica with probe to order after
*
* \return true (to indicate that any further replicas should be processed)
*/
static bool
order_replica_start_after(pe__bundle_replica_t *replica, void *user_data)
{
pe__bundle_replica_t *probed_replica = user_data;
if ((replica == probed_replica) || (replica->container == NULL)) {
return true;
}
pcmk__new_ordering(probed_replica->container,
pcmk__op_key(probed_replica->container->id, RSC_STATUS,
0),
NULL, replica->container,
pcmk__op_key(replica->container->id, RSC_START, 0), NULL,
pe_order_optional|pe_order_same_node,
replica->container->cluster);
return true;
}
/*!
* \internal
* \brief Create probes for a bundle replica's resources
*
* \param[in,out] replica Replica to create probes for
* \param[in,out] user_data struct probe_data
*
* \return true (to indicate that any further replicas should be processed)
*/
static bool
create_replica_probes(pe__bundle_replica_t *replica, void *user_data)
{
struct probe_data *probe_data = user_data;
if ((replica->ip != NULL)
&& replica->ip->cmds->create_probe(replica->ip, probe_data->node)) {
probe_data->any_created = true;
}
if ((replica->child != NULL)
&& pe__same_node(probe_data->node, replica->node)
&& replica->child->cmds->create_probe(replica->child,
probe_data->node)) {
probe_data->any_created = true;
}
if ((replica->container != NULL)
&& replica->container->cmds->create_probe(replica->container,
probe_data->node)) {
probe_data->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 the maximum replicas 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
*/
if (probe_data->bundle->fns->max_per_node(probe_data->bundle) == 1) {
pe__foreach_bundle_replica(probe_data->bundle,
order_replica_start_after, replica);
}
}
if ((replica->container != NULL) && (replica->remote != NULL)
&& replica->remote->cmds->create_probe(replica->remote,
probe_data->node)) {
/* 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,
probe_data->node);
free(probe_uuid);
if (probe != NULL) {
probe_data->any_created = true;
pe_rsc_trace(probe_data->bundle, "Ordering %s probe on %s",
replica->remote->id, pe__node_name(probe_data->node));
pcmk__new_ordering(replica->container,
pcmk__op_key(replica->container->id, RSC_START,
0),
NULL, replica->remote, NULL, probe,
pe_order_probe, probe_data->bundle->cluster);
}
}
return true;
}
/*!
* \internal
*
* \brief Schedule any probes needed for a bundle resource on a node
*
* \param[in,out] rsc Bundle resource to create probes for
* \param[in,out] node Node to create probe on
*
* \return true if any probe was created, otherwise false
*/
bool
pcmk__bundle_create_probe(pe_resource_t *rsc, pe_node_t *node)
{
struct probe_data probe_data = { rsc, node, false };
CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container));
pe__foreach_bundle_replica(rsc, create_replica_probes, &probe_data);
return probe_data.any_created;
}
/*!
* \internal
* \brief Output actions for one bundle replica
*
* \param[in,out] replica Replica to output actions for
* \param[in] user_data Unused
*
* \return true (to indicate that any further replicas should be processed)
*/
static bool
output_replica_actions(pe__bundle_replica_t *replica, void *user_data)
{
if (replica->ip != NULL) {
replica->ip->cmds->output_actions(replica->ip);
}
if (replica->container != NULL) {
replica->container->cmds->output_actions(replica->container);
}
if (replica->remote != NULL) {
replica->remote->cmds->output_actions(replica->remote);
}
if (replica->child != NULL) {
replica->child->cmds->output_actions(replica->child);
}
return true;
}
/*!
* \internal
* \brief Output a summary of scheduled actions for a bundle resource
*
* \param[in,out] rsc Bundle resource to output actions for
*/
void
pcmk__output_bundle_actions(pe_resource_t *rsc)
{
CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container));
pe__foreach_bundle_replica(rsc, output_replica_actions, NULL);
}
// Bundle implementation of resource_alloc_functions_t:add_utilization()
void
pcmk__bundle_add_utilization(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc, GList *all_rscs,
GHashTable *utilization)
{
pe_resource_t *container = NULL;
CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container));
if (!pcmk_is_set(rsc->flags, pe_rsc_provisional)) {
return;
}
/* All bundle replicas are identical, so using the utilization of the first
* is sufficient for any. Only the implicit container resource can have
* utilization values.
*/
container = pe__first_container(rsc);
if (container != NULL) {
container->cmds->add_utilization(container, orig_rsc, all_rscs,
utilization);
}
}
// Bundle implementation of resource_alloc_functions_t:shutdown_lock()
void
pcmk__bundle_shutdown_lock(pe_resource_t *rsc)
{
CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container));
// Bundles currently don't support shutdown locks
}
diff --git a/lib/pacemaker/pcmk_sched_clone.c b/lib/pacemaker/pcmk_sched_clone.c
index 4f5afecfb0..5633f42b27 100644
--- a/lib/pacemaker/pcmk_sched_clone.c
+++ b/lib/pacemaker/pcmk_sched_clone.c
@@ -1,699 +1,691 @@
/*
* Copyright 2004-2023 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 <crm_internal.h>
#include <crm/msg_xml.h>
#include <pacemaker-internal.h>
#include "libpacemaker_private.h"
/*!
* \internal
* \brief Assign a clone resource's instances to nodes
*
* \param[in,out] rsc Clone resource to assign
* \param[in] prefer Node to prefer, if all else is equal
* \param[in] stop_if_fail If \c true and a primitive descendant of \p rsc
* can't be assigned to a node, set the
* descendant's next role to stopped and update
* existing actions
*
* \return NULL (clones are not assigned to a single node)
*
* \note If \p stop_if_fail is \c false, then \c pcmk__unassign_resource() can
* completely undo the assignment. A successful assignment can be either
* undone or left alone as final. A failed assignment has the same effect
* as calling pcmk__unassign_resource(); there are no side effects on
* roles or actions.
*/
pe_node_t *
pcmk__clone_assign(pe_resource_t *rsc, const pe_node_t *prefer,
bool stop_if_fail)
{
CRM_ASSERT(pe_rsc_is_clone(rsc));
if (!pcmk_is_set(rsc->flags, pe_rsc_provisional)) {
return NULL; // Assignment has already been done
}
// Detect assignment loops
if (pcmk_is_set(rsc->flags, pe_rsc_allocating)) {
pe_rsc_debug(rsc, "Breaking assignment loop involving %s", rsc->id);
return NULL;
}
pe__set_resource_flags(rsc, pe_rsc_allocating);
// If this clone is promotable, consider nodes' promotion scores
if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) {
pcmk__add_promotion_scores(rsc);
}
/* If this clone is colocated with any other resources, assign those first.
* Since the this_with_colocations() method boils down to a copy of rsc_cons
* for clones, we can use that here directly for efficiency.
*/
for (GList *iter = rsc->rsc_cons; iter != NULL; iter = iter->next) {
pcmk__colocation_t *constraint = (pcmk__colocation_t *) iter->data;
pe_rsc_trace(rsc, "%s: Assigning colocation %s primary %s first",
rsc->id, constraint->id, constraint->primary->id);
constraint->primary->cmds->assign(constraint->primary, prefer,
stop_if_fail);
}
/* If any resources are colocated with this one, consider their preferences.
* Because the with_this_colocations() method boils down to a copy of
* rsc_cons_lhs for clones, we can use that here directly for efficiency.
*/
g_list_foreach(rsc->rsc_cons_lhs, pcmk__add_dependent_scores, rsc);
pe__show_node_scores(!pcmk_is_set(rsc->cluster->flags, pe_flag_show_scores),
rsc, __func__, rsc->allowed_nodes, rsc->cluster);
rsc->children = g_list_sort(rsc->children, pcmk__cmp_instance);
pcmk__assign_instances(rsc, rsc->children, pe__clone_max(rsc),
pe__clone_node_max(rsc));
if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) {
pcmk__set_instance_roles(rsc);
}
pe__clear_resource_flags(rsc, pe_rsc_provisional|pe_rsc_allocating);
pe_rsc_trace(rsc, "Assigned clone %s", rsc->id);
return NULL;
}
/*!
* \internal
* \brief Create all actions needed for a given clone resource
*
* \param[in,out] rsc Clone resource to create actions for
*/
void
pcmk__clone_create_actions(pe_resource_t *rsc)
{
CRM_ASSERT(pe_rsc_is_clone(rsc));
pe_rsc_trace(rsc, "Creating actions for clone %s", rsc->id);
pcmk__create_instance_actions(rsc, rsc->children);
if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) {
pcmk__create_promotable_actions(rsc);
}
}
/*!
* \internal
* \brief Create implicit constraints needed for a clone resource
*
* \param[in,out] rsc Clone resource to create implicit constraints for
*/
void
pcmk__clone_internal_constraints(pe_resource_t *rsc)
{
bool ordered = false;
CRM_ASSERT(pe_rsc_is_clone(rsc));
pe_rsc_trace(rsc, "Creating internal constraints for clone %s", rsc->id);
// Restart ordering: Stop -> stopped -> start -> started
pcmk__order_resource_actions(rsc, RSC_STOPPED, rsc, RSC_START,
pe_order_optional);
pcmk__order_resource_actions(rsc, RSC_START, rsc, RSC_STARTED,
pe_order_runnable_left);
pcmk__order_resource_actions(rsc, RSC_STOP, rsc, RSC_STOPPED,
pe_order_runnable_left);
// Demoted -> stop and started -> promote
if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) {
pcmk__order_resource_actions(rsc, RSC_DEMOTED, rsc, RSC_STOP,
pe_order_optional);
pcmk__order_resource_actions(rsc, RSC_STARTED, rsc, RSC_PROMOTE,
pe_order_runnable_left);
}
ordered = pe__clone_is_ordered(rsc);
if (ordered) {
/* Ordered clone instances must start and stop by instance number. The
* instances might have been previously shuffled for assignment or
* promotion purposes, so re-sort them.
*/
rsc->children = g_list_sort(rsc->children, pcmk__cmp_instance_number);
}
for (GList *iter = rsc->children; iter != NULL; iter = iter->next) {
pe_resource_t *instance = (pe_resource_t *) iter->data;
instance->cmds->internal_constraints(instance);
// Start clone -> start instance -> clone started
pcmk__order_starts(rsc, instance, pe_order_runnable_left
|pe_order_implies_first_printed);
pcmk__order_resource_actions(instance, RSC_START, rsc, RSC_STARTED,
pe_order_implies_then_printed);
// Stop clone -> stop instance -> clone stopped
pcmk__order_stops(rsc, instance, pe_order_implies_first_printed);
pcmk__order_resource_actions(instance, RSC_STOP, rsc, RSC_STOPPED,
pe_order_implies_then_printed);
/* Instances of ordered clones must be started and stopped by instance
* number. Since only some instances may be starting or stopping, order
* each instance relative to every later instance.
*/
if (ordered) {
for (GList *later = iter->next;
later != NULL; later = later->next) {
pcmk__order_starts(instance, (pe_resource_t *) later->data,
pe_order_optional);
pcmk__order_stops((pe_resource_t *) later->data, instance,
pe_order_optional);
}
}
}
if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) {
pcmk__order_promotable_instances(rsc);
}
}
/*!
* \internal
* \brief Check whether colocated resources can be interleaved
*
* \param[in] colocation Colocation constraint with clone as primary
*
* \return true if colocated resources can be interleaved, otherwise false
*/
static bool
can_interleave(const pcmk__colocation_t *colocation)
{
const pe_resource_t *dependent = colocation->dependent;
// Only colocations between clone or bundle resources use interleaving
if (dependent->variant <= pe_group) {
return false;
}
// Only the dependent needs to be marked for interleaving
if (!crm_is_true(g_hash_table_lookup(dependent->meta,
XML_RSC_ATTR_INTERLEAVE))) {
return false;
}
/* @TODO Do we actually care about multiple primary instances sharing a
* dependent instance?
*/
if (dependent->fns->max_per_node(dependent)
!= colocation->primary->fns->max_per_node(colocation->primary)) {
pcmk__config_err("Cannot interleave %s and %s because they do not "
"support the same number of instances per node",
dependent->id, colocation->primary->id);
return false;
}
return true;
}
/*!
* \internal
* \brief Apply a colocation's score to node scores or resource priority
*
* Given a colocation constraint, apply its score to the dependent's
* allowed node scores (if we are still placing resources) or priority (if
* we are choosing promotable clone instance roles).
*
* \param[in,out] dependent Dependent resource in colocation
* \param[in] primary Primary resource in colocation
* \param[in] colocation Colocation constraint to apply
* \param[in] for_dependent true if called on behalf of dependent
*/
void
pcmk__clone_apply_coloc_score(pe_resource_t *dependent,
const pe_resource_t *primary,
const pcmk__colocation_t *colocation,
bool for_dependent)
{
const GList *iter = NULL;
/* This should never be called for the clone itself as a dependent. Instead,
* we add its colocation constraints to its instances and call the
* apply_coloc_score() method for the instances as dependents.
*/
CRM_ASSERT(!for_dependent);
CRM_ASSERT((colocation != NULL) && pe_rsc_is_clone(primary)
&& (dependent != NULL) && (dependent->variant == pe_native));
if (pcmk_is_set(primary->flags, pe_rsc_provisional)) {
pe_rsc_trace(primary,
"Delaying processing colocation %s "
"because cloned primary %s is still provisional",
colocation->id, primary->id);
return;
}
pe_rsc_trace(primary, "Processing colocation %s (%s with clone %s @%s)",
colocation->id, dependent->id, primary->id,
pcmk_readable_score(colocation->score));
// Apply role-specific colocations
if (pcmk_is_set(primary->flags, pe_rsc_promotable)
&& (colocation->primary_role != RSC_ROLE_UNKNOWN)) {
if (pcmk_is_set(dependent->flags, pe_rsc_provisional)) {
// We're assigning the dependent to a node
pcmk__update_dependent_with_promotable(primary, dependent,
colocation);
return;
}
if (colocation->dependent_role == RSC_ROLE_PROMOTED) {
// We're choosing a role for the dependent
pcmk__update_promotable_dependent_priority(primary, dependent,
colocation);
return;
}
}
// Apply interleaved colocations
if (can_interleave(colocation)) {
const pe_resource_t *primary_instance = NULL;
primary_instance = pcmk__find_compatible_instance(dependent, primary,
RSC_ROLE_UNKNOWN,
false);
if (primary_instance != NULL) {
pe_rsc_debug(primary, "Interleaving %s with %s",
dependent->id, primary_instance->id);
dependent->cmds->apply_coloc_score(dependent, primary_instance,
colocation, true);
} else if (colocation->score >= INFINITY) {
crm_notice("%s cannot run because it cannot interleave with "
"any instance of %s", dependent->id, primary->id);
pcmk__assign_resource(dependent, NULL, true, true);
} else {
pe_rsc_debug(primary,
"%s will not colocate with %s "
"because no instance can interleave with it",
dependent->id, primary->id);
}
return;
}
// Apply mandatory colocations
if (colocation->score >= INFINITY) {
GList *primary_nodes = NULL;
// Dependent can run only where primary will have unblocked instances
for (iter = primary->children; iter != NULL; iter = iter->next) {
const pe_resource_t *instance = iter->data;
pe_node_t *chosen = instance->fns->location(instance, NULL, 0);
if ((chosen != NULL)
&& !is_set_recursive(instance, pe_rsc_block, TRUE)) {
pe_rsc_trace(primary, "Allowing %s: %s %d",
colocation->id, pe__node_name(chosen),
chosen->weight);
primary_nodes = g_list_prepend(primary_nodes, chosen);
}
}
node_list_exclude(dependent->allowed_nodes, primary_nodes, FALSE);
g_list_free(primary_nodes);
return;
}
// Apply optional colocations
for (iter = primary->children; iter != NULL; iter = iter->next) {
const pe_resource_t *instance = iter->data;
instance->cmds->apply_coloc_score(dependent, instance, colocation,
false);
}
}
// Clone implementation of resource_alloc_functions_t:with_this_colocations()
void
pcmk__with_clone_colocations(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc, GList **list)
{
CRM_CHECK((rsc != NULL) && (orig_rsc != NULL) && (list != NULL), return);
- if (rsc == orig_rsc) { // Colocations are wanted for clone itself
- pcmk__add_with_this_list(list, rsc->rsc_cons_lhs, orig_rsc);
- } else {
- pcmk__add_collective_constraints(list, orig_rsc, rsc, true);
- }
+ pcmk__add_with_this_list(list, rsc->rsc_cons_lhs, orig_rsc);
}
// Clone implementation of resource_alloc_functions_t:this_with_colocations()
void
pcmk__clone_with_colocations(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc, GList **list)
{
CRM_CHECK((rsc != NULL) && (orig_rsc != NULL) && (list != NULL), return);
- if (rsc == orig_rsc) { // Colocations are wanted for clone itself
- pcmk__add_this_with_list(list, rsc->rsc_cons, orig_rsc);
- } else {
- pcmk__add_collective_constraints(list, orig_rsc, rsc, false);
- }
+ pcmk__add_this_with_list(list, rsc->rsc_cons, orig_rsc);
}
/*!
* \internal
* \brief Return action flags for a given clone resource action
*
* \param[in,out] action Action to get flags for
* \param[in] node If not NULL, limit effects to this node
*
* \return Flags appropriate to \p action on \p node
*/
uint32_t
pcmk__clone_action_flags(pe_action_t *action, const pe_node_t *node)
{
CRM_ASSERT((action != NULL) && pe_rsc_is_clone(action->rsc));
return pcmk__collective_action_flags(action, action->rsc->children, node);
}
/*!
* \internal
* \brief Apply a location constraint to a clone resource's allowed node scores
*
* \param[in,out] rsc Clone resource to apply constraint to
* \param[in,out] location Location constraint to apply
*/
void
pcmk__clone_apply_location(pe_resource_t *rsc, pe__location_t *location)
{
CRM_CHECK((location != NULL) && pe_rsc_is_clone(rsc), return);
pcmk__apply_location(rsc, location);
for (GList *iter = rsc->children; iter != NULL; iter = iter->next) {
pe_resource_t *instance = (pe_resource_t *) iter->data;
instance->cmds->apply_location(instance, location);
}
}
// GFunc wrapper for calling the action_flags() resource method
static void
call_action_flags(gpointer data, gpointer user_data)
{
pe_resource_t *rsc = user_data;
rsc->cmds->action_flags((pe_action_t *) data, NULL);
}
/*!
* \internal
* \brief Add a clone resource's actions to the transition graph
*
* \param[in,out] rsc Resource whose actions should be added
*/
void
pcmk__clone_add_actions_to_graph(pe_resource_t *rsc)
{
CRM_ASSERT(pe_rsc_is_clone(rsc));
g_list_foreach(rsc->actions, call_action_flags, rsc);
pe__create_clone_notifications(rsc);
for (GList *iter = rsc->children; iter != NULL; iter = iter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) iter->data;
child_rsc->cmds->add_actions_to_graph(child_rsc);
}
pcmk__add_rsc_actions_to_graph(rsc);
pe__free_clone_notification_data(rsc);
}
/*!
* \internal
* \brief Check whether a resource or any children have been probed on a node
*
* \param[in] rsc Resource to check
* \param[in] node Node to check
*
* \return true if \p node is in the known_on table of \p rsc or any of its
* children, otherwise false
*/
static bool
rsc_probed_on(const pe_resource_t *rsc, const pe_node_t *node)
{
if (rsc->children != NULL) {
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_probed_on(child, node)) {
return true;
}
}
return false;
}
if (rsc->known_on != NULL) {
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 (pe__same_node(node, known_node)) {
return true;
}
}
}
return false;
}
/*!
* \internal
* \brief Find clone instance that has been probed on given node
*
* \param[in] clone Clone resource to check
* \param[in] node Node to check
*
* \return Instance of \p clone that has been probed on \p node if any,
* otherwise NULL
*/
static pe_resource_t *
find_probed_instance_on(const pe_resource_t *clone, const pe_node_t *node)
{
for (GList *iter = clone->children; iter != NULL; iter = iter->next) {
pe_resource_t *instance = (pe_resource_t *) iter->data;
if (rsc_probed_on(instance, node)) {
return instance;
}
}
return NULL;
}
/*!
* \internal
* \brief Probe an anonymous clone on a node
*
* \param[in,out] clone Anonymous clone to probe
* \param[in,out] node Node to probe \p clone on
*/
static bool
probe_anonymous_clone(pe_resource_t *clone, pe_node_t *node)
{
// Check whether we already probed an instance on this node
pe_resource_t *child = find_probed_instance_on(clone, node);
// Otherwise, check if we plan to start an instance on this node
for (GList *iter = clone->children; (iter != NULL) && (child == NULL);
iter = iter->next) {
pe_resource_t *instance = (pe_resource_t *) iter->data;
const pe_node_t *instance_node = NULL;
instance_node = instance->fns->location(instance, NULL, 0);
if (pe__same_node(instance_node, node)) {
child = instance;
}
}
// Otherwise, use the first clone instance
if (child == NULL) {
child = clone->children->data;
}
// Anonymous clones only need to probe a single instance
return child->cmds->create_probe(child, node);
}
/*!
* \internal
* \brief Schedule any probes needed for a resource on a node
*
* \param[in,out] rsc Resource to create probe for
* \param[in,out] node Node to create probe on
*
* \return true if any probe was created, otherwise false
*/
bool
pcmk__clone_create_probe(pe_resource_t *rsc, pe_node_t *node)
{
CRM_ASSERT((node != NULL) && pe_rsc_is_clone(rsc));
if (rsc->exclusive_discover) {
/* The clone is configured to be probed only where a location constraint
* exists with resource-discovery set to exclusive.
*
* This check is not strictly necessary here since the instance's
* create_probe() method would also check, but doing it here is more
* efficient (especially for unique clones with a large number of
* instances), and affects the CRM_meta_notify_available_uname variable
* passed with notify actions.
*/
pe_node_t *allowed = g_hash_table_lookup(rsc->allowed_nodes,
node->details->id);
if ((allowed == NULL)
|| (allowed->rsc_discover_mode != pe_discover_exclusive)) {
/* This node is not marked for resource discovery. Remove it from
* allowed_nodes so that notifications contain only nodes that the
* clone can possibly run on.
*/
pe_rsc_trace(rsc,
"Skipping probe for %s on %s because resource has "
"exclusive discovery but is not allowed on node",
rsc->id, pe__node_name(node));
g_hash_table_remove(rsc->allowed_nodes, node->details->id);
return false;
}
}
rsc->children = g_list_sort(rsc->children, pcmk__cmp_instance_number);
if (pcmk_is_set(rsc->flags, pe_rsc_unique)) {
return pcmk__probe_resource_list(rsc->children, node);
} else {
return probe_anonymous_clone(rsc, node);
}
}
/*!
* \internal
* \brief Add meta-attributes relevant to transition graph actions to XML
*
* Add clone-specific meta-attributes needed for transition graph actions.
*
* \param[in] rsc Clone resource whose meta-attributes should be added
* \param[in,out] xml Transition graph action attributes XML to add to
*/
void
pcmk__clone_add_graph_meta(const pe_resource_t *rsc, xmlNode *xml)
{
char *name = NULL;
CRM_ASSERT(pe_rsc_is_clone(rsc) && (xml != NULL));
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, pe__clone_max(rsc));
free(name);
name = crm_meta_name(XML_RSC_ATTR_INCARNATION_NODEMAX);
crm_xml_add_int(xml, name, pe__clone_node_max(rsc));
free(name);
if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) {
int promoted_max = pe__clone_promoted_max(rsc);
int promoted_node_max = pe__clone_promoted_node_max(rsc);
name = crm_meta_name(XML_RSC_ATTR_PROMOTED_MAX);
crm_xml_add_int(xml, name, promoted_max);
free(name);
name = crm_meta_name(XML_RSC_ATTR_PROMOTED_NODEMAX);
crm_xml_add_int(xml, name, 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(PCMK_XA_PROMOTED_MAX_LEGACY);
crm_xml_add_int(xml, name, promoted_max);
free(name);
name = crm_meta_name(PCMK_XA_PROMOTED_NODE_MAX_LEGACY);
crm_xml_add_int(xml, name, promoted_node_max);
free(name);
}
}
// Clone implementation of resource_alloc_functions_t:add_utilization()
void
pcmk__clone_add_utilization(const pe_resource_t *rsc,
const pe_resource_t *orig_rsc, GList *all_rscs,
GHashTable *utilization)
{
bool existing = false;
pe_resource_t *child = NULL;
CRM_ASSERT(pe_rsc_is_clone(rsc) && (orig_rsc != NULL)
&& (utilization != NULL));
if (!pcmk_is_set(rsc->flags, pe_rsc_provisional)) {
return;
}
// Look for any child already existing in the list
for (GList *iter = rsc->children; iter != NULL; iter = iter->next) {
child = (pe_resource_t *) iter->data;
if (g_list_find(all_rscs, child)) {
existing = true; // Keep checking remaining children
} else {
// If this is a clone of a group, look for group's members
for (GList *member_iter = child->children; member_iter != NULL;
member_iter = member_iter->next) {
pe_resource_t *member = (pe_resource_t *) member_iter->data;
if (g_list_find(all_rscs, member) != NULL) {
// Add *child's* utilization, not group member's
child->cmds->add_utilization(child, orig_rsc, all_rscs,
utilization);
existing = true;
break;
}
}
}
}
if (!existing && (rsc->children != NULL)) {
// If nothing was found, still add first child's utilization
child = (pe_resource_t *) rsc->children->data;
child->cmds->add_utilization(child, orig_rsc, all_rscs, utilization);
}
}
// Clone implementation of resource_alloc_functions_t:shutdown_lock()
void
pcmk__clone_shutdown_lock(pe_resource_t *rsc)
{
CRM_ASSERT(pe_rsc_is_clone(rsc));
return; // Clones currently don't support shutdown locks
}
diff --git a/lib/pacemaker/pcmk_sched_instances.c b/lib/pacemaker/pcmk_sched_instances.c
index fdcc139e32..ef6a797c01 100644
--- a/lib/pacemaker/pcmk_sched_instances.c
+++ b/lib/pacemaker/pcmk_sched_instances.c
@@ -1,1764 +1,1668 @@
/*
* Copyright 2004-2023 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.
*/
/* This file is intended for code usable with both clone instances and bundle
* replica containers.
*/
#include <crm_internal.h>
#include <crm/msg_xml.h>
#include <pacemaker-internal.h>
#include "libpacemaker_private.h"
-/*!
- * \internal
- * \brief Check whether a clone or bundle has instances for all available nodes
- *
- * \param[in] collective Clone or bundle to check
- *
- * \return true if \p collective has enough instances for all of its available
- * allowed nodes, otherwise false
- */
-static bool
-can_run_everywhere(const pe_resource_t *collective)
-{
- GHashTableIter iter;
- pe_node_t *node = NULL;
- int available_nodes = 0;
- int max_instances = 0;
-
- switch (collective->variant) {
- case pe_clone:
- max_instances = pe__clone_max(collective);
- break;
- case pe_container:
- max_instances = pe__bundle_max(collective);
- break;
- default:
- return false; // Not actually possible
- }
-
- g_hash_table_iter_init(&iter, collective->allowed_nodes);
- while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) {
- if (pcmk__node_available(node, false, false)
- && (max_instances < ++available_nodes)) {
- return false;
- }
- }
- return true;
-}
-
/*!
* \internal
* \brief Check whether a node is allowed to run an instance
*
* \param[in] instance Clone instance or bundle container to check
* \param[in] node Node to check
* \param[in] max_per_node Maximum number of instances allowed to run on a node
*
* \return true if \p node is allowed to run \p instance, otherwise false
*/
static bool
can_run_instance(const pe_resource_t *instance, const pe_node_t *node,
int max_per_node)
{
pe_node_t *allowed_node = NULL;
if (pcmk_is_set(instance->flags, pe_rsc_orphan)) {
pe_rsc_trace(instance, "%s cannot run on %s: orphaned",
instance->id, pe__node_name(node));
return false;
}
if (!pcmk__node_available(node, false, false)) {
pe_rsc_trace(instance,
"%s cannot run on %s: node cannot run resources",
instance->id, pe__node_name(node));
return false;
}
allowed_node = pcmk__top_allowed_node(instance, node);
if (allowed_node == NULL) {
crm_warn("%s cannot run on %s: node not allowed",
instance->id, pe__node_name(node));
return false;
}
if (allowed_node->weight < 0) {
pe_rsc_trace(instance, "%s cannot run on %s: parent score is %s there",
instance->id, pe__node_name(node),
pcmk_readable_score(allowed_node->weight));
return false;
}
if (allowed_node->count >= max_per_node) {
pe_rsc_trace(instance,
"%s cannot run on %s: node already has %d instance%s",
instance->id, pe__node_name(node), max_per_node,
pcmk__plural_s(max_per_node));
return false;
}
pe_rsc_trace(instance, "%s can run on %s (%d already running)",
instance->id, pe__node_name(node), allowed_node->count);
return true;
}
/*!
* \internal
* \brief Ban a clone instance or bundle replica from unavailable allowed nodes
*
* \param[in,out] instance Clone instance or bundle replica to ban
* \param[in] max_per_node Maximum instances allowed to run on a node
*/
static void
ban_unavailable_allowed_nodes(pe_resource_t *instance, int max_per_node)
{
if (instance->allowed_nodes != NULL) {
GHashTableIter iter;
pe_node_t *node = NULL;
g_hash_table_iter_init(&iter, instance->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) {
if (!can_run_instance(instance, node, max_per_node)) {
pe_rsc_trace(instance, "Banning %s from unavailable node %s",
instance->id, pe__node_name(node));
node->weight = -INFINITY;
for (GList *child_iter = instance->children;
child_iter != NULL; child_iter = child_iter->next) {
pe_resource_t *child = (pe_resource_t *) child_iter->data;
pe_node_t *child_node = NULL;
child_node = pe_hash_table_lookup(child->allowed_nodes,
node->details->id);
if (child_node != NULL) {
pe_rsc_trace(instance,
"Banning %s child %s "
"from unavailable node %s",
instance->id, child->id,
pe__node_name(node));
child_node->weight = -INFINITY;
}
}
}
}
}
}
/*!
* \internal
* \brief Create a hash table with a single node in it
*
* \param[in] node Node to copy into new table
*
* \return Newly created hash table containing a copy of \p node
* \note The caller is responsible for freeing the result with
* g_hash_table_destroy().
*/
static GHashTable *
new_node_table(pe_node_t *node)
{
GHashTable *table = pcmk__strkey_table(NULL, free);
node = pe__copy_node(node);
g_hash_table_insert(table, (gpointer) node->details->id, node);
return table;
}
/*!
* \internal
* \brief Apply a resource's parent's colocation scores to a node table
*
* \param[in] rsc Resource whose colocations should be applied
* \param[in,out] nodes Node table to apply colocations to
*/
static void
apply_parent_colocations(const pe_resource_t *rsc, GHashTable **nodes)
{
GList *iter = NULL;
pcmk__colocation_t *colocation = NULL;
pe_resource_t *other = NULL;
float factor = 0.0;
/* Because the this_with_colocations() and with_this_colocations() methods
* boil down to copies of rsc_cons and rsc_cons_lhs for clones and bundles,
* we can use those here directly for efficiency.
*/
for (iter = rsc->parent->rsc_cons; iter != NULL; iter = iter->next) {
colocation = (pcmk__colocation_t *) iter->data;
other = colocation->primary;
factor = colocation->score / (float) INFINITY,
other->cmds->add_colocated_node_scores(other, rsc->id, nodes,
colocation, factor,
pcmk__coloc_select_default);
}
for (iter = rsc->parent->rsc_cons_lhs; iter != NULL; iter = iter->next) {
colocation = (pcmk__colocation_t *) iter->data;
if (!pcmk__colocation_has_influence(colocation, rsc)) {
continue;
}
other = colocation->dependent;
factor = colocation->score / (float) INFINITY,
other->cmds->add_colocated_node_scores(other, rsc->id, nodes,
colocation, factor,
pcmk__coloc_select_nonnegative);
}
}
/*!
* \internal
* \brief Compare clone or bundle instances based on colocation scores
*
* Determine the relative order in which two clone or bundle instances should be
* assigned to nodes, considering the scores of colocation constraints directly
* or indirectly involving them.
*
* \param[in] instance1 First instance to compare
* \param[in] instance2 Second instance to compare
*
* \return A negative number if \p instance1 should be assigned first,
* a positive number if \p instance2 should be assigned first,
* or 0 if assignment order doesn't matter
*/
static int
cmp_instance_by_colocation(const pe_resource_t *instance1,
const pe_resource_t *instance2)
{
int rc = 0;
pe_node_t *node1 = NULL;
pe_node_t *node2 = NULL;
pe_node_t *current_node1 = pe__current_node(instance1);
pe_node_t *current_node2 = pe__current_node(instance2);
GHashTable *colocated_scores1 = NULL;
GHashTable *colocated_scores2 = NULL;
CRM_ASSERT((instance1 != NULL) && (instance1->parent != NULL)
&& (instance2 != NULL) && (instance2->parent != NULL)
&& (current_node1 != NULL) && (current_node2 != NULL));
// Create node tables initialized with each node
colocated_scores1 = new_node_table(current_node1);
colocated_scores2 = new_node_table(current_node2);
// Apply parental colocations
apply_parent_colocations(instance1, &colocated_scores1);
apply_parent_colocations(instance2, &colocated_scores2);
// Find original nodes again, with scores updated for colocations
node1 = g_hash_table_lookup(colocated_scores1, current_node1->details->id);
node2 = g_hash_table_lookup(colocated_scores2, current_node2->details->id);
// Compare nodes by updated scores
if (node1->weight < node2->weight) {
crm_trace("Assign %s (%d on %s) after %s (%d on %s)",
instance1->id, node1->weight, pe__node_name(node1),
instance2->id, node2->weight, pe__node_name(node2));
rc = 1;
} else if (node1->weight > node2->weight) {
crm_trace("Assign %s (%d on %s) before %s (%d on %s)",
instance1->id, node1->weight, pe__node_name(node1),
instance2->id, node2->weight, pe__node_name(node2));
rc = -1;
}
g_hash_table_destroy(colocated_scores1);
g_hash_table_destroy(colocated_scores2);
return rc;
}
/*!
* \internal
* \brief Check whether a resource or any of its children are failed
*
* \param[in] rsc Resource to check
*
* \return true if \p rsc or any of its children are failed, otherwise false
*/
static bool
did_fail(const pe_resource_t *rsc)
{
if (pcmk_is_set(rsc->flags, pe_rsc_failed)) {
return true;
}
for (GList *iter = rsc->children; iter != NULL; iter = iter->next) {
if (did_fail((const pe_resource_t *) iter->data)) {
return true;
}
}
return false;
}
/*!
* \internal
* \brief Check whether a node is allowed to run a resource
*
* \param[in] rsc Resource to check
* \param[in,out] node Node to check (will be set NULL if not allowed)
*
* \return true if *node is either NULL or allowed for \p rsc, otherwise false
*/
static bool
node_is_allowed(const pe_resource_t *rsc, pe_node_t **node)
{
if (*node != NULL) {
pe_node_t *allowed = pe_hash_table_lookup(rsc->allowed_nodes,
(*node)->details->id);
if ((allowed == NULL) || (allowed->weight < 0)) {
pe_rsc_trace(rsc, "%s: current location (%s) is unavailable",
rsc->id, pe__node_name(*node));
*node = NULL;
return false;
}
}
return true;
}
/*!
* \internal
* \brief Compare two clone or bundle instances' instance numbers
*
* \param[in] a First instance to compare
* \param[in] b Second instance to compare
*
* \return A negative number if \p a's instance number is lower,
* a positive number if \p b's instance number is lower,
* or 0 if their instance numbers are the same
*/
gint
pcmk__cmp_instance_number(gconstpointer a, gconstpointer b)
{
const pe_resource_t *instance1 = (const pe_resource_t *) a;
const pe_resource_t *instance2 = (const pe_resource_t *) b;
char *div1 = NULL;
char *div2 = NULL;
CRM_ASSERT((instance1 != NULL) && (instance2 != NULL));
// Clone numbers are after a colon, bundle numbers after a dash
div1 = strrchr(instance1->id, ':');
if (div1 == NULL) {
div1 = strrchr(instance1->id, '-');
}
div2 = strrchr(instance2->id, ':');
if (div2 == NULL) {
div2 = strrchr(instance2->id, '-');
}
CRM_ASSERT((div1 != NULL) && (div2 != NULL));
return (gint) (strtol(div1 + 1, NULL, 10) - strtol(div2 + 1, NULL, 10));
}
/*!
* \internal
* \brief Compare clone or bundle instances according to assignment order
*
* Compare two clone or bundle instances according to the order they should be
* assigned to nodes, preferring (in order):
*
* - Active instance that is less multiply active
* - Instance that is not active on a disallowed node
* - Instance with higher configured priority
* - Active instance whose current node can run resources
* - Active instance whose parent is allowed on current node
* - Active instance whose current node has fewer other instances
* - Active instance
* - Instance that isn't failed
* - Instance whose colocations result in higher score on current node
* - Instance with lower ID in lexicographic order
*
* \param[in] a First instance to compare
* \param[in] b Second instance to compare
*
* \return A negative number if \p a should be assigned first,
* a positive number if \p b should be assigned first,
* or 0 if assignment order doesn't matter
*/
gint
pcmk__cmp_instance(gconstpointer a, gconstpointer b)
{
int rc = 0;
pe_node_t *node1 = NULL;
pe_node_t *node2 = NULL;
unsigned int nnodes1 = 0;
unsigned int nnodes2 = 0;
bool can1 = true;
bool can2 = true;
const pe_resource_t *instance1 = (const pe_resource_t *) a;
const pe_resource_t *instance2 = (const pe_resource_t *) b;
CRM_ASSERT((instance1 != NULL) && (instance2 != NULL));
node1 = instance1->fns->active_node(instance1, &nnodes1, NULL);
node2 = instance2->fns->active_node(instance2, &nnodes2, NULL);
/* If both instances are running and at least one is multiply
* active, prefer instance that's running on fewer nodes.
*/
if ((nnodes1 > 0) && (nnodes2 > 0)) {
if (nnodes1 < nnodes2) {
crm_trace("Assign %s (active on %d) before %s (active on %d): "
"less multiply active",
instance1->id, nnodes1, instance2->id, nnodes2);
return -1;
} else if (nnodes1 > nnodes2) {
crm_trace("Assign %s (active on %d) after %s (active on %d): "
"more multiply active",
instance1->id, nnodes1, instance2->id, nnodes2);
return 1;
}
}
/* An instance that is either inactive or active on an allowed node is
* preferred over an instance that is active on a no-longer-allowed node.
*/
can1 = node_is_allowed(instance1, &node1);
can2 = node_is_allowed(instance2, &node2);
if (can1 && !can2) {
crm_trace("Assign %s before %s: not active on a disallowed node",
instance1->id, instance2->id);
return -1;
} else if (!can1 && can2) {
crm_trace("Assign %s after %s: active on a disallowed node",
instance1->id, instance2->id);
return 1;
}
// Prefer instance with higher configured priority
if (instance1->priority > instance2->priority) {
crm_trace("Assign %s before %s: priority (%d > %d)",
instance1->id, instance2->id,
instance1->priority, instance2->priority);
return -1;
} else if (instance1->priority < instance2->priority) {
crm_trace("Assign %s after %s: priority (%d < %d)",
instance1->id, instance2->id,
instance1->priority, instance2->priority);
return 1;
}
// Prefer active instance
if ((node1 == NULL) && (node2 == NULL)) {
crm_trace("No assignment preference for %s vs. %s: inactive",
instance1->id, instance2->id);
return 0;
} else if (node1 == NULL) {
crm_trace("Assign %s after %s: active", instance1->id, instance2->id);
return 1;
} else if (node2 == NULL) {
crm_trace("Assign %s before %s: active", instance1->id, instance2->id);
return -1;
}
// Prefer instance whose current node can run resources
can1 = pcmk__node_available(node1, false, false);
can2 = pcmk__node_available(node2, false, false);
if (can1 && !can2) {
crm_trace("Assign %s before %s: current node can run resources",
instance1->id, instance2->id);
return -1;
} else if (!can1 && can2) {
crm_trace("Assign %s after %s: current node can't run resources",
instance1->id, instance2->id);
return 1;
}
// Prefer instance whose parent is allowed to run on instance's current node
node1 = pcmk__top_allowed_node(instance1, node1);
node2 = pcmk__top_allowed_node(instance2, node2);
if ((node1 == NULL) && (node2 == NULL)) {
crm_trace("No assignment preference for %s vs. %s: "
"parent not allowed on either instance's current node",
instance1->id, instance2->id);
return 0;
} else if (node1 == NULL) {
crm_trace("Assign %s after %s: parent not allowed on current node",
instance1->id, instance2->id);
return 1;
} else if (node2 == NULL) {
crm_trace("Assign %s before %s: parent allowed on current node",
instance1->id, instance2->id);
return -1;
}
// Prefer instance whose current node is running fewer other instances
if (node1->count < node2->count) {
crm_trace("Assign %s before %s: fewer active instances on current node",
instance1->id, instance2->id);
return -1;
} else if (node1->count > node2->count) {
crm_trace("Assign %s after %s: more active instances on current node",
instance1->id, instance2->id);
return 1;
}
// Prefer instance that isn't failed
can1 = did_fail(instance1);
can2 = did_fail(instance2);
if (!can1 && can2) {
crm_trace("Assign %s before %s: not failed",
instance1->id, instance2->id);
return -1;
} else if (can1 && !can2) {
crm_trace("Assign %s after %s: failed",
instance1->id, instance2->id);
return 1;
}
// Prefer instance with higher cumulative colocation score on current node
rc = cmp_instance_by_colocation(instance1, instance2);
if (rc != 0) {
return rc;
}
// Prefer instance with lower instance number
rc = pcmk__cmp_instance_number(instance1, instance2);
if (rc < 0) {
crm_trace("Assign %s before %s: instance number",
instance1->id, instance2->id);
} else if (rc > 0) {
crm_trace("Assign %s after %s: instance number",
instance1->id, instance2->id);
} else {
crm_trace("No assignment preference for %s vs. %s",
instance1->id, instance2->id);
}
return rc;
}
/*!
* \internal
* \brief Increment the parent's instance count after assigning an instance
*
* An instance's parent tracks how many instances have been assigned to each
* node via its pe_node_t:count member. After assigning an instance to a node,
* find the corresponding node in the parent's allowed table and increment it.
*
* \param[in,out] instance Instance whose parent to update
* \param[in] assigned_to Node to which the instance was assigned
*/
static void
increment_parent_count(pe_resource_t *instance, const pe_node_t *assigned_to)
{
pe_node_t *allowed = NULL;
if (assigned_to == NULL) {
return;
}
allowed = pcmk__top_allowed_node(instance, assigned_to);
if (allowed == NULL) {
/* The instance is allowed on the node, but its parent isn't. This
* shouldn't be possible if the resource is managed, and we won't be
* able to limit the number of instances assigned to the node.
*/
CRM_LOG_ASSERT(!pcmk_is_set(instance->flags, pe_rsc_managed));
} else {
allowed->count++;
}
}
/*!
* \internal
* \brief Assign an instance to a node
*
* \param[in,out] instance Clone instance or bundle replica container
* \param[in] prefer If not NULL, attempt early assignment to this
* node, if still the best choice; otherwise,
* perform final assignment
* \param[in] max_per_node Assign at most this many instances to one node
*
* \return Node to which \p instance is assigned
*/
static const pe_node_t *
assign_instance(pe_resource_t *instance, const pe_node_t *prefer,
int max_per_node)
{
pe_node_t *chosen = NULL;
pe_rsc_trace(instance, "Assigning %s (preferring %s)", instance->id,
((prefer == NULL)? "no node" : prefer->details->uname));
if (pcmk_is_set(instance->flags, pe_rsc_allocating)) {
pe_rsc_debug(instance,
"Assignment loop detected involving %s colocations",
instance->id);
return NULL;
}
ban_unavailable_allowed_nodes(instance, max_per_node);
// Failed early assignments are reversible (stop_if_fail=false)
chosen = instance->cmds->assign(instance, prefer, (prefer == NULL));
increment_parent_count(instance, chosen);
return chosen;
}
/*!
* \internal
* \brief Try to assign an instance to its current node early
*
* \param[in] rsc Clone or bundle being assigned (for logs only)
* \param[in] instance Clone instance or bundle replica container
* \param[in] current Instance's current node
* \param[in] max_per_node Maximum number of instances per node
* \param[in] available Number of instances still available for assignment
*
* \return \c true if \p instance was successfully assigned to its current node,
* or \c false otherwise
*/
static bool
assign_instance_early(const pe_resource_t *rsc, pe_resource_t *instance,
const pe_node_t *current, int max_per_node, int available)
{
const pe_node_t *chosen = NULL;
int reserved = 0;
pe_resource_t *parent = instance->parent;
GHashTable *allowed_orig = NULL;
GHashTable *allowed_orig_parent = parent->allowed_nodes;
const pe_node_t *allowed_node = g_hash_table_lookup(instance->allowed_nodes,
current->details->id);
pe_rsc_trace(instance, "Trying to assign %s to its current node %s",
instance->id, pe__node_name(current));
if (!pcmk__node_available(allowed_node, true, false)) {
pe_rsc_info(instance,
"Not assigning %s to current node %s: unavailable",
instance->id, pe__node_name(current));
return false;
}
/* On each iteration, if instance gets assigned to a node other than its
* current one, we reserve one instance for the chosen node, unassign
* instance, restore instance's original node tables, and try again. This
* way, instances are proportionally assigned to nodes based on preferences,
* but shuffling of specific instances is minimized. If a node will be
* assigned instances at all, it preferentially receives instances that are
* currently active there.
*
* parent->allowed_nodes tracks the number of instances assigned to each
* node. If a node already has max_per_node instances assigned,
* ban_unavailable_allowed_nodes() marks it as unavailable.
*
* In the end, we restore the original parent->allowed_nodes to undo the
* changes to counts during tentative assignments. If we successfully
* assigned instance to its current node, we increment that node's counter.
*/
// Back up the allowed node tables of instance and its children recursively
pcmk__copy_node_tables(instance, &allowed_orig);
// Update instances-per-node counts in a scratch table
parent->allowed_nodes = pcmk__copy_node_table(parent->allowed_nodes);
while (reserved < available) {
chosen = assign_instance(instance, current, max_per_node);
if (pe__same_node(chosen, current)) {
// Successfully assigned to current node
break;
}
// Assignment updates scores, so restore to original state
pe_rsc_debug(instance, "Rolling back node scores for %s", instance->id);
pcmk__restore_node_tables(instance, allowed_orig);
if (chosen == NULL) {
// Assignment failed, so give up
pe_rsc_info(instance,
"Not assigning %s to current node %s: unavailable",
instance->id, pe__node_name(current));
pe__set_resource_flags(instance, pe_rsc_provisional);
break;
}
// We prefer more strongly to assign an instance to the chosen node
pe_rsc_debug(instance,
"Not assigning %s to current node %s: %s is better",
instance->id, pe__node_name(current),
pe__node_name(chosen));
// Reserve one instance for the chosen node and try again
if (++reserved >= available) {
pe_rsc_info(instance,
"Not assigning %s to current node %s: "
"other assignments are more important",
instance->id, pe__node_name(current));
} else {
pe_rsc_debug(instance,
"Reserved an instance of %s for %s. Retrying "
"assignment of %s to %s",
rsc->id, pe__node_name(chosen), instance->id,
pe__node_name(current));
}
// Clear this assignment (frees chosen); leave instance counts in parent
pcmk__unassign_resource(instance);
chosen = NULL;
}
g_hash_table_destroy(allowed_orig);
// Restore original instances-per-node counts
g_hash_table_destroy(parent->allowed_nodes);
parent->allowed_nodes = allowed_orig_parent;
if (chosen == NULL) {
// Couldn't assign instance to current node
return false;
}
pe_rsc_trace(instance, "Assigned %s to current node %s",
instance->id, pe__node_name(current));
increment_parent_count(instance, chosen);
return true;
}
/*!
* \internal
* \brief Reset the node counts of a resource's allowed nodes to zero
*
* \param[in,out] rsc Resource to reset
*
* \return Number of nodes that are available to run resources
*/
static unsigned int
reset_allowed_node_counts(pe_resource_t *rsc)
{
unsigned int available_nodes = 0;
pe_node_t *node = NULL;
GHashTableIter iter;
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) {
node->count = 0;
if (pcmk__node_available(node, false, false)) {
available_nodes++;
}
}
return available_nodes;
}
/*!
* \internal
* \brief Check whether an instance has a preferred node
*
* \param[in] rsc Clone or bundle being assigned (for logs only)
* \param[in] instance Clone instance or bundle replica container
* \param[in] optimal_per_node Optimal number of instances per node
*
* \return Instance's current node if still available, otherwise NULL
*/
static const pe_node_t *
preferred_node(const pe_resource_t *rsc, const pe_resource_t *instance,
int optimal_per_node)
{
const pe_node_t *node = NULL;
const pe_node_t *parent_node = NULL;
// Check whether instance is active, healthy, and not yet assigned
if ((instance->running_on == NULL)
|| !pcmk_is_set(instance->flags, pe_rsc_provisional)
|| pcmk_is_set(instance->flags, pe_rsc_failed)) {
return NULL;
}
// Check whether instance's current node can run resources
node = pe__current_node(instance);
if (!pcmk__node_available(node, true, false)) {
pe_rsc_trace(rsc, "Not assigning %s to %s early (unavailable)",
instance->id, pe__node_name(node));
return NULL;
}
// Check whether node already has optimal number of instances assigned
parent_node = pcmk__top_allowed_node(instance, node);
if ((parent_node != NULL) && (parent_node->count >= optimal_per_node)) {
pe_rsc_trace(rsc,
"Not assigning %s to %s early "
"(optimal instances already assigned)",
instance->id, pe__node_name(node));
return NULL;
}
return node;
}
/*!
* \internal
* \brief Assign collective instances to nodes
*
* \param[in,out] collective Clone or bundle resource being assigned
* \param[in,out] instances List of clone instances or bundle containers
* \param[in] max_total Maximum instances to assign in total
* \param[in] max_per_node Maximum instances to assign to any one node
*/
void
pcmk__assign_instances(pe_resource_t *collective, GList *instances,
int max_total, int max_per_node)
{
// Reuse node count to track number of assigned instances
unsigned int available_nodes = reset_allowed_node_counts(collective);
int optimal_per_node = 0;
int assigned = 0;
GList *iter = NULL;
pe_resource_t *instance = NULL;
const pe_node_t *current = NULL;
if (available_nodes > 0) {
optimal_per_node = max_total / available_nodes;
}
if (optimal_per_node < 1) {
optimal_per_node = 1;
}
pe_rsc_debug(collective,
"Assigning up to %d %s instance%s to up to %u node%s "
"(at most %d per host, %d optimal)",
max_total, collective->id, pcmk__plural_s(max_total),
available_nodes, pcmk__plural_s(available_nodes),
max_per_node, optimal_per_node);
// Assign as many instances as possible to their current location
for (iter = instances; (iter != NULL) && (assigned < max_total);
iter = iter->next) {
int available = max_total - assigned;
instance = iter->data;
if (!pcmk_is_set(instance->flags, pe_rsc_provisional)) {
continue; // Already assigned
}
current = preferred_node(collective, instance, optimal_per_node);
if ((current != NULL)
&& assign_instance_early(collective, instance, current,
max_per_node, available)) {
assigned++;
}
}
pe_rsc_trace(collective, "Assigned %d of %d instance%s to current node",
assigned, max_total, pcmk__plural_s(max_total));
for (iter = instances; iter != NULL; iter = iter->next) {
instance = (pe_resource_t *) iter->data;
if (!pcmk_is_set(instance->flags, pe_rsc_provisional)) {
continue; // Already assigned
}
if (instance->running_on != NULL) {
current = pe__current_node(instance);
if (pcmk__top_allowed_node(instance, current) == NULL) {
const char *unmanaged = "";
if (!pcmk_is_set(instance->flags, pe_rsc_managed)) {
unmanaged = "Unmanaged resource ";
}
crm_notice("%s%s is running on %s which is no longer allowed",
unmanaged, instance->id, pe__node_name(current));
}
}
if (assigned >= max_total) {
pe_rsc_debug(collective,
"Not assigning %s because maximum %d instances "
"already assigned",
instance->id, max_total);
resource_location(instance, NULL, -INFINITY,
"collective_limit_reached", collective->cluster);
} else if (assign_instance(instance, NULL, max_per_node) != NULL) {
assigned++;
}
}
pe_rsc_debug(collective, "Assigned %d of %d possible instance%s of %s",
assigned, max_total, pcmk__plural_s(max_total),
collective->id);
}
enum instance_state {
instance_starting = (1 << 0),
instance_stopping = (1 << 1),
/* This indicates that some instance is restarting. It's not the same as
* instance_starting|instance_stopping, which would indicate that some
* instance is starting, and some instance (not necessarily the same one) is
* stopping.
*/
instance_restarting = (1 << 2),
instance_active = (1 << 3),
instance_all = instance_starting|instance_stopping
|instance_restarting|instance_active,
};
/*!
* \internal
* \brief Check whether an instance is active, starting, and/or stopping
*
* \param[in] instance Clone instance or bundle replica container
* \param[in,out] state Whether any instance is starting, stopping, etc.
*/
static void
check_instance_state(const pe_resource_t *instance, uint32_t *state)
{
const GList *iter = NULL;
uint32_t instance_state = 0; // State of just this instance
// No need to check further if all conditions have already been detected
if (pcmk_all_flags_set(*state, instance_all)) {
return;
}
// If instance is a collective (a cloned group), check its children instead
if (instance->variant > pe_native) {
for (iter = instance->children;
(iter != NULL) && !pcmk_all_flags_set(*state, instance_all);
iter = iter->next) {
check_instance_state((const pe_resource_t *) iter->data, state);
}
return;
}
// If we get here, instance is a primitive
if (instance->running_on != NULL) {
instance_state |= instance_active;
}
// Check each of the instance's actions for runnable start or stop
for (iter = instance->actions;
(iter != NULL) && !pcmk_all_flags_set(instance_state,
instance_starting
|instance_stopping);
iter = iter->next) {
const pe_action_t *action = (const pe_action_t *) iter->data;
const bool optional = pcmk_is_set(action->flags, pe_action_optional);
if (pcmk__str_eq(RSC_START, action->task, pcmk__str_none)) {
if (!optional && pcmk_is_set(action->flags, pe_action_runnable)) {
pe_rsc_trace(instance, "Instance is starting due to %s",
action->uuid);
instance_state |= instance_starting;
} else {
pe_rsc_trace(instance, "%s doesn't affect %s state (%s)",
action->uuid, instance->id,
(optional? "optional" : "unrunnable"));
}
} else if (pcmk__str_eq(RSC_STOP, action->task, pcmk__str_none)) {
/* Only stop actions can be pseudo-actions for primitives. That
* indicates that the node they are on is being fenced, so the stop
* is implied rather than actually executed.
*/
if (!optional
&& pcmk_any_flags_set(action->flags,
pe_action_pseudo|pe_action_runnable)) {
pe_rsc_trace(instance, "Instance is stopping due to %s",
action->uuid);
instance_state |= instance_stopping;
} else {
pe_rsc_trace(instance, "%s doesn't affect %s state (%s)",
action->uuid, instance->id,
(optional? "optional" : "unrunnable"));
}
}
}
if (pcmk_all_flags_set(instance_state,
instance_starting|instance_stopping)) {
instance_state |= instance_restarting;
}
*state |= instance_state;
}
/*!
* \internal
* \brief Create actions for collective resource instances
*
* \param[in,out] collective Clone or bundle resource to create actions for
* \param[in,out] instances List of clone instances or bundle containers
*/
void
pcmk__create_instance_actions(pe_resource_t *collective, GList *instances)
{
uint32_t state = 0;
pe_action_t *stop = NULL;
pe_action_t *stopped = NULL;
pe_action_t *start = NULL;
pe_action_t *started = NULL;
pe_rsc_trace(collective, "Creating collective instance actions for %s",
collective->id);
// Create actions for each instance appropriate to its variant
for (GList *iter = instances; iter != NULL; iter = iter->next) {
pe_resource_t *instance = (pe_resource_t *) iter->data;
instance->cmds->create_actions(instance);
check_instance_state(instance, &state);
}
// Create pseudo-actions for rsc start and started
start = pe__new_rsc_pseudo_action(collective, RSC_START,
!pcmk_is_set(state, instance_starting),
true);
started = pe__new_rsc_pseudo_action(collective, RSC_STARTED,
!pcmk_is_set(state, instance_starting),
false);
started->priority = INFINITY;
if (pcmk_any_flags_set(state, instance_active|instance_starting)) {
pe__set_action_flags(started, pe_action_runnable);
}
// Create pseudo-actions for rsc stop and stopped
stop = pe__new_rsc_pseudo_action(collective, RSC_STOP,
!pcmk_is_set(state, instance_stopping),
true);
stopped = pe__new_rsc_pseudo_action(collective, RSC_STOPPED,
!pcmk_is_set(state, instance_stopping),
true);
stopped->priority = INFINITY;
if (!pcmk_is_set(state, instance_restarting)) {
pe__set_action_flags(stop, pe_action_migrate_runnable);
}
if (collective->variant == pe_clone) {
pe__create_clone_notif_pseudo_ops(collective, start, started, stop,
stopped);
}
}
/*!
* \internal
* \brief Get a list of clone instances or bundle replica containers
*
* \param[in] rsc Clone or bundle resource
*
* \return Clone instances if \p rsc is a clone, or a newly created list of
* \p rsc's replica containers if \p rsc is a bundle
* \note The caller must call free_instance_list() on the result when the list
* is no longer needed.
*/
static inline GList *
get_instance_list(const pe_resource_t *rsc)
{
if (rsc->variant == pe_container) {
return pe__bundle_containers(rsc);
} else {
return rsc->children;
}
}
/*!
* \internal
* \brief Free any memory created by get_instance_list()
*
* \param[in] rsc Clone or bundle resource passed to get_instance_list()
* \param[in,out] list Return value of get_instance_list() for \p rsc
*/
static inline void
free_instance_list(const pe_resource_t *rsc, GList *list)
{
if (list != rsc->children) {
g_list_free(list);
}
}
/*!
* \internal
* \brief Check whether an instance is compatible with a role and node
*
* \param[in] instance Clone instance or bundle replica container
* \param[in] node Instance must match this node
* \param[in] role If not RSC_ROLE_UNKNOWN, instance must match this role
* \param[in] current If true, compare instance's original node and role,
* otherwise compare assigned next node and role
*
* \return true if \p instance is compatible with \p node and \p role,
* otherwise false
*/
bool
pcmk__instance_matches(const pe_resource_t *instance, const pe_node_t *node,
enum rsc_role_e role, bool current)
{
pe_node_t *instance_node = NULL;
CRM_CHECK((instance != NULL) && (node != NULL), return false);
if ((role != RSC_ROLE_UNKNOWN)
&& (role != instance->fns->state(instance, current))) {
pe_rsc_trace(instance,
"%s is not a compatible instance (role is not %s)",
instance->id, role2text(role));
return false;
}
if (!is_set_recursive(instance, pe_rsc_block, true)) {
// We only want instances that haven't failed
instance_node = instance->fns->location(instance, NULL, current);
}
if (instance_node == NULL) {
pe_rsc_trace(instance,
"%s is not a compatible instance (not assigned to a node)",
instance->id);
return false;
}
if (!pe__same_node(instance_node, node)) {
pe_rsc_trace(instance,
"%s is not a compatible instance (assigned to %s not %s)",
instance->id, pe__node_name(instance_node),
pe__node_name(node));
return false;
}
return true;
}
/*!
* \internal
* \brief Find an instance that matches a given resource by node and role
*
* \param[in] match_rsc Resource that instance must match (for logging only)
* \param[in] rsc Clone or bundle resource to check for matching instance
* \param[in] node Instance must match this node
* \param[in] role If not RSC_ROLE_UNKNOWN, instance must match this role
* \param[in] current If true, compare instance's original node and role,
* otherwise compare assigned next node and role
*
* \return \p rsc instance matching \p node and \p role if any, otherwise NULL
*/
static pe_resource_t *
find_compatible_instance_on_node(const pe_resource_t *match_rsc,
const pe_resource_t *rsc,
const pe_node_t *node, enum rsc_role_e role,
bool current)
{
GList *instances = NULL;
instances = get_instance_list(rsc);
for (GList *iter = instances; iter != NULL; iter = iter->next) {
pe_resource_t *instance = (pe_resource_t *) iter->data;
if (pcmk__instance_matches(instance, node, role, current)) {
pe_rsc_trace(match_rsc,
"Found %s %s instance %s compatible with %s on %s",
role == RSC_ROLE_UNKNOWN? "matching" : role2text(role),
rsc->id, instance->id, match_rsc->id,
pe__node_name(node));
free_instance_list(rsc, instances); // Only frees list, not contents
return instance;
}
}
free_instance_list(rsc, instances);
pe_rsc_trace(match_rsc, "No %s %s instance found compatible with %s on %s",
((role == RSC_ROLE_UNKNOWN)? "matching" : role2text(role)),
rsc->id, match_rsc->id, pe__node_name(node));
return NULL;
}
/*!
* \internal
* \brief Find a clone instance or bundle container compatible with a resource
*
* \param[in] match_rsc Resource that instance must match
* \param[in] rsc Clone or bundle resource to check for matching instance
* \param[in] role If not RSC_ROLE_UNKNOWN, instance must match this role
* \param[in] current If true, compare instance's original node and role,
* otherwise compare assigned next node and role
*
* \return Compatible (by \p role and \p match_rsc location) instance of \p rsc
* if any, otherwise NULL
*/
pe_resource_t *
pcmk__find_compatible_instance(const pe_resource_t *match_rsc,
const pe_resource_t *rsc, enum rsc_role_e role,
bool current)
{
pe_resource_t *instance = NULL;
GList *nodes = NULL;
const pe_node_t *node = match_rsc->fns->location(match_rsc, NULL, current);
// If match_rsc has a node, check only that node
if (node != NULL) {
return find_compatible_instance_on_node(match_rsc, rsc, node, role,
current);
}
// Otherwise check for an instance matching any of match_rsc's allowed nodes
nodes = pcmk__sort_nodes(g_hash_table_get_values(match_rsc->allowed_nodes),
NULL);
for (GList *iter = nodes; (iter != NULL) && (instance == NULL);
iter = iter->next) {
instance = find_compatible_instance_on_node(match_rsc, rsc,
(pe_node_t *) iter->data,
role, current);
}
if (instance == NULL) {
pe_rsc_debug(rsc, "No %s instance found compatible with %s",
rsc->id, match_rsc->id);
}
g_list_free(nodes);
return instance;
}
/*!
* \internal
* \brief Unassign an instance if mandatory ordering has no interleave match
*
* \param[in] first 'First' action in an ordering
* \param[in] then 'Then' action in an ordering
* \param[in,out] then_instance 'Then' instance that has no interleave match
* \param[in] type Group of enum pe_ordering flags to apply
* \param[in] current If true, "then" action is stopped or demoted
*
* \return true if \p then_instance was unassigned, otherwise false
*/
static bool
unassign_if_mandatory(const pe_action_t *first, const pe_action_t *then,
pe_resource_t *then_instance, uint32_t type, bool current)
{
// Allow "then" instance to go down even without an interleave match
if (current) {
pe_rsc_trace(then->rsc,
"%s has no instance to order before stopping "
"or demoting %s",
first->rsc->id, then_instance->id);
/* If the "first" action must be runnable, but there is no "first"
* instance, the "then" instance must not be allowed to come up.
*/
} else if (pcmk_any_flags_set(type, pe_order_runnable_left
|pe_order_implies_then)) {
pe_rsc_info(then->rsc,
"Inhibiting %s from being active "
"because there is no %s instance to interleave",
then_instance->id, first->rsc->id);
return pcmk__assign_resource(then_instance, NULL, true, true);
}
return false;
}
/*!
* \internal
* \brief Find first matching action for a clone instance or bundle container
*
* \param[in] action Action in an interleaved ordering
* \param[in] instance Clone instance or bundle container being interleaved
* \param[in] action_name Action to look for
* \param[in] node If not NULL, require action to be on this node
* \param[in] for_first If true, \p instance is the 'first' resource in the
* ordering, otherwise it is the 'then' resource
*
* \return First action for \p instance (or in some cases if \p instance is a
* bundle container, its containerized resource) that matches
* \p action_name and \p node if any, otherwise NULL
*/
static pe_action_t *
find_instance_action(const pe_action_t *action, const pe_resource_t *instance,
const char *action_name, const pe_node_t *node,
bool for_first)
{
const pe_resource_t *rsc = NULL;
pe_action_t *matching_action = NULL;
/* If instance is a bundle container, sometimes we should interleave the
* action for the container itself, and sometimes for the containerized
* resource.
*
* For example, given "start bundle A then bundle B", B likely requires the
* service inside A's container to be active, rather than just the
* container, so we should interleave the action for A's containerized
* resource. On the other hand, it's possible B's container itself requires
* something from A, so we should interleave the action for B's container.
*
* Essentially, for 'first', we should use the containerized resource for
* everything except stop, and for 'then', we should use the container for
* everything except promote and demote (which can only be performed on the
* containerized resource).
*/
if ((for_first && !pcmk__str_any_of(action->task, CRMD_ACTION_STOP,
CRMD_ACTION_STOPPED, NULL))
|| (!for_first && pcmk__str_any_of(action->task, CRMD_ACTION_PROMOTE,
CRMD_ACTION_PROMOTED,
CRMD_ACTION_DEMOTE,
CRMD_ACTION_DEMOTED, NULL))) {
rsc = pe__get_rsc_in_container(instance);
}
if (rsc == NULL) {
rsc = instance; // No containerized resource, use instance itself
} else {
node = NULL; // Containerized actions are on bundle-created guest
}
matching_action = find_first_action(rsc->actions, NULL, action_name, node);
if (matching_action != NULL) {
return matching_action;
}
if (pcmk_is_set(instance->flags, pe_rsc_orphan)
|| pcmk__str_any_of(action_name, RSC_STOP, RSC_DEMOTE, NULL)) {
crm_trace("No %s action found for %s%s",
action_name,
pcmk_is_set(instance->flags, pe_rsc_orphan)? "orphan " : "",
instance->id);
} else {
crm_err("No %s action found for %s to interleave (bug?)",
action_name, instance->id);
}
return NULL;
}
/*!
* \internal
* \brief Get the original action name of a bundle or clone action
*
* Given an action for a bundle or clone, get the original action name,
* mapping notify to the action being notified, and if the instances are
* primitives, mapping completion actions to the action that was completed
* (for example, stopped to stop).
*
* \param[in] action Clone or bundle action to check
*
* \return Original action name for \p action
*/
static const char *
orig_action_name(const pe_action_t *action)
{
const pe_resource_t *instance = action->rsc->children->data; // Any instance
char *action_type = NULL;
const char *action_name = action->task;
enum action_tasks orig_task = no_action;
if (pcmk__strcase_any_of(action->task, CRMD_ACTION_NOTIFY,
CRMD_ACTION_NOTIFIED, NULL)) {
// action->uuid is RSC_(confirmed-){pre,post}_notify_ACTION_INTERVAL
CRM_CHECK(parse_op_key(action->uuid, NULL, &action_type, NULL),
return task2text(no_action));
action_name = strstr(action_type, "_notify_");
CRM_CHECK(action_name != NULL, return task2text(no_action));
action_name += strlen("_notify_");
}
orig_task = get_complex_task(instance, action_name);
free(action_type);
return task2text(orig_task);
}
/*!
* \internal
* \brief Update two interleaved actions according to an ordering between them
*
* Given information about an ordering of two interleaved actions, update the
* actions' flags (and runnable_before members if appropriate) as appropriate
* for the ordering. Effects may cascade to other orderings involving the
* actions as well.
*
* \param[in,out] first 'First' action in an ordering
* \param[in,out] then 'Then' action in an ordering
* \param[in] node If not NULL, limit scope of ordering to this node
* \param[in] filter Action flags to limit scope of certain updates (may
* include pe_action_optional to affect only mandatory
* actions, and pe_action_runnable to affect only
* runnable actions)
* \param[in] type Group of enum pe_ordering flags to apply
*
* \return Group of enum pcmk__updated flags indicating what was updated
*/
static uint32_t
update_interleaved_actions(pe_action_t *first, pe_action_t *then,
const pe_node_t *node, uint32_t filter,
uint32_t type)
{
GList *instances = NULL;
uint32_t changed = pcmk__updated_none;
const char *orig_first_task = orig_action_name(first);
// Stops and demotes must be interleaved with instance on current node
bool current = pcmk__ends_with(first->uuid, "_" CRMD_ACTION_STOPPED "_0")
|| pcmk__ends_with(first->uuid,
"_" CRMD_ACTION_DEMOTED "_0");
// Update the specified actions for each "then" instance individually
instances = get_instance_list(then->rsc);
for (GList *iter = instances; iter != NULL; iter = iter->next) {
pe_resource_t *first_instance = NULL;
pe_resource_t *then_instance = iter->data;
pe_action_t *first_action = NULL;
pe_action_t *then_action = NULL;
// Find a "first" instance to interleave with this "then" instance
first_instance = pcmk__find_compatible_instance(then_instance,
first->rsc,
RSC_ROLE_UNKNOWN,
current);
if (first_instance == NULL) { // No instance can be interleaved
if (unassign_if_mandatory(first, then, then_instance, type,
current)) {
pcmk__set_updated_flags(changed, first, pcmk__updated_then);
}
continue;
}
first_action = find_instance_action(first, first_instance,
orig_first_task, node, true);
if (first_action == NULL) {
continue;
}
then_action = find_instance_action(then, then_instance, then->task,
node, false);
if (then_action == NULL) {
continue;
}
if (order_actions(first_action, then_action, type)) {
pcmk__set_updated_flags(changed, first,
pcmk__updated_first|pcmk__updated_then);
}
changed |= then_instance->cmds->update_ordered_actions(
first_action, then_action, node,
first_instance->cmds->action_flags(first_action, node), filter,
type, then->rsc->cluster);
}
free_instance_list(then->rsc, instances);
return changed;
}
/*!
* \internal
* \brief Check whether two actions in an ordering can be interleaved
*
* \param[in] first 'First' action in the ordering
* \param[in] then 'Then' action in the ordering
*
* \return true if \p first and \p then can be interleaved, otherwise false
*/
static bool
can_interleave_actions(const pe_action_t *first, const pe_action_t *then)
{
bool interleave = false;
pe_resource_t *rsc = NULL;
if ((first->rsc == NULL) || (then->rsc == NULL)) {
crm_trace("Not interleaving %s with %s: not resource actions",
first->uuid, then->uuid);
return false;
}
if (first->rsc == then->rsc) {
crm_trace("Not interleaving %s with %s: same resource",
first->uuid, then->uuid);
return false;
}
if ((first->rsc->variant < pe_clone) || (then->rsc->variant < pe_clone)) {
crm_trace("Not interleaving %s with %s: not 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 = crm_is_true(g_hash_table_lookup(rsc->meta,
XML_RSC_ATTR_INTERLEAVE));
pe_rsc_trace(rsc, "'%s then %s' will %sbe interleaved (based on %s)",
first->uuid, then->uuid, (interleave? "" : "not "), rsc->id);
return interleave;
}
/*!
* \internal
* \brief Update non-interleaved instance actions according to an ordering
*
* Given information about an ordering of two non-interleaved actions, update
* the actions' flags (and runnable_before members if appropriate) as
* appropriate for the ordering. Effects may cascade to other orderings
* involving the actions as well.
*
* \param[in,out] instance Clone instance or bundle container
* \param[in,out] first "First" action in ordering
* \param[in] then "Then" action in ordering (for \p instance's parent)
* \param[in] node If not NULL, limit scope of ordering to this node
* \param[in] flags Action flags for \p first for ordering purposes
* \param[in] filter Action flags to limit scope of certain updates (may
* include pe_action_optional to affect only mandatory
* actions, and pe_action_runnable to affect only
* runnable actions)
* \param[in] type Group of enum pe_ordering flags to apply
*
* \return Group of enum pcmk__updated flags indicating what was updated
*/
static uint32_t
update_noninterleaved_actions(pe_resource_t *instance, pe_action_t *first,
const pe_action_t *then, const pe_node_t *node,
uint32_t flags, uint32_t filter, uint32_t type)
{
pe_action_t *instance_action = NULL;
uint32_t instance_flags = 0;
uint32_t changed = pcmk__updated_none;
// Check whether instance has an equivalent of "then" action
instance_action = find_first_action(instance->actions, NULL, then->task,
node);
if (instance_action == NULL) {
return changed;
}
// Check whether action is runnable
instance_flags = instance->cmds->action_flags(instance_action, node);
if (!pcmk_is_set(instance_flags, pe_action_runnable)) {
return changed;
}
// If so, update actions for the instance
changed = instance->cmds->update_ordered_actions(first, instance_action,
node, flags, filter, type,
instance->cluster);
// Propagate any changes to later actions
if (pcmk_is_set(changed, pcmk__updated_then)) {
for (GList *after_iter = instance_action->actions_after;
after_iter != NULL; after_iter = after_iter->next) {
pe_action_wrapper_t *after = after_iter->data;
pcmk__update_action_for_orderings(after->action, instance->cluster);
}
}
return changed;
}
/*!
* \internal
* \brief Update two actions according to an ordering between them
*
* Given information about an ordering of two clone or bundle actions, update
* the actions' flags (and runnable_before members if appropriate) as
* appropriate for the ordering. Effects may cascade to other orderings
* involving the actions as well.
*
* \param[in,out] first 'First' action in an ordering
* \param[in,out] then 'Then' action in an ordering
* \param[in] node If not NULL, limit scope of ordering to this node
* (only used when interleaving instances)
* \param[in] flags Action flags for \p first for ordering purposes
* \param[in] filter Action flags to limit scope of certain updates (may
* include pe_action_optional to affect only mandatory
* actions, and pe_action_runnable to affect only
* runnable actions)
* \param[in] type Group of enum pe_ordering flags to apply
* \param[in,out] data_set Cluster working set
*
* \return Group of enum pcmk__updated flags indicating what was updated
*/
uint32_t
pcmk__instance_update_ordered_actions(pe_action_t *first, pe_action_t *then,
const pe_node_t *node, uint32_t flags,
uint32_t filter, uint32_t type,
pe_working_set_t *data_set)
{
CRM_ASSERT((first != NULL) && (then != NULL) && (data_set != NULL));
if (then->rsc == NULL) {
return pcmk__updated_none;
} else if (can_interleave_actions(first, then)) {
return update_interleaved_actions(first, then, node, filter, type);
} else {
uint32_t changed = pcmk__updated_none;
GList *instances = get_instance_list(then->rsc);
// Update actions for the clone or bundle resource itself
changed |= pcmk__update_ordered_actions(first, then, node, flags,
filter, type, data_set);
// Update the 'then' clone instances or bundle containers individually
for (GList *iter = instances; iter != NULL; iter = iter->next) {
pe_resource_t *instance = iter->data;
changed |= update_noninterleaved_actions(instance, first, then,
node, flags, filter, type);
}
free_instance_list(then->rsc, instances);
return changed;
}
}
#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)
/*!
* \internal
* \brief Return action flags for a given clone or bundle action
*
* \param[in,out] action Action for a clone or bundle
* \param[in] instances Clone instances or bundle containers
* \param[in] node If not NULL, limit effects to this node
*
* \return Flags appropriate to \p action on \p node
*/
uint32_t
pcmk__collective_action_flags(pe_action_t *action, const GList *instances,
const pe_node_t *node)
{
bool any_runnable = false;
const char *action_name = orig_action_name(action);
// Set original assumptions (optional and runnable may be cleared below)
uint32_t flags = pe_action_optional|pe_action_runnable|pe_action_pseudo;
for (const GList *iter = instances; iter != NULL; iter = iter->next) {
const pe_resource_t *instance = iter->data;
const pe_node_t *instance_node = NULL;
pe_action_t *instance_action = NULL;
uint32_t instance_flags;
// Node is relevant only to primitive instances
if (instance->variant == pe_native) {
instance_node = node;
}
instance_action = find_first_action(instance->actions, NULL,
action_name, instance_node);
if (instance_action == NULL) {
pe_rsc_trace(action->rsc, "%s has no %s action on %s",
instance->id, action_name, pe__node_name(node));
continue;
}
pe_rsc_trace(action->rsc, "%s has %s for %s on %s",
instance->id, instance_action->uuid, action_name,
pe__node_name(node));
instance_flags = instance->cmds->action_flags(instance_action, node);
// If any instance action is mandatory, so is the collective action
if (pcmk_is_set(flags, pe_action_optional)
&& !pcmk_is_set(instance_flags, pe_action_optional)) {
pe_rsc_trace(instance, "%s is mandatory because %s is",
action->uuid, instance_action->uuid);
pe__clear_action_summary_flags(flags, action, pe_action_optional);
pe__clear_action_flags(action, pe_action_optional);
}
// If any instance action is runnable, so is the collective action
if (pcmk_is_set(instance_flags, pe_action_runnable)) {
any_runnable = true;
}
}
if (!any_runnable) {
pe_rsc_trace(action->rsc,
"%s is not runnable because no instance can run %s",
action->uuid, action_name);
pe__clear_action_summary_flags(flags, action, pe_action_runnable);
if (node == NULL) {
pe__clear_action_flags(action, pe_action_runnable);
}
}
return flags;
}
-
-/*!
- * \internal
- * \brief Add a collective resource's colocations to a list for an instance
- *
- * \param[in,out] list Colocation list to add to
- * \param[in] instance Clone or bundle instance or instance group member
- * \param[in] collective Clone or bundle resource with colocations to add
- * \param[in] with_this If true, add collective's "with this" colocations,
- * otherwise add its "this with" colocations
- */
-void
-pcmk__add_collective_constraints(GList **list, const pe_resource_t *instance,
- const pe_resource_t *collective,
- bool with_this)
-{
- const GList *colocations = NULL;
- bool everywhere = false;
-
- CRM_CHECK((list != NULL) && (instance != NULL), return);
-
- if (collective == NULL) {
- return;
- }
- switch (collective->variant) {
- case pe_clone:
- case pe_container:
- break;
- default:
- return;
- }
-
- everywhere = can_run_everywhere(collective);
-
- if (with_this) {
- colocations = collective->rsc_cons_lhs;
- } else {
- colocations = collective->rsc_cons;
- }
-
- for (const GList *iter = colocations; iter != NULL; iter = iter->next) {
- const pcmk__colocation_t *colocation = iter->data;
-
- if (with_this
- && !pcmk__colocation_has_influence(colocation, instance)) {
- continue;
- }
- if (!everywhere || (colocation->score < 0)
- || (!with_this && (colocation->score == INFINITY))) {
-
- if (with_this) {
- pcmk__add_with_this(list, colocation, instance);
- } else {
- pcmk__add_this_with(list, colocation, instance);
- }
- }
- }
-}

File Metadata

Mime Type
text/x-diff
Expires
Wed, Jun 4, 6:23 AM (6 h, 40 m)
Storage Engine
blob
Storage Format
Raw Data
Storage Handle
1854732
Default Alt Text
(169 KB)

Event Timeline