Page MenuHomeClusterLabs Projects
Files F7609267

No OneTemporary
Actions

View Options

This file is larger than 256 KB, so syntax highlighting was skipped.
diff --git a/daemons/controld/controld_membership.c b/daemons/controld/controld_membership.c
index 1f619fe1e8..75cd3e541c 100644
--- a/daemons/controld/controld_membership.c
+++ b/daemons/controld/controld_membership.c
@@ -1,468 +1,468 @@
/*
* Copyright 2004-2024 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.
*/
/* put these first so that uuid_t is defined without conflicts */
#include <crm_internal.h>
#include <string.h>
#include <crm/crm.h>
#include <crm/msg_xml.h>
#include <crm/common/xml.h>
#include <crm/common/xml_internal.h>
#include <crm/cluster/internal.h>
#include <pacemaker-controld.h>
void post_cache_update(int instance);
extern gboolean check_join_state(enum crmd_fsa_state cur_state, const char *source);
static void
reap_dead_nodes(gpointer key, gpointer value, gpointer user_data)
{
crm_node_t *node = value;
if (crm_is_peer_active(node) == FALSE) {
crm_update_peer_join(__func__, node, crm_join_none);
if(node && node->uname) {
if (pcmk__str_eq(controld_globals.our_nodename, node->uname,
pcmk__str_casei)) {
crm_err("We're not part of the cluster anymore");
register_fsa_input(C_FSA_INTERNAL, I_ERROR, NULL);
} else if (!AM_I_DC
&& pcmk__str_eq(node->uname, controld_globals.dc_name,
pcmk__str_casei)) {
crm_warn("Our DC node (%s) left the cluster", node->uname);
register_fsa_input(C_FSA_INTERNAL, I_ELECTION, NULL);
}
}
if ((controld_globals.fsa_state == S_INTEGRATION)
|| (controld_globals.fsa_state == S_FINALIZE_JOIN)) {
check_join_state(controld_globals.fsa_state, __func__);
}
if ((node != NULL) && (node->uuid != NULL)) {
fail_incompletable_actions(controld_globals.transition_graph,
node->uuid);
}
}
}
void
post_cache_update(int instance)
{
xmlNode *no_op = NULL;
crm_peer_seq = instance;
crm_debug("Updated cache after membership event %d.", instance);
g_hash_table_foreach(crm_peer_cache, reap_dead_nodes, NULL);
controld_set_fsa_input_flags(R_MEMBERSHIP);
if (AM_I_DC) {
populate_cib_nodes(node_update_quick | node_update_cluster | node_update_peer |
node_update_expected, __func__);
}
/*
* If we lost nodes, we should re-check the election status
* Safe to call outside of an election
*/
controld_set_fsa_action_flags(A_ELECTION_CHECK);
controld_trigger_fsa();
/* Membership changed, remind everyone we're here.
* This will aid detection of duplicate DCs
*/
no_op = create_request(CRM_OP_NOOP, NULL, NULL, CRM_SYSTEM_CRMD,
AM_I_DC ? CRM_SYSTEM_DC : CRM_SYSTEM_CRMD, NULL);
send_cluster_message(NULL, crm_msg_crmd, no_op, FALSE);
free_xml(no_op);
}
static void
crmd_node_update_complete(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data)
{
fsa_data_t *msg_data = NULL;
if (rc == pcmk_ok) {
crm_trace("Node update %d complete", call_id);
} else if(call_id < pcmk_ok) {
crm_err("Node update failed: %s (%d)", pcmk_strerror(call_id), call_id);
crm_log_xml_debug(msg, "failed");
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
} else {
crm_err("Node update %d failed: %s (%d)", call_id, pcmk_strerror(rc), rc);
crm_log_xml_debug(msg, "failed");
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
}
}
/*!
* \internal
* \brief Create an XML node state tag with updates
*
* \param[in,out] node Node whose state will be used for update
* \param[in] flags Bitmask of node_update_flags indicating what to update
* \param[in,out] parent XML node to contain update (or NULL)
* \param[in] source Who requested the update (only used for logging)
*
* \return Pointer to created node state tag
*/
xmlNode *
create_node_state_update(crm_node_t *node, int flags, xmlNode *parent,
const char *source)
{
const char *value = NULL;
xmlNode *node_state;
if (!node->state) {
crm_info("Node update for %s cancelled: no state, not seen yet", node->uname);
return NULL;
}
node_state = create_xml_node(parent, PCMK__XE_NODE_STATE);
if (pcmk_is_set(node->flags, crm_remote_node)) {
pcmk__xe_set_bool_attr(node_state, PCMK__XA_REMOTE_NODE, true);
}
if (crm_xml_add(node_state, PCMK_XA_ID, crm_peer_uuid(node)) == NULL) {
crm_info("Node update for %s cancelled: no ID", node->uname);
free_xml(node_state);
return NULL;
}
crm_xml_add(node_state, PCMK_XA_UNAME, node->uname);
if ((flags & node_update_cluster) && node->state) {
if (compare_version(controld_globals.dc_version, "3.18.0") >= 0) {
// A value 0 means the node is not a cluster member.
crm_xml_add_ll(node_state, PCMK__XA_IN_CCM, node->when_member);
} else {
pcmk__xe_set_bool_attr(node_state, PCMK__XA_IN_CCM,
pcmk__str_eq(node->state, CRM_NODE_MEMBER,
pcmk__str_casei));
}
}
if (!pcmk_is_set(node->flags, crm_remote_node)) {
if (flags & node_update_peer) {
if (compare_version(controld_globals.dc_version, "3.18.0") >= 0) {
// A value 0 means the peer is offline in CPG.
- crm_xml_add_ll(node_state, PCMK__XA_CRMD, node->when_online);
+ crm_xml_add_ll(node_state, PCMK_XA_CRMD, node->when_online);
} else {
// @COMPAT DCs < 2.1.7 use online/offline rather than timestamp
value = OFFLINESTATUS;
if (pcmk_is_set(node->processes, crm_get_cluster_proc())) {
value = ONLINESTATUS;
}
- crm_xml_add(node_state, PCMK__XA_CRMD, value);
+ crm_xml_add(node_state, PCMK_XA_CRMD, value);
}
}
if (flags & node_update_join) {
if (node->join <= crm_join_none) {
value = CRMD_JOINSTATE_DOWN;
} else {
value = CRMD_JOINSTATE_MEMBER;
}
crm_xml_add(node_state, PCMK__XA_JOIN, value);
}
if (flags & node_update_expected) {
crm_xml_add(node_state, PCMK_XA_EXPECTED, node->expected);
}
}
crm_xml_add(node_state, PCMK_XA_CRM_DEBUG_ORIGIN, source);
return node_state;
}
static void
remove_conflicting_node_callback(xmlNode * msg, int call_id, int rc,
xmlNode * output, void *user_data)
{
char *node_uuid = user_data;
do_crm_log_unlikely(rc == 0 ? LOG_DEBUG : LOG_NOTICE,
"Deletion of the unknown conflicting node \"%s\": %s (rc=%d)",
node_uuid, pcmk_strerror(rc), rc);
}
static void
search_conflicting_node_callback(xmlNode * msg, int call_id, int rc,
xmlNode * output, void *user_data)
{
char *new_node_uuid = user_data;
xmlNode *node_xml = NULL;
if (rc != pcmk_ok) {
if (rc != -ENXIO) {
crm_notice("Searching conflicting nodes for %s failed: %s (%d)",
new_node_uuid, pcmk_strerror(rc), rc);
}
return;
} else if (output == NULL) {
return;
}
if (pcmk__xe_is(output, PCMK_XE_NODE)) {
node_xml = output;
} else {
node_xml = pcmk__xml_first_child(output);
}
for (; node_xml != NULL; node_xml = pcmk__xml_next(node_xml)) {
const char *node_uuid = NULL;
const char *node_uname = NULL;
GHashTableIter iter;
crm_node_t *node = NULL;
gboolean known = FALSE;
if (!pcmk__xe_is(node_xml, PCMK_XE_NODE)) {
continue;
}
node_uuid = crm_element_value(node_xml, PCMK_XA_ID);
node_uname = crm_element_value(node_xml, PCMK_XA_UNAME);
if (node_uuid == NULL || node_uname == NULL) {
continue;
}
g_hash_table_iter_init(&iter, crm_peer_cache);
while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) {
if (node->uuid
&& pcmk__str_eq(node->uuid, node_uuid, pcmk__str_casei)
&& node->uname
&& pcmk__str_eq(node->uname, node_uname, pcmk__str_casei)) {
known = TRUE;
break;
}
}
if (known == FALSE) {
cib_t *cib_conn = controld_globals.cib_conn;
int delete_call_id = 0;
xmlNode *node_state_xml = NULL;
crm_notice("Deleting unknown node %s/%s which has conflicting uname with %s",
node_uuid, node_uname, new_node_uuid);
delete_call_id = cib_conn->cmds->remove(cib_conn, PCMK_XE_NODES,
node_xml, cib_scope_local);
fsa_register_cib_callback(delete_call_id, strdup(node_uuid),
remove_conflicting_node_callback);
node_state_xml = create_xml_node(NULL, PCMK__XE_NODE_STATE);
crm_xml_add(node_state_xml, PCMK_XA_ID, node_uuid);
crm_xml_add(node_state_xml, PCMK_XA_UNAME, node_uname);
delete_call_id = cib_conn->cmds->remove(cib_conn, PCMK_XE_STATUS,
node_state_xml,
cib_scope_local);
fsa_register_cib_callback(delete_call_id, strdup(node_uuid),
remove_conflicting_node_callback);
free_xml(node_state_xml);
}
}
}
static void
node_list_update_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data)
{
fsa_data_t *msg_data = NULL;
if(call_id < pcmk_ok) {
crm_err("Node list update failed: %s (%d)", pcmk_strerror(call_id), call_id);
crm_log_xml_debug(msg, "update:failed");
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
} else if(rc < pcmk_ok) {
crm_err("Node update %d failed: %s (%d)", call_id, pcmk_strerror(rc), rc);
crm_log_xml_debug(msg, "update:failed");
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
}
}
void
populate_cib_nodes(enum node_update_flags flags, const char *source)
{
cib_t *cib_conn = controld_globals.cib_conn;
int call_id = 0;
gboolean from_hashtable = TRUE;
xmlNode *node_list = create_xml_node(NULL, PCMK_XE_NODES);
#if SUPPORT_COROSYNC
if (!pcmk_is_set(flags, node_update_quick) && is_corosync_cluster()) {
from_hashtable = pcmk__corosync_add_nodes(node_list);
}
#endif
if (from_hashtable) {
GHashTableIter iter;
crm_node_t *node = NULL;
GString *xpath = NULL;
g_hash_table_iter_init(&iter, crm_peer_cache);
while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) {
xmlNode *new_node = NULL;
if ((node->uuid != NULL) && (node->uname != NULL)) {
crm_trace("Creating node entry for %s/%s", node->uname, node->uuid);
if (xpath == NULL) {
xpath = g_string_sized_new(512);
} else {
g_string_truncate(xpath, 0);
}
/* We need both to be valid */
new_node = create_xml_node(node_list, PCMK_XE_NODE);
crm_xml_add(new_node, PCMK_XA_ID, node->uuid);
crm_xml_add(new_node, PCMK_XA_UNAME, node->uname);
/* Search and remove unknown nodes with the conflicting uname from CIB */
pcmk__g_strcat(xpath,
"/" PCMK_XE_CIB "/" PCMK_XE_CONFIGURATION
"/" PCMK_XE_NODES "/" PCMK_XE_NODE
"[@" PCMK_XA_UNAME "='", node->uname, "']"
"[@" PCMK_XA_ID "!='", node->uuid, "']", NULL);
call_id = cib_conn->cmds->query(cib_conn,
(const char *) xpath->str,
NULL,
cib_scope_local|cib_xpath);
fsa_register_cib_callback(call_id, strdup(node->uuid),
search_conflicting_node_callback);
}
}
if (xpath != NULL) {
g_string_free(xpath, TRUE);
}
}
crm_trace("Populating <nodes> section from %s", from_hashtable ? "hashtable" : "cluster");
if ((controld_update_cib(PCMK_XE_NODES, node_list, cib_scope_local,
node_list_update_callback) == pcmk_rc_ok)
&& (crm_peer_cache != NULL) && AM_I_DC) {
/*
* There is no need to update the local CIB with our values if
* we've not seen valid membership data
*/
GHashTableIter iter;
crm_node_t *node = NULL;
free_xml(node_list);
node_list = create_xml_node(NULL, PCMK_XE_STATUS);
g_hash_table_iter_init(&iter, crm_peer_cache);
while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) {
create_node_state_update(node, flags, node_list, source);
}
if (crm_remote_peer_cache) {
g_hash_table_iter_init(&iter, crm_remote_peer_cache);
while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) {
create_node_state_update(node, flags, node_list, source);
}
}
controld_update_cib(PCMK_XE_STATUS, node_list, cib_scope_local,
crmd_node_update_complete);
}
free_xml(node_list);
}
static void
cib_quorum_update_complete(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data)
{
fsa_data_t *msg_data = NULL;
if (rc == pcmk_ok) {
crm_trace("Quorum update %d complete", call_id);
} else {
crm_err("Quorum update %d failed: %s (%d)", call_id, pcmk_strerror(rc), rc);
crm_log_xml_debug(msg, "failed");
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
}
}
void
crm_update_quorum(gboolean quorum, gboolean force_update)
{
bool has_quorum = pcmk_is_set(controld_globals.flags, controld_has_quorum);
if (quorum) {
controld_set_global_flags(controld_ever_had_quorum);
} else if (pcmk_all_flags_set(controld_globals.flags,
controld_ever_had_quorum
|controld_no_quorum_suicide)) {
pcmk__panic(__func__);
}
if (AM_I_DC
&& ((has_quorum && !quorum) || (!has_quorum && quorum)
|| force_update)) {
xmlNode *update = NULL;
update = create_xml_node(NULL, PCMK_XE_CIB);
crm_xml_add_int(update, PCMK_XA_HAVE_QUORUM, quorum);
crm_xml_add(update, PCMK_XA_DC_UUID, controld_globals.our_uuid);
crm_debug("Updating quorum status to %s", pcmk__btoa(quorum));
controld_update_cib(PCMK_XE_CIB, update, cib_scope_local,
cib_quorum_update_complete);
free_xml(update);
/* Quorum changes usually cause a new transition via other activity:
* quorum gained via a node joining will abort via the node join,
* and quorum lost via a node leaving will usually abort via resource
* activity and/or fencing.
*
* However, it is possible that nothing else causes a transition (e.g.
* someone forces quorum via corosync-cmaptcl, or quorum is lost due to
* a node in standby shutting down cleanly), so here ensure a new
* transition is triggered.
*/
if (quorum) {
/* If quorum was gained, abort after a short delay, in case multiple
* nodes are joining around the same time, so the one that brings us
* to quorum doesn't cause all the remaining ones to be fenced.
*/
abort_after_delay(INFINITY, pcmk__graph_restart, "Quorum gained",
5000);
} else {
abort_transition(INFINITY, pcmk__graph_restart, "Quorum lost",
NULL);
}
}
if (quorum) {
controld_set_global_flags(controld_has_quorum);
} else {
controld_clear_global_flags(controld_has_quorum);
}
}
diff --git a/daemons/controld/controld_messages.c b/daemons/controld/controld_messages.c
index 676a94fa62..1c56722bba 100644
--- a/daemons/controld/controld_messages.c
+++ b/daemons/controld/controld_messages.c
@@ -1,1337 +1,1337 @@
/*
* Copyright 2004-2024 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 <sys/param.h>
#include <string.h>
#include <time.h>
#include <crm/crm.h>
#include <crm/msg_xml.h>
#include <crm/common/xml.h>
#include <crm/cluster/internal.h>
#include <crm/cib.h>
#include <crm/common/ipc_internal.h>
#include <pacemaker-controld.h>
static enum crmd_fsa_input handle_message(xmlNode *msg,
enum crmd_fsa_cause cause);
static void handle_response(xmlNode *stored_msg);
static enum crmd_fsa_input handle_request(xmlNode *stored_msg,
enum crmd_fsa_cause cause);
static enum crmd_fsa_input handle_shutdown_request(xmlNode *stored_msg);
static void send_msg_via_ipc(xmlNode * msg, const char *sys);
/* debug only, can wrap all it likes */
static int last_data_id = 0;
void
register_fsa_error_adv(enum crmd_fsa_cause cause, enum crmd_fsa_input input,
fsa_data_t * cur_data, void *new_data, const char *raised_from)
{
/* save the current actions if any */
if (controld_globals.fsa_actions != A_NOTHING) {
register_fsa_input_adv(cur_data ? cur_data->fsa_cause : C_FSA_INTERNAL,
I_NULL, cur_data ? cur_data->data : NULL,
controld_globals.fsa_actions, TRUE, __func__);
}
/* reset the action list */
crm_info("Resetting the current action list");
fsa_dump_actions(controld_globals.fsa_actions, "Drop");
controld_globals.fsa_actions = A_NOTHING;
/* register the error */
register_fsa_input_adv(cause, input, new_data, A_NOTHING, TRUE, raised_from);
}
void
register_fsa_input_adv(enum crmd_fsa_cause cause, enum crmd_fsa_input input,
void *data, uint64_t with_actions,
gboolean prepend, const char *raised_from)
{
unsigned old_len = g_list_length(controld_globals.fsa_message_queue);
fsa_data_t *fsa_data = NULL;
if (raised_from == NULL) {
raised_from = "<unknown>";
}
if (input == I_NULL && with_actions == A_NOTHING /* && data == NULL */ ) {
/* no point doing anything */
crm_err("Cannot add entry to queue: no input and no action");
return;
}
if (input == I_WAIT_FOR_EVENT) {
controld_set_global_flags(controld_fsa_is_stalled);
crm_debug("Stalling the FSA pending further input: source=%s cause=%s data=%p queue=%d",
raised_from, fsa_cause2string(cause), data, old_len);
if (old_len > 0) {
fsa_dump_queue(LOG_TRACE);
prepend = FALSE;
}
if (data == NULL) {
controld_set_fsa_action_flags(with_actions);
fsa_dump_actions(with_actions, "Restored");
return;
}
/* Store everything in the new event and reset
* controld_globals.fsa_actions
*/
with_actions |= controld_globals.fsa_actions;
controld_globals.fsa_actions = A_NOTHING;
}
last_data_id++;
crm_trace("%s %s FSA input %d (%s) due to %s, %s data",
raised_from, (prepend? "prepended" : "appended"), last_data_id,
fsa_input2string(input), fsa_cause2string(cause),
(data? "with" : "without"));
fsa_data = calloc(1, sizeof(fsa_data_t));
fsa_data->id = last_data_id;
fsa_data->fsa_input = input;
fsa_data->fsa_cause = cause;
fsa_data->origin = raised_from;
fsa_data->data = NULL;
fsa_data->data_type = fsa_dt_none;
fsa_data->actions = with_actions;
if (with_actions != A_NOTHING) {
crm_trace("Adding actions %.16llx to input",
(unsigned long long) with_actions);
}
if (data != NULL) {
switch (cause) {
case C_FSA_INTERNAL:
case C_CRMD_STATUS_CALLBACK:
case C_IPC_MESSAGE:
case C_HA_MESSAGE:
CRM_CHECK(((ha_msg_input_t *) data)->msg != NULL,
crm_err("Bogus data from %s", raised_from));
crm_trace("Copying %s data from %s as cluster message data",
fsa_cause2string(cause), raised_from);
fsa_data->data = copy_ha_msg_input(data);
fsa_data->data_type = fsa_dt_ha_msg;
break;
case C_LRM_OP_CALLBACK:
crm_trace("Copying %s data from %s as lrmd_event_data_t",
fsa_cause2string(cause), raised_from);
fsa_data->data = lrmd_copy_event((lrmd_event_data_t *) data);
fsa_data->data_type = fsa_dt_lrm;
break;
case C_TIMER_POPPED:
case C_SHUTDOWN:
case C_UNKNOWN:
case C_STARTUP:
crm_crit("Copying %s data (from %s) is not yet implemented",
fsa_cause2string(cause), raised_from);
crmd_exit(CRM_EX_SOFTWARE);
break;
}
}
/* make sure to free it properly later */
if (prepend) {
controld_globals.fsa_message_queue
= g_list_prepend(controld_globals.fsa_message_queue, fsa_data);
} else {
controld_globals.fsa_message_queue
= g_list_append(controld_globals.fsa_message_queue, fsa_data);
}
crm_trace("FSA message queue length is %d",
g_list_length(controld_globals.fsa_message_queue));
/* fsa_dump_queue(LOG_TRACE); */
if (old_len == g_list_length(controld_globals.fsa_message_queue)) {
crm_err("Couldn't add message to the queue");
}
if (input != I_WAIT_FOR_EVENT) {
controld_trigger_fsa();
}
}
void
fsa_dump_queue(int log_level)
{
int offset = 0;
for (GList *iter = controld_globals.fsa_message_queue; iter != NULL;
iter = iter->next) {
fsa_data_t *data = (fsa_data_t *) iter->data;
do_crm_log_unlikely(log_level,
"queue[%d.%d]: input %s raised by %s(%p.%d)\t(cause=%s)",
offset++, data->id, fsa_input2string(data->fsa_input),
data->origin, data->data, data->data_type,
fsa_cause2string(data->fsa_cause));
}
}
ha_msg_input_t *
copy_ha_msg_input(ha_msg_input_t * orig)
{
ha_msg_input_t *copy = calloc(1, sizeof(ha_msg_input_t));
CRM_ASSERT(copy != NULL);
copy->msg = (orig && orig->msg)? copy_xml(orig->msg) : NULL;
copy->xml = get_message_xml(copy->msg, PCMK__XE_CRM_XML);
return copy;
}
void
delete_fsa_input(fsa_data_t * fsa_data)
{
lrmd_event_data_t *op = NULL;
xmlNode *foo = NULL;
if (fsa_data == NULL) {
return;
}
crm_trace("About to free %s data", fsa_cause2string(fsa_data->fsa_cause));
if (fsa_data->data != NULL) {
switch (fsa_data->data_type) {
case fsa_dt_ha_msg:
delete_ha_msg_input(fsa_data->data);
break;
case fsa_dt_xml:
foo = fsa_data->data;
free_xml(foo);
break;
case fsa_dt_lrm:
op = (lrmd_event_data_t *) fsa_data->data;
lrmd_free_event(op);
break;
case fsa_dt_none:
if (fsa_data->data != NULL) {
crm_err("Don't know how to free %s data from %s",
fsa_cause2string(fsa_data->fsa_cause), fsa_data->origin);
crmd_exit(CRM_EX_SOFTWARE);
}
break;
}
crm_trace("%s data freed", fsa_cause2string(fsa_data->fsa_cause));
}
free(fsa_data);
}
/* returns the next message */
fsa_data_t *
get_message(void)
{
fsa_data_t *message
= (fsa_data_t *) controld_globals.fsa_message_queue->data;
controld_globals.fsa_message_queue
= g_list_remove(controld_globals.fsa_message_queue, message);
crm_trace("Processing input %d", message->id);
return message;
}
void *
fsa_typed_data_adv(fsa_data_t * fsa_data, enum fsa_data_type a_type, const char *caller)
{
void *ret_val = NULL;
if (fsa_data == NULL) {
crm_err("%s: No FSA data available", caller);
} else if (fsa_data->data == NULL) {
crm_err("%s: No message data available. Origin: %s", caller, fsa_data->origin);
} else if (fsa_data->data_type != a_type) {
crm_crit("%s: Message data was the wrong type! %d vs. requested=%d. Origin: %s",
caller, fsa_data->data_type, a_type, fsa_data->origin);
CRM_ASSERT(fsa_data->data_type == a_type);
} else {
ret_val = fsa_data->data;
}
return ret_val;
}
/* A_MSG_ROUTE */
void
do_msg_route(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state,
enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
ha_msg_input_t *input = fsa_typed_data(fsa_dt_ha_msg);
route_message(msg_data->fsa_cause, input->msg);
}
void
route_message(enum crmd_fsa_cause cause, xmlNode * input)
{
ha_msg_input_t fsa_input;
enum crmd_fsa_input result = I_NULL;
fsa_input.msg = input;
CRM_CHECK(cause == C_IPC_MESSAGE || cause == C_HA_MESSAGE, return);
/* try passing the buck first */
if (relay_message(input, cause == C_IPC_MESSAGE)) {
return;
}
/* handle locally */
result = handle_message(input, cause);
/* done or process later? */
switch (result) {
case I_NULL:
case I_CIB_OP:
case I_ROUTER:
case I_NODE_JOIN:
case I_JOIN_REQUEST:
case I_JOIN_RESULT:
break;
default:
/* Defering local processing of message */
register_fsa_input_later(cause, result, &fsa_input);
return;
}
if (result != I_NULL) {
/* add to the front of the queue */
register_fsa_input(cause, result, &fsa_input);
}
}
gboolean
relay_message(xmlNode * msg, gboolean originated_locally)
{
enum crm_ais_msg_types dest = crm_msg_ais;
bool is_for_dc = false;
bool is_for_dcib = false;
bool is_for_te = false;
bool is_for_crm = false;
bool is_for_cib = false;
bool is_local = false;
bool broadcast = false;
const char *host_to = NULL;
const char *sys_to = NULL;
const char *sys_from = NULL;
const char *type = NULL;
const char *task = NULL;
const char *ref = NULL;
crm_node_t *node_to = NULL;
CRM_CHECK(msg != NULL, return TRUE);
host_to = crm_element_value(msg, PCMK__XA_CRM_HOST_TO);
sys_to = crm_element_value(msg, PCMK__XA_CRM_SYS_TO);
sys_from = crm_element_value(msg, PCMK__XA_CRM_SYS_FROM);
type = crm_element_value(msg, PCMK__XA_T);
task = crm_element_value(msg, PCMK__XA_CRM_TASK);
ref = crm_element_value(msg, PCMK_XA_REFERENCE);
broadcast = pcmk__str_empty(host_to);
if (ref == NULL) {
ref = "without reference ID";
}
if (pcmk__str_eq(task, CRM_OP_HELLO, pcmk__str_casei)) {
crm_trace("Received hello %s from %s (no processing needed)",
ref, pcmk__s(sys_from, "unidentified source"));
crm_log_xml_trace(msg, "hello");
return TRUE;
}
// Require message type (set by create_request())
if (!pcmk__str_eq(type, PCMK__VALUE_CRMD, pcmk__str_none)) {
crm_warn("Ignoring invalid message %s with type '%s' "
"(not '" PCMK__VALUE_CRMD "')",
ref, pcmk__s(type, ""));
crm_log_xml_trace(msg, "ignored");
return TRUE;
}
// Require a destination subsystem (also set by create_request())
if (sys_to == NULL) {
crm_warn("Ignoring invalid message %s with no " PCMK__XA_CRM_SYS_TO,
ref);
crm_log_xml_trace(msg, "ignored");
return TRUE;
}
// Get the message type appropriate to the destination subsystem
if (is_corosync_cluster()) {
dest = text2msg_type(sys_to);
if ((dest < crm_msg_ais) || (dest > crm_msg_stonith_ng)) {
/* Unrecognized value, use a sane default
*
* @TODO Maybe we should bail instead
*/
dest = crm_msg_crmd;
}
}
is_for_dc = (strcasecmp(CRM_SYSTEM_DC, sys_to) == 0);
is_for_dcib = (strcasecmp(CRM_SYSTEM_DCIB, sys_to) == 0);
is_for_te = (strcasecmp(CRM_SYSTEM_TENGINE, sys_to) == 0);
is_for_cib = (strcasecmp(CRM_SYSTEM_CIB, sys_to) == 0);
is_for_crm = (strcasecmp(CRM_SYSTEM_CRMD, sys_to) == 0);
// Check whether message should be processed locally
is_local = false;
if (broadcast) {
if (is_for_dc || is_for_te) {
is_local = false;
} else if (is_for_crm) {
if (pcmk__strcase_any_of(task, CRM_OP_NODE_INFO,
PCMK__CONTROLD_CMD_NODES, NULL)) {
/* Node info requests do not specify a host, which is normally
* treated as "all hosts", because the whole point is that the
* client may not know the local node name. Always handle these
* requests locally.
*/
is_local = true;
} else {
is_local = !originated_locally;
}
} else {
is_local = true;
}
} else if (pcmk__str_eq(controld_globals.our_nodename, host_to,
pcmk__str_casei)) {
is_local = true;
} else if (is_for_crm && pcmk__str_eq(task, CRM_OP_LRM_DELETE, pcmk__str_casei)) {
xmlNode *msg_data = get_message_xml(msg, PCMK__XE_CRM_XML);
const char *mode = crm_element_value(msg_data, PCMK__XA_MODE);
if (pcmk__str_eq(mode, PCMK__VALUE_CIB, pcmk__str_none)) {
// Local delete of an offline node's resource history
is_local = true;
}
}
// Check whether message should be relayed
if (is_for_dc || is_for_dcib || is_for_te) {
if (AM_I_DC) {
if (is_for_te) {
crm_trace("Route message %s locally as transition request",
ref);
crm_log_xml_trace(msg, sys_to);
send_msg_via_ipc(msg, sys_to);
return TRUE; // No further processing of message is needed
}
crm_trace("Route message %s locally as DC request", ref);
return FALSE; // More to be done by caller
}
if (originated_locally
&& !pcmk__strcase_any_of(sys_from, CRM_SYSTEM_PENGINE,
CRM_SYSTEM_TENGINE, NULL)) {
crm_trace("Relay message %s to DC (via %s)",
ref, pcmk__s(host_to, "broadcast"));
crm_log_xml_trace(msg, "relayed");
if (!broadcast) {
node_to = pcmk__get_node(0, host_to, NULL,
pcmk__node_search_cluster);
}
send_cluster_message(node_to, dest, msg, TRUE);
return TRUE;
}
/* Transition engine and scheduler messages are sent only to the DC on
* the same node. If we are no longer the DC, discard this message.
*/
crm_trace("Ignoring message %s because we are no longer DC", ref);
crm_log_xml_trace(msg, "ignored");
return TRUE; // No further processing of message is needed
}
if (is_local) {
if (is_for_crm || is_for_cib) {
crm_trace("Route message %s locally as controller request", ref);
return FALSE; // More to be done by caller
}
crm_trace("Relay message %s locally to %s", ref, sys_to);
crm_log_xml_trace(msg, "IPC-relay");
send_msg_via_ipc(msg, sys_to);
return TRUE;
}
if (!broadcast) {
node_to = pcmk__search_node_caches(0, host_to,
pcmk__node_search_cluster);
if (node_to == NULL) {
crm_warn("Ignoring message %s because node %s is unknown",
ref, host_to);
crm_log_xml_trace(msg, "ignored");
return TRUE;
}
}
crm_trace("Relay message %s to %s",
ref, pcmk__s(host_to, "all peers"));
crm_log_xml_trace(msg, "relayed");
send_cluster_message(node_to, dest, msg, TRUE);
return TRUE;
}
// Return true if field contains a positive integer
static bool
authorize_version(xmlNode *message_data, const char *field,
const char *client_name, const char *ref, const char *uuid)
{
const char *version = crm_element_value(message_data, field);
long long version_num;
if ((pcmk__scan_ll(version, &version_num, -1LL) != pcmk_rc_ok)
|| (version_num < 0LL)) {
crm_warn("Rejected IPC hello from %s: '%s' is not a valid protocol %s "
CRM_XS " ref=%s uuid=%s",
client_name, ((version == NULL)? "" : version),
field, (ref? ref : "none"), uuid);
return false;
}
return true;
}
/*!
* \internal
* \brief Check whether a client IPC message is acceptable
*
* If a given client IPC message is a hello, "authorize" it by ensuring it has
* valid information such as a protocol version, and return false indicating
* that nothing further needs to be done with the message. If the message is not
* a hello, just return true to indicate it needs further processing.
*
* \param[in] client_msg XML of IPC message
* \param[in,out] curr_client If IPC is not proxied, client that sent message
* \param[in] proxy_session If IPC is proxied, the session ID
*
* \return true if message needs further processing, false if it doesn't
*/
bool
controld_authorize_ipc_message(const xmlNode *client_msg, pcmk__client_t *curr_client,
const char *proxy_session)
{
xmlNode *message_data = NULL;
const char *client_name = NULL;
const char *op = crm_element_value(client_msg, PCMK__XA_CRM_TASK);
const char *ref = crm_element_value(client_msg, PCMK_XA_REFERENCE);
const char *uuid = (curr_client? curr_client->id : proxy_session);
if (uuid == NULL) {
crm_warn("IPC message from client rejected: No client identifier "
CRM_XS " ref=%s", (ref? ref : "none"));
goto rejected;
}
if (!pcmk__str_eq(CRM_OP_HELLO, op, pcmk__str_casei)) {
// Only hello messages need to be authorized
return true;
}
message_data = get_message_xml(client_msg, PCMK__XE_CRM_XML);
client_name = crm_element_value(message_data, PCMK__XA_CLIENT_NAME);
if (pcmk__str_empty(client_name)) {
crm_warn("IPC hello from client rejected: No client name",
CRM_XS " ref=%s uuid=%s", (ref? ref : "none"), uuid);
goto rejected;
}
if (!authorize_version(message_data, PCMK__XA_MAJOR_VERSION, client_name,
ref, uuid)) {
goto rejected;
}
if (!authorize_version(message_data, PCMK__XA_MINOR_VERSION, client_name,
ref, uuid)) {
goto rejected;
}
crm_trace("Validated IPC hello from client %s", client_name);
crm_log_xml_trace(client_msg, "hello");
if (curr_client) {
curr_client->userdata = strdup(client_name);
}
controld_trigger_fsa();
return false;
rejected:
crm_log_xml_trace(client_msg, "rejected");
if (curr_client) {
qb_ipcs_disconnect(curr_client->ipcs);
}
return false;
}
static enum crmd_fsa_input
handle_message(xmlNode *msg, enum crmd_fsa_cause cause)
{
const char *type = NULL;
CRM_CHECK(msg != NULL, return I_NULL);
type = crm_element_value(msg, PCMK__XA_SUBT);
if (pcmk__str_eq(type, PCMK__VALUE_REQUEST, pcmk__str_none)) {
return handle_request(msg, cause);
}
if (pcmk__str_eq(type, PCMK__VALUE_RESPONSE, pcmk__str_none)) {
handle_response(msg);
return I_NULL;
}
crm_warn("Ignoring message with unknown " PCMK__XA_SUBT" '%s'",
pcmk__s(type, ""));
crm_log_xml_trace(msg, "bad");
return I_NULL;
}
static enum crmd_fsa_input
handle_failcount_op(xmlNode * stored_msg)
{
const char *rsc = NULL;
const char *uname = NULL;
const char *op = NULL;
char *interval_spec = NULL;
guint interval_ms = 0;
gboolean is_remote_node = FALSE;
xmlNode *xml_op = get_message_xml(stored_msg, PCMK__XE_CRM_XML);
if (xml_op) {
xmlNode *xml_rsc = first_named_child(xml_op, PCMK_XE_PRIMITIVE);
xmlNode *xml_attrs = first_named_child(xml_op, PCMK__XE_ATTRIBUTES);
if (xml_rsc) {
rsc = ID(xml_rsc);
}
if (xml_attrs) {
op = crm_element_value(xml_attrs,
CRM_META "_" PCMK__META_CLEAR_FAILURE_OP);
crm_element_value_ms(xml_attrs,
CRM_META "_" PCMK__META_CLEAR_FAILURE_INTERVAL,
&interval_ms);
}
}
uname = crm_element_value(xml_op, PCMK__META_ON_NODE);
if ((rsc == NULL) || (uname == NULL)) {
crm_log_xml_warn(stored_msg, "invalid failcount op");
return I_NULL;
}
if (crm_element_value(xml_op, PCMK__XA_ROUTER_NODE)) {
is_remote_node = TRUE;
}
crm_debug("Clearing failures for %s-interval %s on %s "
"from attribute manager, CIB, and executor state",
pcmk__readable_interval(interval_ms), rsc, uname);
if (interval_ms) {
interval_spec = crm_strdup_printf("%ums", interval_ms);
}
update_attrd_clear_failures(uname, rsc, op, interval_spec, is_remote_node);
free(interval_spec);
controld_cib_delete_last_failure(rsc, uname, op, interval_ms);
lrm_clear_last_failure(rsc, uname, op, interval_ms);
return I_NULL;
}
static enum crmd_fsa_input
handle_lrm_delete(xmlNode *stored_msg)
{
const char *mode = NULL;
xmlNode *msg_data = get_message_xml(stored_msg, PCMK__XE_CRM_XML);
CRM_CHECK(msg_data != NULL, return I_NULL);
/* CRM_OP_LRM_DELETE has two distinct modes. The default behavior is to
* relay the operation to the affected node, which will unregister the
* resource from the local executor, clear the resource's history from the
* CIB, and do some bookkeeping in the controller.
*
* However, if the affected node is offline, the client will specify
* mode=PCMK__VALUE_CIB which means the controller receiving the operation
* should clear the resource's history from the CIB and nothing else. This
* is used to clear shutdown locks.
*/
mode = crm_element_value(msg_data, PCMK__XA_MODE);
if (!pcmk__str_eq(mode, PCMK__VALUE_CIB, pcmk__str_none)) {
// Relay to affected node
crm_xml_add(stored_msg, PCMK__XA_CRM_SYS_TO, CRM_SYSTEM_LRMD);
return I_ROUTER;
} else {
// Delete CIB history locally (compare with do_lrm_delete())
const char *from_sys = NULL;
const char *user_name = NULL;
const char *rsc_id = NULL;
const char *node = NULL;
xmlNode *rsc_xml = NULL;
int rc = pcmk_rc_ok;
rsc_xml = first_named_child(msg_data, PCMK_XE_PRIMITIVE);
CRM_CHECK(rsc_xml != NULL, return I_NULL);
rsc_id = ID(rsc_xml);
from_sys = crm_element_value(stored_msg, PCMK__XA_CRM_SYS_FROM);
node = crm_element_value(msg_data, PCMK__META_ON_NODE);
user_name = pcmk__update_acl_user(stored_msg, PCMK__XA_CRM_USER, NULL);
crm_debug("Handling " CRM_OP_LRM_DELETE " for %s on %s locally%s%s "
"(clearing CIB resource history only)", rsc_id, node,
(user_name? " for user " : ""), (user_name? user_name : ""));
rc = controld_delete_resource_history(rsc_id, node, user_name,
cib_dryrun|cib_sync_call);
if (rc == pcmk_rc_ok) {
rc = controld_delete_resource_history(rsc_id, node, user_name,
crmd_cib_smart_opt());
}
/* Notify client. Also notify tengine if mode=PCMK__VALUE_CIB and
* op=CRM_OP_LRM_DELETE.
*/
if (from_sys) {
lrmd_event_data_t *op = NULL;
const char *from_host = crm_element_value(stored_msg, PCMK__XA_SRC);
const char *transition;
if (strcmp(from_sys, CRM_SYSTEM_TENGINE)) {
transition = crm_element_value(msg_data,
PCMK__XA_TRANSITION_KEY);
} else {
transition = crm_element_value(stored_msg,
PCMK__XA_TRANSITION_KEY);
}
crm_info("Notifying %s on %s that %s was%s deleted",
from_sys, (from_host? from_host : "local node"), rsc_id,
((rc == pcmk_rc_ok)? "" : " not"));
op = lrmd_new_event(rsc_id, PCMK_ACTION_DELETE, 0);
op->type = lrmd_event_exec_complete;
op->user_data = strdup(transition? transition : FAKE_TE_ID);
op->params = pcmk__strkey_table(free, free);
g_hash_table_insert(op->params, strdup(PCMK_XA_CRM_FEATURE_SET),
strdup(CRM_FEATURE_SET));
controld_rc2event(op, rc);
controld_ack_event_directly(from_host, from_sys, NULL, op, rsc_id);
lrmd_free_event(op);
controld_trigger_delete_refresh(from_sys, rsc_id);
}
return I_NULL;
}
}
/*!
* \brief Handle a CRM_OP_REMOTE_STATE message by updating remote peer cache
*
* \param[in] msg Message XML
*
* \return Next FSA input
*/
static enum crmd_fsa_input
handle_remote_state(const xmlNode *msg)
{
const char *conn_host = NULL;
const char *remote_uname = ID(msg);
crm_node_t *remote_peer;
bool remote_is_up = false;
int rc = pcmk_rc_ok;
rc = pcmk__xe_get_bool_attr(msg, PCMK__XA_IN_CCM, &remote_is_up);
CRM_CHECK(remote_uname && rc == pcmk_rc_ok, return I_NULL);
remote_peer = crm_remote_peer_get(remote_uname);
CRM_CHECK(remote_peer, return I_NULL);
pcmk__update_peer_state(__func__, remote_peer,
remote_is_up ? CRM_NODE_MEMBER : CRM_NODE_LOST,
0);
conn_host = crm_element_value(msg, PCMK__XA_CONN_HOST);
if (conn_host) {
pcmk__str_update(&remote_peer->conn_host, conn_host);
} else if (remote_peer->conn_host) {
free(remote_peer->conn_host);
remote_peer->conn_host = NULL;
}
return I_NULL;
}
/*!
* \brief Handle a CRM_OP_PING message
*
* \param[in] msg Message XML
*
* \return Next FSA input
*/
static enum crmd_fsa_input
handle_ping(const xmlNode *msg)
{
const char *value = NULL;
xmlNode *ping = NULL;
xmlNode *reply = NULL;
// Build reply
ping = create_xml_node(NULL, PCMK__XE_PING_RESPONSE);
value = crm_element_value(msg, PCMK__XA_CRM_SYS_TO);
crm_xml_add(ping, PCMK__XA_CRM_SUBSYSTEM, value);
// Add controller state
value = fsa_state2string(controld_globals.fsa_state);
crm_xml_add(ping, PCMK__XA_CRMD_STATE, value);
crm_notice("Current ping state: %s", value); // CTS needs this
// Add controller health
// @TODO maybe do some checks to determine meaningful status
crm_xml_add(ping, PCMK_XA_RESULT, "ok");
// Send reply
reply = create_reply(msg, ping);
free_xml(ping);
if (reply != NULL) {
(void) relay_message(reply, TRUE);
free_xml(reply);
}
// Nothing further to do
return I_NULL;
}
/*!
* \brief Handle a PCMK__CONTROLD_CMD_NODES message
*
* \param[in] request Message XML
*
* \return Next FSA input
*/
static enum crmd_fsa_input
handle_node_list(const xmlNode *request)
{
GHashTableIter iter;
crm_node_t *node = NULL;
xmlNode *reply = NULL;
xmlNode *reply_data = NULL;
// Create message data for reply
reply_data = create_xml_node(NULL, PCMK_XE_NODES);
g_hash_table_iter_init(&iter, crm_peer_cache);
while (g_hash_table_iter_next(&iter, NULL, (gpointer *) & node)) {
xmlNode *xml = create_xml_node(reply_data, PCMK_XE_NODE);
crm_xml_add_ll(xml, PCMK_XA_ID, (long long) node->id); // uint32_t
crm_xml_add(xml, PCMK_XA_UNAME, node->uname);
crm_xml_add(xml, PCMK__XA_IN_CCM, node->state);
}
// Create and send reply
reply = create_reply(request, reply_data);
free_xml(reply_data);
if (reply) {
(void) relay_message(reply, TRUE);
free_xml(reply);
}
// Nothing further to do
return I_NULL;
}
/*!
* \brief Handle a CRM_OP_NODE_INFO request
*
* \param[in] msg Message XML
*
* \return Next FSA input
*/
static enum crmd_fsa_input
handle_node_info_request(const xmlNode *msg)
{
const char *value = NULL;
crm_node_t *node = NULL;
int node_id = 0;
xmlNode *reply = NULL;
xmlNode *reply_data = NULL;
// Build reply
reply_data = create_xml_node(NULL, PCMK_XE_NODE);
crm_xml_add(reply_data, PCMK__XA_CRM_SUBSYSTEM, CRM_SYSTEM_CRMD);
// Add whether current partition has quorum
pcmk__xe_set_bool_attr(reply_data, PCMK_XA_HAVE_QUORUM,
pcmk_is_set(controld_globals.flags,
controld_has_quorum));
// Check whether client requested node info by ID and/or name
crm_element_value_int(msg, PCMK_XA_ID, &node_id);
if (node_id < 0) {
node_id = 0;
}
value = crm_element_value(msg, PCMK_XA_UNAME);
// Default to local node if none given
if ((node_id == 0) && (value == NULL)) {
value = controld_globals.our_nodename;
}
node = pcmk__search_node_caches(node_id, value, pcmk__node_search_any);
if (node) {
crm_xml_add(reply_data, PCMK_XA_ID, node->uuid);
crm_xml_add(reply_data, PCMK_XA_UNAME, node->uname);
- crm_xml_add(reply_data, PCMK__XA_CRMD, node->state);
+ crm_xml_add(reply_data, PCMK_XA_CRMD, node->state);
pcmk__xe_set_bool_attr(reply_data, PCMK__XA_REMOTE_NODE,
pcmk_is_set(node->flags, crm_remote_node));
}
// Send reply
reply = create_reply(msg, reply_data);
free_xml(reply_data);
if (reply != NULL) {
(void) relay_message(reply, TRUE);
free_xml(reply);
}
// Nothing further to do
return I_NULL;
}
static void
verify_feature_set(xmlNode *msg)
{
const char *dc_version = crm_element_value(msg, PCMK_XA_CRM_FEATURE_SET);
if (dc_version == NULL) {
/* All we really know is that the DC feature set is older than 3.1.0,
* but that's also all that really matters.
*/
dc_version = "3.0.14";
}
if (feature_set_compatible(dc_version, CRM_FEATURE_SET)) {
crm_trace("Local feature set (%s) is compatible with DC's (%s)",
CRM_FEATURE_SET, dc_version);
} else {
crm_err("Local feature set (%s) is incompatible with DC's (%s)",
CRM_FEATURE_SET, dc_version);
// Nothing is likely to improve without administrator involvement
controld_set_fsa_input_flags(R_STAYDOWN);
crmd_exit(CRM_EX_FATAL);
}
}
// DC gets own shutdown all-clear
static enum crmd_fsa_input
handle_shutdown_self_ack(xmlNode *stored_msg)
{
const char *host_from = crm_element_value(stored_msg, PCMK__XA_SRC);
if (pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN)) {
// The expected case -- we initiated own shutdown sequence
crm_info("Shutting down controller");
return I_STOP;
}
if (pcmk__str_eq(host_from, controld_globals.dc_name, pcmk__str_casei)) {
// Must be logic error -- DC confirming its own unrequested shutdown
crm_err("Shutting down controller immediately due to "
"unexpected shutdown confirmation");
return I_TERMINATE;
}
if (controld_globals.fsa_state != S_STOPPING) {
// Shouldn't happen -- non-DC confirming unrequested shutdown
crm_err("Starting new DC election because %s is "
"confirming shutdown we did not request",
(host_from? host_from : "another node"));
return I_ELECTION;
}
// Shouldn't happen, but we are already stopping anyway
crm_debug("Ignoring unexpected shutdown confirmation from %s",
(host_from? host_from : "another node"));
return I_NULL;
}
// Non-DC gets shutdown all-clear from DC
static enum crmd_fsa_input
handle_shutdown_ack(xmlNode *stored_msg)
{
const char *host_from = crm_element_value(stored_msg, PCMK__XA_SRC);
if (host_from == NULL) {
crm_warn("Ignoring shutdown request without origin specified");
return I_NULL;
}
if (pcmk__str_eq(host_from, controld_globals.dc_name,
pcmk__str_null_matches|pcmk__str_casei)) {
if (pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN)) {
crm_info("Shutting down controller after confirmation from %s",
host_from);
} else {
crm_err("Shutting down controller after unexpected "
"shutdown request from %s", host_from);
controld_set_fsa_input_flags(R_STAYDOWN);
}
return I_STOP;
}
crm_warn("Ignoring shutdown request from %s because DC is %s",
host_from, controld_globals.dc_name);
return I_NULL;
}
static enum crmd_fsa_input
handle_request(xmlNode *stored_msg, enum crmd_fsa_cause cause)
{
xmlNode *msg = NULL;
const char *op = crm_element_value(stored_msg, PCMK__XA_CRM_TASK);
/* Optimize this for the DC - it has the most to do */
crm_log_xml_trace(stored_msg, "request");
if (op == NULL) {
crm_warn("Ignoring request without " PCMK__XA_CRM_TASK);
return I_NULL;
}
if (strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0) {
const char *from = crm_element_value(stored_msg, PCMK__XA_SRC);
crm_node_t *node = pcmk__search_node_caches(0, from,
pcmk__node_search_cluster);
pcmk__update_peer_expected(__func__, node, CRMD_JOINSTATE_DOWN);
if(AM_I_DC == FALSE) {
return I_NULL; /* Done */
}
}
/*========== DC-Only Actions ==========*/
if (AM_I_DC) {
if (strcmp(op, CRM_OP_JOIN_ANNOUNCE) == 0) {
return I_NODE_JOIN;
} else if (strcmp(op, CRM_OP_JOIN_REQUEST) == 0) {
return I_JOIN_REQUEST;
} else if (strcmp(op, CRM_OP_JOIN_CONFIRM) == 0) {
return I_JOIN_RESULT;
} else if (strcmp(op, CRM_OP_SHUTDOWN) == 0) {
return handle_shutdown_self_ack(stored_msg);
} else if (strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0) {
// Another controller wants to shut down its node
return handle_shutdown_request(stored_msg);
}
}
/*========== common actions ==========*/
if (strcmp(op, CRM_OP_NOVOTE) == 0) {
ha_msg_input_t fsa_input;
fsa_input.msg = stored_msg;
register_fsa_input_adv(C_HA_MESSAGE, I_NULL, &fsa_input,
A_ELECTION_COUNT | A_ELECTION_CHECK, FALSE,
__func__);
} else if (strcmp(op, CRM_OP_REMOTE_STATE) == 0) {
/* a remote connection host is letting us know the node state */
return handle_remote_state(stored_msg);
} else if (strcmp(op, CRM_OP_THROTTLE) == 0) {
throttle_update(stored_msg);
if (AM_I_DC && (controld_globals.transition_graph != NULL)
&& !controld_globals.transition_graph->complete) {
crm_debug("The throttle changed. Trigger a graph.");
trigger_graph();
}
return I_NULL;
} else if (strcmp(op, CRM_OP_CLEAR_FAILCOUNT) == 0) {
return handle_failcount_op(stored_msg);
} else if (strcmp(op, CRM_OP_VOTE) == 0) {
/* count the vote and decide what to do after that */
ha_msg_input_t fsa_input;
fsa_input.msg = stored_msg;
register_fsa_input_adv(C_HA_MESSAGE, I_NULL, &fsa_input,
A_ELECTION_COUNT | A_ELECTION_CHECK, FALSE,
__func__);
/* Sometimes we _must_ go into S_ELECTION */
if (controld_globals.fsa_state == S_HALT) {
crm_debug("Forcing an election from S_HALT");
return I_ELECTION;
}
} else if (strcmp(op, CRM_OP_JOIN_OFFER) == 0) {
verify_feature_set(stored_msg);
crm_debug("Raising I_JOIN_OFFER: join-%s",
crm_element_value(stored_msg, PCMK__XA_JOIN_ID));
return I_JOIN_OFFER;
} else if (strcmp(op, CRM_OP_JOIN_ACKNAK) == 0) {
crm_debug("Raising I_JOIN_RESULT: join-%s",
crm_element_value(stored_msg, PCMK__XA_JOIN_ID));
return I_JOIN_RESULT;
} else if (strcmp(op, CRM_OP_LRM_DELETE) == 0) {
return handle_lrm_delete(stored_msg);
} else if ((strcmp(op, CRM_OP_LRM_FAIL) == 0)
|| (strcmp(op, CRM_OP_LRM_REFRESH) == 0) // @COMPAT
|| (strcmp(op, CRM_OP_REPROBE) == 0)) {
crm_xml_add(stored_msg, PCMK__XA_CRM_SYS_TO, CRM_SYSTEM_LRMD);
return I_ROUTER;
} else if (strcmp(op, CRM_OP_NOOP) == 0) {
return I_NULL;
} else if (strcmp(op, CRM_OP_PING) == 0) {
return handle_ping(stored_msg);
} else if (strcmp(op, CRM_OP_NODE_INFO) == 0) {
return handle_node_info_request(stored_msg);
} else if (strcmp(op, CRM_OP_RM_NODE_CACHE) == 0) {
int id = 0;
const char *name = NULL;
crm_element_value_int(stored_msg, PCMK_XA_ID, &id);
name = crm_element_value(stored_msg, PCMK_XA_UNAME);
if(cause == C_IPC_MESSAGE) {
msg = create_request(CRM_OP_RM_NODE_CACHE, NULL, NULL, CRM_SYSTEM_CRMD, CRM_SYSTEM_CRMD, NULL);
if (send_cluster_message(NULL, crm_msg_crmd, msg, TRUE) == FALSE) {
crm_err("Could not instruct peers to remove references to node %s/%u", name, id);
} else {
crm_notice("Instructing peers to remove references to node %s/%u", name, id);
}
free_xml(msg);
} else {
reap_crm_member(id, name);
/* If we're forgetting this node, also forget any failures to fence
* it, so we don't carry that over to any node added later with the
* same name.
*/
st_fail_count_reset(name);
}
} else if (strcmp(op, CRM_OP_MAINTENANCE_NODES) == 0) {
xmlNode *xml = get_message_xml(stored_msg, PCMK__XE_CRM_XML);
remote_ra_process_maintenance_nodes(xml);
} else if (strcmp(op, PCMK__CONTROLD_CMD_NODES) == 0) {
return handle_node_list(stored_msg);
/*========== (NOT_DC)-Only Actions ==========*/
} else if (!AM_I_DC) {
if (strcmp(op, CRM_OP_SHUTDOWN) == 0) {
return handle_shutdown_ack(stored_msg);
}
} else {
crm_err("Unexpected request (%s) sent to %s", op, AM_I_DC ? "the DC" : "non-DC node");
crm_log_xml_err(stored_msg, "Unexpected");
}
return I_NULL;
}
static void
handle_response(xmlNode *stored_msg)
{
const char *op = crm_element_value(stored_msg, PCMK__XA_CRM_TASK);
crm_log_xml_trace(stored_msg, "reply");
if (op == NULL) {
crm_warn("Ignoring reply without " PCMK__XA_CRM_TASK);
} else if (AM_I_DC && strcmp(op, CRM_OP_PECALC) == 0) {
// Check whether scheduler answer been superseded by subsequent request
const char *msg_ref = crm_element_value(stored_msg, PCMK_XA_REFERENCE);
if (msg_ref == NULL) {
crm_err("%s - Ignoring calculation with no reference", op);
} else if (pcmk__str_eq(msg_ref, controld_globals.fsa_pe_ref,
pcmk__str_none)) {
ha_msg_input_t fsa_input;
controld_stop_sched_timer();
fsa_input.msg = stored_msg;
register_fsa_input_later(C_IPC_MESSAGE, I_PE_SUCCESS, &fsa_input);
} else {
crm_info("%s calculation %s is obsolete", op, msg_ref);
}
} else if (strcmp(op, CRM_OP_VOTE) == 0
|| strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0 || strcmp(op, CRM_OP_SHUTDOWN) == 0) {
} else {
const char *host_from = crm_element_value(stored_msg, PCMK__XA_SRC);
crm_err("Unexpected response (op=%s, src=%s) sent to the %s",
op, host_from, AM_I_DC ? "DC" : "controller");
}
}
static enum crmd_fsa_input
handle_shutdown_request(xmlNode * stored_msg)
{
/* handle here to avoid potential version issues
* where the shutdown message/procedure may have
* been changed in later versions.
*
* This way the DC is always in control of the shutdown
*/
char *now_s = NULL;
const char *host_from = crm_element_value(stored_msg, PCMK__XA_SRC);
if (host_from == NULL) {
/* we're shutting down and the DC */
host_from = controld_globals.our_nodename;
}
crm_info("Creating shutdown request for %s (state=%s)", host_from,
fsa_state2string(controld_globals.fsa_state));
crm_log_xml_trace(stored_msg, "message");
now_s = pcmk__ttoa(time(NULL));
update_attrd(host_from, PCMK__NODE_ATTR_SHUTDOWN, now_s, NULL, FALSE);
free(now_s);
/* will be picked up by the TE as long as its running */
return I_NULL;
}
static void
send_msg_via_ipc(xmlNode * msg, const char *sys)
{
pcmk__client_t *client_channel = NULL;
CRM_CHECK(sys != NULL, return);
client_channel = pcmk__find_client_by_id(sys);
if (crm_element_value(msg, PCMK__XA_SRC) == NULL) {
crm_xml_add(msg, PCMK__XA_SRC, controld_globals.our_nodename);
}
if (client_channel != NULL) {
/* Transient clients such as crmadmin */
pcmk__ipc_send_xml(client_channel, 0, msg, crm_ipc_server_event);
} else if (pcmk__str_eq(sys, CRM_SYSTEM_TENGINE, pcmk__str_none)) {
xmlNode *data = get_message_xml(msg, PCMK__XE_CRM_XML);
process_te_message(msg, data);
} else if (pcmk__str_eq(sys, CRM_SYSTEM_LRMD, pcmk__str_none)) {
fsa_data_t fsa_data;
ha_msg_input_t fsa_input;
fsa_input.msg = msg;
fsa_input.xml = get_message_xml(msg, PCMK__XE_CRM_XML);
fsa_data.id = 0;
fsa_data.actions = 0;
fsa_data.data = &fsa_input;
fsa_data.fsa_input = I_MESSAGE;
fsa_data.fsa_cause = C_IPC_MESSAGE;
fsa_data.origin = __func__;
fsa_data.data_type = fsa_dt_ha_msg;
do_lrm_invoke(A_LRM_INVOKE, C_IPC_MESSAGE, controld_globals.fsa_state,
I_MESSAGE, &fsa_data);
} else if (crmd_is_proxy_session(sys)) {
crmd_proxy_send(sys, msg);
} else {
crm_info("Received invalid request: unknown subsystem '%s'", sys);
}
}
void
delete_ha_msg_input(ha_msg_input_t * orig)
{
if (orig == NULL) {
return;
}
free_xml(orig->msg);
free(orig);
}
/*!
* \internal
* \brief Notify the cluster of a remote node state change
*
* \param[in] node_name Node's name
* \param[in] node_up true if node is up, false if down
*/
void
broadcast_remote_state_message(const char *node_name, bool node_up)
{
xmlNode *msg = create_request(CRM_OP_REMOTE_STATE, NULL, NULL,
CRM_SYSTEM_CRMD, CRM_SYSTEM_CRMD, NULL);
crm_info("Notifying cluster of Pacemaker Remote node %s %s",
node_name, node_up? "coming up" : "going down");
crm_xml_add(msg, PCMK_XA_ID, node_name);
pcmk__xe_set_bool_attr(msg, PCMK__XA_IN_CCM, node_up);
if (node_up) {
crm_xml_add(msg, PCMK__XA_CONN_HOST, controld_globals.our_nodename);
}
send_cluster_message(NULL, crm_msg_crmd, msg, TRUE);
free_xml(msg);
}
diff --git a/include/crm/msg_xml.h b/include/crm/msg_xml.h
index ee0f3ea169..14620c5e48 100644
--- a/include/crm/msg_xml.h
+++ b/include/crm/msg_xml.h
@@ -1,410 +1,411 @@
/*
* Copyright 2004-2024 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__CRM_MSG_XML__H
# define PCMK__CRM_MSG_XML__H
# include <crm/common/xml.h>
#if !defined(PCMK_ALLOW_DEPRECATED) || (PCMK_ALLOW_DEPRECATED == 1)
#include <crm/msg_xml_compat.h>
#endif
#ifdef __cplusplus
extern "C" {
#endif
/* This file defines constants for various XML syntax (mainly element and
* attribute names).
*
* For consistency, new constants should start with "PCMK_", followed by:
* * "XE" for XML element names
* * "XA" for XML attribute names
* * "OPT" for cluster option (property) names
* * "META" for meta-attribute names
* * "VALUE" for enumerated values for various options
*
* Old names that don't follow this policy should eventually be deprecated and
* replaced with names that do.
*
* Symbols should be public if the user may specify them somewhere (especially
* the CIB) or if they're part of a well-defined structure that a user may need
* to parse. They should be internal if they're used only internally to
* Pacemaker (such as daemon IPC/CPG message XML).
*
* Constants belong in the following locations:
* * Public "XE" and "XA": msg_xml.h
* * Internal "XE" and "XA": crm_internal.h
* * Public "OPT", "META", and "VALUE": options.h
* * Internal "OPT", "META", and "VALUE": options_internal.h
*
* For meta-attributes that can be specified as either XML attributes or nvpair
* names, use "META" unless using both "XA" and "META" constants adds clarity.
* An example is operation attributes, which can be specified either as
* attributes of the PCMK_XE_OP element or as nvpairs in a meta-attribute set
* beneath the PCMK_XE_OP element.
*/
/*
* XML elements
*/
#define PCMK_XE_ACL_GROUP "acl_group"
#define PCMK_XE_ACL_PERMISSION "acl_permission"
#define PCMK_XE_ACL_ROLE "acl_role"
#define PCMK_XE_ACL_TARGET "acl_target"
#define PCMK_XE_ACLS "acls"
#define PCMK_XE_ACTION "action"
#define PCMK_XE_ACTIONS "actions"
#define PCMK_XE_AGENT "agent"
#define PCMK_XE_AGENT_STATUS "agent-status"
#define PCMK_XE_AGENTS "agents"
#define PCMK_XE_ALERT "alert"
#define PCMK_XE_ALERTS "alerts"
#define PCMK_XE_ALLOCATIONS "allocations"
#define PCMK_XE_ALLOCATIONS_UTILIZATIONS "allocations_utilizations"
#define PCMK_XE_ATTRIBUTE "attribute"
#define PCMK_XE_BAN "ban"
#define PCMK_XE_BANS "bans"
#define PCMK_XE_BUNDLE "bundle"
#define PCMK_XE_CAPACITY "capacity"
#define PCMK_XE_CHANGE "change"
#define PCMK_XE_CHANGE_ATTR "change-attr"
#define PCMK_XE_CHANGE_LIST "change-list"
#define PCMK_XE_CHANGE_RESULT "change-result"
#define PCMK_XE_CHECK "check"
#define PCMK_XE_CIB "cib"
#define PCMK_XE_CLONE "clone"
#define PCMK_XE_CLUSTER_ACTION "cluster_action"
#define PCMK_XE_CLUSTER_INFO "cluster-info"
#define PCMK_XE_CLUSTER_OPTIONS "cluster_options"
#define PCMK_XE_CLUSTER_PROPERTY_SET "cluster_property_set"
#define PCMK_XE_CLUSTER_STATUS "cluster_status"
#define PCMK_XE_COMMAND "command"
#define PCMK_XE_CONFIGURATION "configuration"
#define PCMK_XE_CONSTRAINTS "constraints"
#define PCMK_XE_CONTENT "content"
#define PCMK_XE_CRM_CONFIG "crm_config"
#define PCMK_XE_CRM_MON "crm_mon"
#define PCMK_XE_CRM_MON_DISCONNECTED "crm-mon-disconnected"
#define PCMK_XE_CURRENT_DC "current_dc"
#define PCMK_XE_DATE_EXPRESSION "date_expression"
#define PCMK_XE_DATE_SPEC "date_spec"
#define PCMK_XE_DC "dc"
#define PCMK_XE_DIFF "diff"
#define PCMK_XE_DIGEST "digest"
#define PCMK_XE_DIGESTS "digests"
#define PCMK_XE_DOCKER "docker"
#define PCMK_XE_DURATION "duration"
#define PCMK_XE_ERROR "error"
#define PCMK_XE_ERRORS "errors"
#define PCMK_XE_EXPRESSION "expression"
#define PCMK_XE_FAILURE "failure"
#define PCMK_XE_FAILURES "failures"
#define PCMK_XE_FEATURE "feature"
#define PCMK_XE_FEATURES "features"
#define PCMK_XE_FENCE_EVENT "fence_event"
#define PCMK_XE_FENCE_HISTORY "fence_history"
#define PCMK_XE_FENCING_ACTION "fencing_action"
#define PCMK_XE_FENCING_LEVEL "fencing-level"
#define PCMK_XE_FENCING_TOPOLOGY "fencing-topology"
#define PCMK_XE_GROUP "group"
#define PCMK_XE_INJECT_ATTR "inject_attr"
#define PCMK_XE_INJECT_SPEC "inject_spec"
#define PCMK_XE_INSTANCE_ATTRIBUTES "instance_attributes"
#define PCMK_XE_INSTRUCTION "instruction"
#define PCMK_XE_ITEM "item"
#define PCMK_XE_LAST_CHANGE "last_change"
#define PCMK_XE_LAST_FENCED "last-fenced"
#define PCMK_XE_LAST_UPDATE "last_update"
#define PCMK_XE_LIST "list"
#define PCMK_XE_LONGDESC "longdesc"
#define PCMK_XE_META_ATTRIBUTES "meta_attributes"
#define PCMK_XE_METADATA "metadata"
#define PCMK_XE_MODIFICATIONS "modifications"
#define PCMK_XE_MODIFY_NODE "modify_node"
#define PCMK_XE_MODIFY_TICKET "modify_ticket"
#define PCMK_XE_NETWORK "network"
#define PCMK_XE_NODE "node"
#define PCMK_XE_NODE_ACTION "node_action"
#define PCMK_XE_NODE_ATTRIBUTES "node_attributes"
#define PCMK_XE_NODE_HISTORY "node_history"
#define PCMK_XE_NODE_INFO "node-info"
#define PCMK_XE_NODES "nodes"
#define PCMK_XE_NVPAIR "nvpair"
#define PCMK_XE_OBJ_REF "obj_ref"
#define PCMK_XE_OP "op"
#define PCMK_XE_OP_DEFAULTS "op_defaults"
#define PCMK_XE_OP_EXPRESSION "op_expression"
#define PCMK_XE_OPERATION "operation"
#define PCMK_XE_OPERATIONS "operations"
#define PCMK_XE_OPTION "option"
#define PCMK_XE_OUTPUT "output"
#define PCMK_XE_OVERRIDE "override"
#define PCMK_XE_OVERRIDES "overrides"
#define PCMK_XE_PACEMAKER_RESULT "pacemaker-result"
#define PCMK_XE_PACEMAKERD "pacemakerd"
#define PCMK_XE_PARAMETER "parameter"
#define PCMK_XE_PARAMETERS "parameters"
#define PCMK_XE_PORT_MAPPING "port-mapping"
#define PCMK_XE_POSITION "position"
#define PCMK_XE_PRIMITIVE "primitive"
#define PCMK_XE_RECIPIENT "recipient"
#define PCMK_XE_REPLICA "replica"
#define PCMK_XE_RESOURCE "resource"
#define PCMK_XE_RESOURCE_AGENT "resource-agent"
#define PCMK_XE_RESOURCE_AGENT_ACTION "resource-agent-action"
#define PCMK_XE_RESOURCE_CONFIG "resource_config"
#define PCMK_XE_RESOURCE_REF "resource_ref"
#define PCMK_XE_RESOURCE_SET "resource_set"
#define PCMK_XE_RESOURCES "resources"
#define PCMK_XE_RESULT_CODE "result-code"
#define PCMK_XE_REVISED_CLUSTER_STATUS "revised_cluster_status"
#define PCMK_XE_ROLE "role"
#define PCMK_XE_RSC_ACTION "rsc_action"
#define PCMK_XE_RSC_COLOCATION "rsc_colocation"
#define PCMK_XE_RSC_DEFAULTS "rsc_defaults"
#define PCMK_XE_RSC_EXPRESSION "rsc_expression"
#define PCMK_XE_RSC_LOCATION "rsc_location"
#define PCMK_XE_RSC_ORDER "rsc_order"
#define PCMK_XE_RSC_TICKET "rsc_ticket"
#define PCMK_XE_RULE "rule"
#define PCMK_XE_SELECT "select"
#define PCMK_XE_SELECT_ATTRIBUTES "select_attributes"
#define PCMK_XE_SELECT_FENCING "select_fencing"
#define PCMK_XE_SELECT_NODES "select_nodes"
#define PCMK_XE_SELECT_RESOURCES "select_resources"
#define PCMK_XE_SHORTDESC "shortdesc"
#define PCMK_XE_SOURCE "source"
#define PCMK_XE_STATUS "status"
#define PCMK_XE_STORAGE "storage"
#define PCMK_XE_STORAGE_MAPPING "storage-mapping"
#define PCMK_XE_SUMMARY "summary"
#define PCMK_XE_TAG "tag"
#define PCMK_XE_TAGS "tags"
#define PCMK_XE_TARGET "target"
#define PCMK_XE_TEMPLATE "template"
#define PCMK_XE_TICKET "ticket"
#define PCMK_XE_TICKETS "tickets"
#define PCMK_XE_TRANSITION "transition"
#define PCMK_XE_UTILIZATION "utilization"
#define PCMK_XE_UTILIZATIONS "utilizations"
#define PCMK_XE_VERSION "version"
/*
* XML attributes
*/
#define PCMK_XA_ACTION "action"
#define PCMK_XA_ACTIVE "active"
#define PCMK_XA_ADD_HOST "add-host"
#define PCMK_XA_ADMIN_EPOCH "admin_epoch"
#define PCMK_XA_AGENT "agent"
#define PCMK_XA_API_VERSION "api-version"
#define PCMK_XA_ATTRIBUTE "attribute"
#define PCMK_XA_AUTHOR "author"
#define PCMK_XA_BLOCKED "blocked"
#define PCMK_XA_BOOLEAN_OP "boolean-op"
#define PCMK_XA_BUILD "build"
#define PCMK_XA_CACHED "cached"
#define PCMK_XA_CALL "call"
#define PCMK_XA_CIB_LAST_WRITTEN "cib-last-written"
#define PCMK_XA_CIB_NODE "cib_node"
#define PCMK_XA_CLASS "class"
#define PCMK_XA_CLIENT "client"
#define PCMK_XA_CODE "code"
#define PCMK_XA_COMMENT "comment"
#define PCMK_XA_COMPLETED "completed"
#define PCMK_XA_CONTROL_PORT "control-port"
#define PCMK_XA_COUNT "count"
#define PCMK_XA_CRM_DEBUG_ORIGIN "crm-debug-origin"
#define PCMK_XA_CRM_FEATURE_SET "crm_feature_set"
#define PCMK_XA_CRM_TIMESTAMP "crm-timestamp"
+#define PCMK_XA_CRMD "crmd"
#define PCMK_XA_DAYS "days"
#define PCMK_XA_DC_UUID "dc-uuid"
#define PCMK_XA_DEFAULT "default"
#define PCMK_XA_DELEGATE "delegate"
#define PCMK_XA_DESCRIPTION "description"
#define PCMK_XA_DEST "dest"
#define PCMK_XA_DEVICES "devices"
#define PCMK_XA_DISABLED "disabled"
#define PCMK_XA_DURATION "duration"
#define PCMK_XA_END "end"
#define PCMK_XA_EPOCH "epoch"
#define PCMK_XA_EXEC "exec"
#define PCMK_XA_EXEC_TIME "exec-time"
#define PCMK_XA_EXECUTION_CODE "execution_code"
#define PCMK_XA_EXECUTION_DATE "execution-date"
#define PCMK_XA_EXECUTION_MESSAGE "execution_message"
#define PCMK_XA_EXIT_REASON "exit-reason"
#define PCMK_XA_EXITCODE "exitcode"
#define PCMK_XA_EXITREASON "exitreason"
#define PCMK_XA_EXITSTATUS "exitstatus"
#define PCMK_XA_EXPECTED "expected"
#define PCMK_XA_EXPECTED_UP "expected_up"
#define PCMK_XA_EXTENDED_STATUS "extended-status"
#define PCMK_XA_FAIL_COUNT "fail-count"
#define PCMK_XA_FAILED "failed"
#define PCMK_XA_FAILURE_IGNORED "failure_ignored"
#define PCMK_XA_FEATURE_SET "feature_set"
#define PCMK_XA_FEATURES "features"
#define PCMK_XA_FILE "file"
#define PCMK_XA_FIRST "first"
#define PCMK_XA_FIRST_ACTION "first-action"
#define PCMK_XA_FOR "for"
#define PCMK_XA_FORMAT "format"
#define PCMK_XA_FUNCTION "function"
#define PCMK_XA_HASH "hash"
#define PCMK_XA_HAVE_QUORUM "have-quorum"
#define PCMK_XA_HEALTH "health"
#define PCMK_XA_HOST "host"
#define PCMK_XA_HOST_INTERFACE "host-interface"
#define PCMK_XA_HOST_NETMASK "host-netmask"
#define PCMK_XA_HOURS "hours"
#define PCMK_XA_ID "id"
#define PCMK_XA_ID_AS_RESOURCE "id_as_resource"
#define PCMK_XA_ID_REF "id-ref"
#define PCMK_XA_IMAGE "image"
#define PCMK_XA_INDEX "index"
#define PCMK_XA_INFLUENCE "influence"
#define PCMK_XA_INSTANCE "instance"
#define PCMK_XA_INTERNAL_PORT "internal-port"
#define PCMK_XA_INTERVAL "interval"
#define PCMK_XA_IP_RANGE_START "ip-range-start"
#define PCMK_XA_IS_DC "is_dc"
#define PCMK_XA_KIND "kind"
#define PCMK_XA_LANG "lang"
#define PCMK_XA_LAST_FAILURE "last-failure"
#define PCMK_XA_LAST_GRANTED "last-granted"
#define PCMK_XA_LAST_RC_CHANGE "last-rc-change"
#define PCMK_XA_LAST_UPDATED "last_updated"
#define PCMK_XA_LOCKED_TO "locked_to"
#define PCMK_XA_LOCKED_TO_HYPHEN "locked-to"
#define PCMK_XA_LOSS_POLICY "loss-policy"
#define PCMK_XA_MAINTENANCE "maintenance"
#define PCMK_XA_MAINTENANCE_MODE "maintenance-mode"
#define PCMK_XA_MANAGED "managed"
#define PCMK_XA_MESSAGE "message"
#define PCMK_XA_MINUTES "minutes"
#define PCMK_XA_MIXED_VERSION "mixed_version"
#define PCMK_XA_MONTHDAYS "monthdays"
#define PCMK_XA_MONTHS "months"
#define PCMK_XA_MULTI_STATE "multi_state"
#define PCMK_XA_NAME "name"
#define PCMK_XA_NETWORK "network"
#define PCMK_XA_NEXT_ROLE "next-role"
#define PCMK_XA_NO_QUORUM_PANIC "no-quorum-panic"
#define PCMK_XA_NO_QUORUM_POLICY "no-quorum-policy"
#define PCMK_XA_NODE "node"
#define PCMK_XA_NODE_ATTRIBUTE "node-attribute"
#define PCMK_XA_NODEID "nodeid"
#define PCMK_XA_NODES_RUNNING_ON "nodes_running_on"
#define PCMK_XA_NUM_UPDATES "num_updates"
#define PCMK_XA_NUMBER "number"
#define PCMK_XA_NUMBER_RESOURCES "number_resources"
#define PCMK_XA_OBJECT_TYPE "object-type"
#define PCMK_XA_ONLINE "online"
#define PCMK_XA_OP "op"
#define PCMK_XA_OPERATION "operation"
#define PCMK_XA_OPTIONS "options"
#define PCMK_XA_ORIGIN "origin"
#define PCMK_XA_ORPHANED "orphaned"
#define PCMK_XA_PATH "path"
#define PCMK_XA_PENDING "pending"
#define PCMK_XA_PORT "port"
#define PCMK_XA_PRESENT "present"
#define PCMK_XA_PRIORITY_FENCING_DELAY_MS "priority-fencing-delay-ms"
#define PCMK_XA_PROGRAM "program"
#define PCMK_XA_PROMOTED_MAX "promoted-max"
#define PCMK_XA_PROMOTED_ONLY "promoted-only"
#define PCMK_XA_PROVIDER "provider"
#define PCMK_XA_QUEUE_TIME "queue-time"
#define PCMK_XA_RANGE "range"
#define PCMK_XA_REASON "reason"
#define PCMK_XA_REFERENCE "reference"
#define PCMK_XA_RELOADABLE "reloadable"
#define PCMK_XA_REMOTE_CLEAR_PORT "remote-clear-port"
#define PCMK_XA_REMOTE_TLS_PORT "remote-tls-port"
#define PCMK_XA_REPLICAS "replicas"
#define PCMK_XA_REPLICAS_PER_HOST "replicas-per-host"
#define PCMK_XA_REQUEST "request"
#define PCMK_XA_REQUIRE_ALL "require-all"
#define PCMK_XA_RESOURCE "resource"
#define PCMK_XA_RESOURCE_AGENT "resource_agent"
#define PCMK_XA_RESOURCE_DISCOVERY "resource-discovery"
#define PCMK_XA_RESOURCES_RUNNING "resources_running"
#define PCMK_XA_RESULT "result"
#define PCMK_XA_ROLE "role"
#define PCMK_XA_RSC "rsc"
#define PCMK_XA_RSC_PATTERN "rsc-pattern"
#define PCMK_XA_RSC_ROLE "rsc-role"
#define PCMK_XA_RUN_COMMAND "run-command"
#define PCMK_XA_RUNNING "running"
#define PCMK_XA_SCOPE "scope"
#define PCMK_XA_SCORE "score"
#define PCMK_XA_SCORE_ATTRIBUTE "score-attribute"
#define PCMK_XA_SEQUENTIAL "sequential"
#define PCMK_XA_SECONDS "seconds"
#define PCMK_XA_SHUTDOWN "shutdown"
#define PCMK_XA_SOURCE "source"
#define PCMK_XA_SOURCE_DIR "source-dir"
#define PCMK_XA_SOURCE_DIR_ROOT "source-dir-root"
#define PCMK_XA_STANDBY "standby"
#define PCMK_XA_STANDBY_ONFAIL "standby_onfail"
#define PCMK_XA_START "start"
#define PCMK_XA_STATUS "status"
#define PCMK_XA_STONITH_ENABLED "stonith-enabled"
#define PCMK_XA_STONITH_TIMEOUT_MS "stonith-timeout-ms"
#define PCMK_XA_STOP_ALL_RESOURCES "stop-all-resources"
#define PCMK_XA_SYMMETRIC_CLUSTER "symmetric-cluster"
#define PCMK_XA_SYMMETRICAL "symmetrical"
#define PCMK_XA_TARGET "target"
#define PCMK_XA_TARGET_ATTRIBUTE "target-attribute"
#define PCMK_XA_TARGET_DIR "target-dir"
#define PCMK_XA_TARGET_PATTERN "target-pattern"
#define PCMK_XA_TARGET_ROLE "target_role"
#define PCMK_XA_TARGET_VALUE "target-value"
#define PCMK_XA_TEMPLATE "template"
#define PCMK_XA_TICKET "ticket"
#define PCMK_XA_TIME "time"
#define PCMK_XA_THEN "then"
#define PCMK_XA_THEN_ACTION "then-action"
#define PCMK_XA_TYPE "type"
#define PCMK_XA_UNAME "uname"
#define PCMK_XA_UNCLEAN "unclean"
#define PCMK_XA_UNIQUE "unique"
#define PCMK_XA_UPDATE_CLIENT "update-client"
#define PCMK_XA_UPDATE_ORIGIN "update-origin"
#define PCMK_XA_UPDATE_USER "update-user"
#define PCMK_XA_USER "user"
#define PCMK_XA_VALIDATE_WITH "validate-with"
#define PCMK_XA_VALUE "value"
#define PCMK_XA_VALUE_SOURCE "value-source"
#define PCMK_XA_VERSION "version"
#define PCMK_XA_WEEKDAYS "weekdays"
#define PCMK_XA_WEEKS "weeks"
#define PCMK_XA_WEEKYEARS "weekyears"
#define PCMK_XA_WEIGHT "weight"
#define PCMK_XA_WHEN "when"
#define PCMK_XA_WITH_QUORUM "with_quorum"
#define PCMK_XA_WITH_RSC "with-rsc"
#define PCMK_XA_WITH_RSC_ROLE "with-rsc-role"
#define PCMK_XA_XPATH "xpath"
#define PCMK_XA_YEARDAYS "yeardays"
#define PCMK_XA_YEARS "years"
# define ID(x) crm_element_value(x, PCMK_XA_ID)
#ifdef __cplusplus
}
#endif
#endif
diff --git a/include/crm_internal.h b/include/crm_internal.h
index ea8e440713..20a9b983c2 100644
--- a/include/crm_internal.h
+++ b/include/crm_internal.h
@@ -1,268 +1,267 @@
/*
* Copyright 2006-2024 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 CRM_INTERNAL__H
# define CRM_INTERNAL__H
# ifndef PCMK__CONFIG_H
# define PCMK__CONFIG_H
# include <config.h>
# endif
# include <portability.h>
/* Our minimum glib dependency is 2.42. Define that as both the minimum and
* maximum glib APIs that are allowed (i.e. APIs that were already deprecated
* in 2.42, and APIs introduced after 2.42, cannot be used by Pacemaker code).
*/
#define GLIB_VERSION_MIN_REQUIRED GLIB_VERSION_2_42
#define GLIB_VERSION_MAX_ALLOWED GLIB_VERSION_2_42
# include <glib.h>
# include <stdbool.h>
# include <libxml/tree.h>
/* Public API headers can guard including deprecated API headers with this
* symbol, thus preventing internal code (which includes this header) from using
* deprecated APIs, while still allowing external code to use them by default.
*/
#define PCMK_ALLOW_DEPRECATED 0
# include <crm/lrmd.h>
# include <crm/common/logging.h>
# include <crm/common/logging_internal.h>
# include <crm/common/ipc_internal.h>
# include <crm/common/options_internal.h>
# include <crm/common/output_internal.h>
# include <crm/common/scheduler_internal.h>
# include <crm/common/xml_internal.h>
# include <crm/common/xml_names_internal.h>
# include <crm/common/internal.h>
# include <locale.h>
# include <gettext.h>
#define N_(String) (String)
#ifdef ENABLE_NLS
# define _(String) gettext(String)
#else
# define _(String) (String)
#endif
/*
* XML element names used only by internal code
*/
#define PCMK__XE_ATTRIBUTES "attributes"
#define PCMK__XE_CRM_EVENT "crm_event"
#define PCMK__XE_CRM_XML "crm_xml"
#define PCMK__XE_DOWNED "downed"
#define PCMK__XE_FAILED "failed"
#define PCMK__XE_FAILED_UPDATE "failed_update"
#define PCMK__XE_GENERATION_TUPLE "generation_tuple"
#define PCMK__XE_LRM "lrm"
#define PCMK__XE_LRM_RESOURCE "lrm_resource"
#define PCMK__XE_LRM_RESOURCES "lrm_resources"
#define PCMK__XE_LRM_RSC_OP "lrm_rsc_op"
#define PCMK__XE_MAINTENANCE "maintenance"
#define PCMK__XE_META "meta"
#define PCMK__XE_NODE_STATE "node_state"
#define PCMK__XE_OPTIONS "options"
#define PCMK__XE_PARAM "param"
#define PCMK__XE_PING "ping"
#define PCMK__XE_PING_RESPONSE "ping_response"
#define PCMK__XE_PSEUDO_EVENT "pseudo_event"
#define PCMK__XE_RSC_OP "rsc_op"
#define PCMK__XE_SHUTDOWN "shutdown"
#define PCMK__XE_TICKET_STATE "ticket_state"
#define PCMK__XE_TRANSIENT_ATTRIBUTES "transient_attributes"
#define PCMK__XE_TRANSITION_GRAPH "transition_graph"
// @COMPAT Deprecated since 1.1.12
#define PCMK__XE_ACL_USER "acl_user"
/* @TODO Deprecate somehow. It's undocumented and behaves the same as
* PCMK__XE_CIB in places where it's recognized.
*/
#define PCMK__XE_ALL "all"
// @COMPAT Deprecated since 2.1.7
#define PCMK__XE_DIFF_ADDED "diff-added"
// @COMPAT Deprecated since 2.1.7
#define PCMK__XE_DIFF_REMOVED "diff-removed"
// @COMPAT Deprecated since 1.0.8 (commit 4cb100f)
#define PCMK__XE_LIFETIME "lifetime"
/* @COMPAT Deprecated since 2.0.0; alias for <clone> with PCMK_META_PROMOTABLE
* set to "true"
*/
#define PCMK__XE_PROMOTABLE_LEGACY "master"
// @COMPAT Deprecated since 1.1.12
#define PCMK__XE_ROLE_REF "role_ref"
/*
* XML attribute names used only by internal code
*/
#define PCMK__XA_ATTR_DAMPENING "attr_dampening"
#define PCMK__XA_ATTR_FORCE "attrd_is_force_write"
#define PCMK__XA_ATTR_INTERVAL "attr_clear_interval"
#define PCMK__XA_ATTR_IS_PRIVATE "attr_is_private"
#define PCMK__XA_ATTR_IS_REMOTE "attr_is_remote"
#define PCMK__XA_ATTR_NAME "attr_name"
#define PCMK__XA_ATTR_NODE_ID "attr_host_id"
#define PCMK__XA_ATTR_NODE_NAME "attr_host"
#define PCMK__XA_ATTR_OPERATION "attr_clear_operation"
#define PCMK__XA_ATTR_PATTERN "attr_regex"
#define PCMK__XA_ATTR_RESOURCE "attr_resource"
#define PCMK__XA_ATTR_SECTION "attr_section"
#define PCMK__XA_ATTR_SET "attr_set"
#define PCMK__XA_ATTR_SET_TYPE "attr_set_type"
#define PCMK__XA_ATTR_SYNC_POINT "attr_sync_point"
#define PCMK__XA_ATTR_USER "attr_user"
#define PCMK__XA_ATTR_UUID "attr_key"
#define PCMK__XA_ATTR_VALUE "attr_value"
#define PCMK__XA_ATTR_VERSION "attr_version"
#define PCMK__XA_ATTR_WRITER "attr_writer"
#define PCMK__XA_CALL_ID "call-id"
#define PCMK__XA_CLIENT_NAME "client_name"
#define PCMK__XA_CLIENT_UUID "client_uuid"
#define PCMK__XA_CONFIG_ERRORS "config-errors"
#define PCMK__XA_CONFIG_WARNINGS "config-warnings"
#define PCMK__XA_CONFIRM "confirm"
#define PCMK__XA_CONN_HOST "connection_host"
#define PCMK__XA_CONTENT "content"
-#define PCMK__XA_CRMD "crmd"
#define PCMK__XA_CRMD_STATE "crmd_state"
#define PCMK__XA_CRM_HOST_TO "crm_host_to"
#define PCMK__XA_CRM_LIMIT_MAX "crm-limit-max"
#define PCMK__XA_CRM_LIMIT_MODE "crm-limit-mode"
#define PCMK__XA_CRM_SUBSYSTEM "crm_subsystem"
#define PCMK__XA_CRM_SYS_FROM "crm_sys_from"
#define PCMK__XA_CRM_SYS_TO "crm_sys_to"
#define PCMK__XA_CRM_TASK "crm_task"
#define PCMK__XA_CRM_TGRAPH_IN "crm-tgraph-in"
#define PCMK__XA_CRM_USER "crm_user"
#define PCMK__XA_DC_LEAVING "dc-leaving"
#define PCMK__XA_DIGEST "digest"
#define PCMK__XA_ELECTION_AGE_SEC "election-age-sec"
#define PCMK__XA_ELECTION_AGE_NANO_SEC "election-age-nano-sec"
#define PCMK__XA_ELECTION_ID "election-id"
#define PCMK__XA_ELECTION_OWNER "election-owner"
#define PCMK__XA_GRANTED "granted"
#define PCMK__XA_GRAPH_ERRORS "graph-errors"
#define PCMK__XA_GRAPH_WARNINGS "graph-warnings"
#define PCMK__XA_HIDDEN "hidden"
#define PCMK__XA_IN_CCM "in_ccm"
#define PCMK__XA_JOIN "join"
#define PCMK__XA_JOIN_ID "join_id"
#define PCMK__XA_LONG_ID "long-id"
#define PCMK__XA_MAJOR_VERSION "major_version"
#define PCMK__XA_MINOR_VERSION "minor_version"
#define PCMK__XA_MODE "mode"
#define PCMK__XA_MOON "moon"
#define PCMK__XA_NODE_FENCED "node_fenced"
#define PCMK__XA_NODE_IN_MAINTENANCE "node_in_maintenance"
#define PCMK__XA_NODE_START_STATE "node_start_state"
#define PCMK__XA_NODE_STATE "node_state"
#define PCMK__XA_OP_DIGEST "op-digest"
#define PCMK__XA_OP_FORCE_RESTART "op-force-restart"
#define PCMK__XA_OP_RESTART_DIGEST "op-restart-digest"
#define PCMK__XA_OP_SECURE_DIGEST "op-secure-digest"
#define PCMK__XA_OP_SECURE_PARAMS "op-secure-params"
#define PCMK__XA_OP_STATUS "op-status"
#define PCMK__XA_PACEMAKERD_STATE "pacemakerd_state"
#define PCMK__XA_PASSWORD "password"
#define PCMK__XA_PRIORITY "priority"
#define PCMK__XA_RC_CODE "rc-code"
#define PCMK__XA_REAP "reap"
#define PCMK__XA_REMOTE_NODE "remote_node"
/* Actions to be executed on Pacemaker Remote nodes are routed through the
* controller on the cluster node hosting the remote connection. That cluster
* node is considered the router node for the action.
*/
#define PCMK__XA_ROUTER_NODE "router_node"
#define PCMK__XA_RSC_ID "rsc-id"
#define PCMK__XA_SCHEMA "schema"
#define PCMK__XA_SCHEMAS "schemas"
#define PCMK__XA_SRC "src"
#define PCMK__XA_SUBT "subt" // subtype
#define PCMK__XA_T "t" // type
#define PCMK__XA_TASK "task"
#define PCMK__XA_TRANSITION_KEY "transition-key"
#define PCMK__XA_TRANSITION_MAGIC "transition-magic"
#define PCMK__XA_UPTIME "uptime"
// @COMPAT Used only with v1 patchsets
#define PCMK__XA_CRM_DIFF_MARKER "__crm_diff_marker__"
// @COMPAT Deprecated since 2.1.5
#define PCMK__XA_FIRST_INSTANCE "first-instance"
// @COMPAT Deprecated since 2.1.7
#define PCMK__XA_ORDERING "ordering"
// @COMPAT Deprecated alias for PCMK_XA_PROMOTED_MAX since 2.0.0
#define PCMK__XA_PROMOTED_MAX_LEGACY "masters"
// @COMPAT Deprecated alias for PCMK_XA_PROMOTED_ONLY since 2.0.0
#define PCMK__XA_PROMOTED_ONLY_LEGACY "master_only"
// @COMPAT Deprecated since 1.1.12
#define PCMK__XA_REF "ref"
// @COMPAT Deprecated since 2.1.6
#define PCMK__XA_REPLACE "replace"
// @COMPAT Deprecated since 2.1.5
#define PCMK__XA_RSC_INSTANCE "rsc-instance"
// @COMPAT Deprecated since 1.1.12
#define PCMK__XA_TAG "tag"
// @COMPAT Deprecated since 2.1.5
#define PCMK__XA_THEN_INSTANCE "then-instance"
// @COMPAT Deprecated since 2.1.5
#define PCMK__XA_WITH_RSC_INSTANCE "with-rsc-instance"
/*
* IPC service names that are only used internally
*/
# define PCMK__SERVER_BASED_RO "cib_ro"
# define PCMK__SERVER_BASED_RW "cib_rw"
# define PCMK__SERVER_BASED_SHM "cib_shm"
/*
* IPC commands that can be sent to Pacemaker daemons
*/
#define PCMK__ATTRD_CMD_PEER_REMOVE "peer-remove"
#define PCMK__ATTRD_CMD_UPDATE "update"
#define PCMK__ATTRD_CMD_UPDATE_BOTH "update-both"
#define PCMK__ATTRD_CMD_UPDATE_DELAY "update-delay"
#define PCMK__ATTRD_CMD_QUERY "query"
#define PCMK__ATTRD_CMD_REFRESH "refresh"
#define PCMK__ATTRD_CMD_FLUSH "flush"
#define PCMK__ATTRD_CMD_SYNC "sync"
#define PCMK__ATTRD_CMD_SYNC_RESPONSE "sync-response"
#define PCMK__ATTRD_CMD_CLEAR_FAILURE "clear-failure"
#define PCMK__ATTRD_CMD_CONFIRM "confirm"
#define PCMK__CONTROLD_CMD_NODES "list-nodes"
#endif /* CRM_INTERNAL__H */
diff --git a/lib/common/ipc_controld.c b/lib/common/ipc_controld.c
index ed75032b2e..1da0787605 100644
--- a/lib/common/ipc_controld.c
+++ b/lib/common/ipc_controld.c
@@ -1,660 +1,660 @@
/*
* Copyright 2020-2024 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.
*/
#include <crm_internal.h>
#include <stdio.h>
#include <stdbool.h>
#include <errno.h>
#include <libxml/tree.h>
#include <crm/crm.h>
#include <crm/msg_xml.h>
#include <crm/common/xml.h>
#include <crm/common/ipc.h>
#include <crm/common/ipc_internal.h>
#include <crm/common/ipc_controld.h>
#include "crmcommon_private.h"
struct controld_api_private_s {
char *client_uuid;
unsigned int replies_expected;
};
/*!
* \internal
* \brief Get a string representation of a controller API reply type
*
* \param[in] reply Controller API reply type
*
* \return String representation of a controller API reply type
*/
const char *
pcmk__controld_api_reply2str(enum pcmk_controld_api_reply reply)
{
switch (reply) {
case pcmk_controld_reply_reprobe:
return "reprobe";
case pcmk_controld_reply_info:
return "info";
case pcmk_controld_reply_resource:
return "resource";
case pcmk_controld_reply_ping:
return "ping";
case pcmk_controld_reply_nodes:
return "nodes";
default:
return "unknown";
}
}
// \return Standard Pacemaker return code
static int
new_data(pcmk_ipc_api_t *api)
{
struct controld_api_private_s *private = NULL;
api->api_data = calloc(1, sizeof(struct controld_api_private_s));
if (api->api_data == NULL) {
return errno;
}
private = api->api_data;
/* This is set to the PID because that's how it was always done, but PIDs
* are not unique because clients can be remote. The value appears to be
* unused other than as part of PCMK__XA_CRM_SYS_FROM in IPC requests, which
* is only compared against the internal system names (CRM_SYSTEM_TENGINE,
* etc.), so it shouldn't be a problem.
*/
private->client_uuid = pcmk__getpid_s();
/* @TODO Implement a call ID model similar to the CIB, executor, and fencer
* IPC APIs, so that requests and replies can be matched, and
* duplicate replies can be discarded.
*/
return pcmk_rc_ok;
}
static void
free_data(void *data)
{
free(((struct controld_api_private_s *) data)->client_uuid);
free(data);
}
// \return Standard Pacemaker return code
static int
post_connect(pcmk_ipc_api_t *api)
{
/* The controller currently requires clients to register via a hello
* request, but does not reply back.
*/
struct controld_api_private_s *private = api->api_data;
const char *client_name = crm_system_name? crm_system_name : "client";
xmlNode *hello;
int rc;
hello = create_hello_message(private->client_uuid, client_name,
PCMK__CONTROLD_API_MAJOR,
PCMK__CONTROLD_API_MINOR);
rc = pcmk__send_ipc_request(api, hello);
free_xml(hello);
if (rc != pcmk_rc_ok) {
crm_info("Could not send IPC hello to %s: %s " CRM_XS " rc=%s",
pcmk_ipc_name(api, true), pcmk_rc_str(rc), rc);
} else {
crm_debug("Sent IPC hello to %s", pcmk_ipc_name(api, true));
}
return rc;
}
static void
set_node_info_data(pcmk_controld_api_reply_t *data, xmlNode *msg_data)
{
data->reply_type = pcmk_controld_reply_info;
if (msg_data == NULL) {
return;
}
data->data.node_info.have_quorum =
pcmk__xe_attr_is_true(msg_data, PCMK_XA_HAVE_QUORUM);
data->data.node_info.is_remote =
pcmk__xe_attr_is_true(msg_data, PCMK__XA_REMOTE_NODE);
/* Integer node_info.id is currently valid only for Corosync nodes.
*
* @TODO: Improve handling after crm_node_t is refactored to handle layer-
* specific data better.
*/
crm_element_value_int(msg_data, PCMK_XA_ID, &(data->data.node_info.id));
data->data.node_info.uuid = crm_element_value(msg_data, PCMK_XA_ID);
data->data.node_info.uname = crm_element_value(msg_data, PCMK_XA_UNAME);
- data->data.node_info.state = crm_element_value(msg_data, PCMK__XA_CRMD);
+ data->data.node_info.state = crm_element_value(msg_data, PCMK_XA_CRMD);
}
static void
set_ping_data(pcmk_controld_api_reply_t *data, xmlNode *msg_data)
{
data->reply_type = pcmk_controld_reply_ping;
if (msg_data == NULL) {
return;
}
data->data.ping.sys_from = crm_element_value(msg_data,
PCMK__XA_CRM_SUBSYSTEM);
data->data.ping.fsa_state = crm_element_value(msg_data,
PCMK__XA_CRMD_STATE);
data->data.ping.result = crm_element_value(msg_data, PCMK_XA_RESULT);
}
static void
set_nodes_data(pcmk_controld_api_reply_t *data, xmlNode *msg_data)
{
pcmk_controld_api_node_t *node_info;
data->reply_type = pcmk_controld_reply_nodes;
for (xmlNode *node = first_named_child(msg_data, PCMK_XE_NODE);
node != NULL; node = crm_next_same_xml(node)) {
long long id_ll = 0;
node_info = calloc(1, sizeof(pcmk_controld_api_node_t));
crm_element_value_ll(node, PCMK_XA_ID, &id_ll);
if (id_ll > 0) {
node_info->id = id_ll;
}
node_info->uname = crm_element_value(node, PCMK_XA_UNAME);
node_info->state = crm_element_value(node, PCMK__XA_IN_CCM);
data->data.nodes = g_list_prepend(data->data.nodes, node_info);
}
}
static bool
reply_expected(pcmk_ipc_api_t *api, const xmlNode *request)
{
// We only need to handle commands that API functions can send
return pcmk__str_any_of(crm_element_value(request, PCMK__XA_CRM_TASK),
PCMK__CONTROLD_CMD_NODES,
CRM_OP_LRM_DELETE,
CRM_OP_LRM_FAIL,
CRM_OP_NODE_INFO,
CRM_OP_PING,
CRM_OP_REPROBE,
CRM_OP_RM_NODE_CACHE,
NULL);
}
static bool
dispatch(pcmk_ipc_api_t *api, xmlNode *reply)
{
struct controld_api_private_s *private = api->api_data;
crm_exit_t status = CRM_EX_OK;
xmlNode *msg_data = NULL;
const char *value = NULL;
pcmk_controld_api_reply_t reply_data = {
pcmk_controld_reply_unknown, NULL, NULL,
};
if (pcmk__xe_is(reply, "ack")) {
/* ACKs are trivial responses that do not count toward expected replies,
* and do not have all the fields that validation requires, so skip that
* processing.
*/
return private->replies_expected > 0;
}
if (private->replies_expected > 0) {
private->replies_expected--;
}
// Do some basic validation of the reply
/* @TODO We should be able to verify that value is always a response, but
* currently the controller doesn't always properly set the type. Even
* if we fix the controller, we'll still need to handle replies from
* old versions (feature set could be used to differentiate).
*/
value = crm_element_value(reply, PCMK__XA_SUBT);
if (!pcmk__str_any_of(value, PCMK__VALUE_REQUEST, PCMK__VALUE_RESPONSE,
NULL)) {
crm_info("Unrecognizable message from controller: "
"invalid message type '%s'", pcmk__s(value, ""));
status = CRM_EX_PROTOCOL;
goto done;
}
if (pcmk__str_empty(crm_element_value(reply, PCMK_XA_REFERENCE))) {
crm_info("Unrecognizable message from controller: no reference");
status = CRM_EX_PROTOCOL;
goto done;
}
value = crm_element_value(reply, PCMK__XA_CRM_TASK);
if (pcmk__str_empty(value)) {
crm_info("Unrecognizable message from controller: no command name");
status = CRM_EX_PROTOCOL;
goto done;
}
// Parse useful info from reply
reply_data.feature_set = crm_element_value(reply, PCMK_XA_VERSION);
reply_data.host_from = crm_element_value(reply, PCMK__XA_SRC);
msg_data = get_message_xml(reply, PCMK__XE_CRM_XML);
if (!strcmp(value, CRM_OP_REPROBE)) {
reply_data.reply_type = pcmk_controld_reply_reprobe;
} else if (!strcmp(value, CRM_OP_NODE_INFO)) {
set_node_info_data(&reply_data, msg_data);
} else if (!strcmp(value, CRM_OP_INVOKE_LRM)) {
reply_data.reply_type = pcmk_controld_reply_resource;
reply_data.data.resource.node_state = msg_data;
} else if (!strcmp(value, CRM_OP_PING)) {
set_ping_data(&reply_data, msg_data);
} else if (!strcmp(value, PCMK__CONTROLD_CMD_NODES)) {
set_nodes_data(&reply_data, msg_data);
} else {
crm_info("Unrecognizable message from controller: unknown command '%s'",
value);
status = CRM_EX_PROTOCOL;
}
done:
pcmk__call_ipc_callback(api, pcmk_ipc_event_reply, status, &reply_data);
// Free any reply data that was allocated
if (pcmk__str_eq(value, PCMK__CONTROLD_CMD_NODES, pcmk__str_casei)) {
g_list_free_full(reply_data.data.nodes, free);
}
return false; // No further replies needed
}
pcmk__ipc_methods_t *
pcmk__controld_api_methods(void)
{
pcmk__ipc_methods_t *cmds = calloc(1, sizeof(pcmk__ipc_methods_t));
if (cmds != NULL) {
cmds->new_data = new_data;
cmds->free_data = free_data;
cmds->post_connect = post_connect;
cmds->reply_expected = reply_expected;
cmds->dispatch = dispatch;
}
return cmds;
}
/*!
* \internal
* \brief Create XML for a controller IPC request
*
* \param[in] api Controller connection
* \param[in] op Controller IPC command name
* \param[in] node Node name to set as destination host
* \param[in] msg_data XML to attach to request as message data
*
* \return Newly allocated XML for request
*/
static xmlNode *
create_controller_request(const pcmk_ipc_api_t *api, const char *op,
const char *node, xmlNode *msg_data)
{
struct controld_api_private_s *private = NULL;
const char *sys_to = NULL;
if (api == NULL) {
return NULL;
}
private = api->api_data;
if ((node == NULL) && !strcmp(op, CRM_OP_PING)) {
sys_to = CRM_SYSTEM_DC;
} else {
sys_to = CRM_SYSTEM_CRMD;
}
return create_request(op, msg_data, node, sys_to,
(crm_system_name? crm_system_name : "client"),
private->client_uuid);
}
// \return Standard Pacemaker return code
static int
send_controller_request(pcmk_ipc_api_t *api, const xmlNode *request,
bool reply_is_expected)
{
if (crm_element_value(request, PCMK_XA_REFERENCE) == NULL) {
return EINVAL;
}
if (reply_is_expected) {
struct controld_api_private_s *private = api->api_data;
private->replies_expected++;
}
return pcmk__send_ipc_request(api, request);
}
static xmlNode *
create_reprobe_message_data(const char *target_node, const char *router_node)
{
xmlNode *msg_data;
msg_data = create_xml_node(NULL, "data_for_" CRM_OP_REPROBE);
crm_xml_add(msg_data, PCMK__META_ON_NODE, target_node);
if ((router_node != NULL) && !pcmk__str_eq(router_node, target_node, pcmk__str_casei)) {
crm_xml_add(msg_data, PCMK__XA_ROUTER_NODE, router_node);
}
return msg_data;
}
/*!
* \brief Send a reprobe controller operation
*
* \param[in,out] api Controller connection
* \param[in] target_node Name of node to reprobe
* \param[in] router_node Router node for host
*
* \return Standard Pacemaker return code
* \note Event callback will get a reply of type pcmk_controld_reply_reprobe.
*/
int
pcmk_controld_api_reprobe(pcmk_ipc_api_t *api, const char *target_node,
const char *router_node)
{
xmlNode *request;
xmlNode *msg_data;
int rc = pcmk_rc_ok;
if (api == NULL) {
return EINVAL;
}
if (router_node == NULL) {
router_node = target_node;
}
crm_debug("Sending %s IPC request to reprobe %s via %s",
pcmk_ipc_name(api, true), pcmk__s(target_node, "local node"),
pcmk__s(router_node, "local node"));
msg_data = create_reprobe_message_data(target_node, router_node);
request = create_controller_request(api, CRM_OP_REPROBE, router_node,
msg_data);
rc = send_controller_request(api, request, true);
free_xml(msg_data);
free_xml(request);
return rc;
}
/*!
* \brief Send a "node info" controller operation
*
* \param[in,out] api Controller connection
* \param[in] nodeid ID of node to get info for (or 0 for local node)
*
* \return Standard Pacemaker return code
* \note Event callback will get a reply of type pcmk_controld_reply_info.
*/
int
pcmk_controld_api_node_info(pcmk_ipc_api_t *api, uint32_t nodeid)
{
xmlNode *request;
int rc = pcmk_rc_ok;
request = create_controller_request(api, CRM_OP_NODE_INFO, NULL, NULL);
if (request == NULL) {
return EINVAL;
}
if (nodeid > 0) {
crm_xml_set_id(request, "%lu", (unsigned long) nodeid);
}
rc = send_controller_request(api, request, true);
free_xml(request);
return rc;
}
/*!
* \brief Ask the controller for status
*
* \param[in,out] api Controller connection
* \param[in] node_name Name of node whose status is desired (NULL for DC)
*
* \return Standard Pacemaker return code
* \note Event callback will get a reply of type pcmk_controld_reply_ping.
*/
int
pcmk_controld_api_ping(pcmk_ipc_api_t *api, const char *node_name)
{
xmlNode *request;
int rc = pcmk_rc_ok;
request = create_controller_request(api, CRM_OP_PING, node_name, NULL);
if (request == NULL) {
return EINVAL;
}
rc = send_controller_request(api, request, true);
free_xml(request);
return rc;
}
/*!
* \brief Ask the controller for cluster information
*
* \param[in,out] api Controller connection
*
* \return Standard Pacemaker return code
* \note Event callback will get a reply of type pcmk_controld_reply_nodes.
*/
int
pcmk_controld_api_list_nodes(pcmk_ipc_api_t *api)
{
xmlNode *request;
int rc = EINVAL;
request = create_controller_request(api, PCMK__CONTROLD_CMD_NODES, NULL,
NULL);
if (request != NULL) {
rc = send_controller_request(api, request, true);
free_xml(request);
}
return rc;
}
// \return Standard Pacemaker return code
static int
controller_resource_op(pcmk_ipc_api_t *api, const char *op,
const char *target_node, const char *router_node,
bool cib_only, const char *rsc_id,
const char *rsc_long_id, const char *standard,
const char *provider, const char *type)
{
int rc = pcmk_rc_ok;
char *key;
xmlNode *request, *msg_data, *xml_rsc, *params;
if (api == NULL) {
return EINVAL;
}
if (router_node == NULL) {
router_node = target_node;
}
msg_data = create_xml_node(NULL, PCMK__XE_RSC_OP);
/* The controller logs the transition key from resource op requests, so we
* need to have *something* for it.
* @TODO don't use "crm-resource"
*/
key = pcmk__transition_key(0, getpid(), 0,
"xxxxxxxx-xrsc-opxx-xcrm-resourcexxxx");
crm_xml_add(msg_data, PCMK__XA_TRANSITION_KEY, key);
free(key);
crm_xml_add(msg_data, PCMK__META_ON_NODE, target_node);
if (!pcmk__str_eq(router_node, target_node, pcmk__str_casei)) {
crm_xml_add(msg_data, PCMK__XA_ROUTER_NODE, router_node);
}
if (cib_only) {
// Indicate that only the CIB needs to be cleaned
crm_xml_add(msg_data, PCMK__XA_MODE, PCMK__VALUE_CIB);
}
xml_rsc = create_xml_node(msg_data, PCMK_XE_PRIMITIVE);
crm_xml_add(xml_rsc, PCMK_XA_ID, rsc_id);
crm_xml_add(xml_rsc, PCMK__XA_LONG_ID, rsc_long_id);
crm_xml_add(xml_rsc, PCMK_XA_CLASS, standard);
crm_xml_add(xml_rsc, PCMK_XA_PROVIDER, provider);
crm_xml_add(xml_rsc, PCMK_XA_TYPE, type);
params = create_xml_node(msg_data, PCMK__XE_ATTRIBUTES);
crm_xml_add(params, PCMK_XA_CRM_FEATURE_SET, CRM_FEATURE_SET);
// The controller parses the timeout from the request
key = crm_meta_name(PCMK_META_TIMEOUT);
crm_xml_add(params, key, "60000"); /* 1 minute */ //@TODO pass as arg
free(key);
request = create_controller_request(api, op, router_node, msg_data);
rc = send_controller_request(api, request, true);
free_xml(msg_data);
free_xml(request);
return rc;
}
/*!
* \brief Ask the controller to fail a resource
*
* \param[in,out] api Controller connection
* \param[in] target_node Name of node resource is on
* \param[in] router_node Router node for target
* \param[in] rsc_id ID of resource to fail
* \param[in] rsc_long_id Long ID of resource (if any)
* \param[in] standard Standard of resource
* \param[in] provider Provider of resource (if any)
* \param[in] type Type of resource to fail
*
* \return Standard Pacemaker return code
* \note Event callback will get a reply of type pcmk_controld_reply_resource.
*/
int
pcmk_controld_api_fail(pcmk_ipc_api_t *api,
const char *target_node, const char *router_node,
const char *rsc_id, const char *rsc_long_id,
const char *standard, const char *provider,
const char *type)
{
crm_debug("Sending %s IPC request to fail %s (a.k.a. %s) on %s via %s",
pcmk_ipc_name(api, true), pcmk__s(rsc_id, "unknown resource"),
pcmk__s(rsc_long_id, "no other names"),
pcmk__s(target_node, "unspecified node"),
pcmk__s(router_node, "unspecified node"));
return controller_resource_op(api, CRM_OP_LRM_FAIL, target_node,
router_node, false, rsc_id, rsc_long_id,
standard, provider, type);
}
/*!
* \brief Ask the controller to refresh a resource
*
* \param[in,out] api Controller connection
* \param[in] target_node Name of node resource is on
* \param[in] router_node Router node for target
* \param[in] rsc_id ID of resource to refresh
* \param[in] rsc_long_id Long ID of resource (if any)
* \param[in] standard Standard of resource
* \param[in] provider Provider of resource (if any)
* \param[in] type Type of resource
* \param[in] cib_only If true, clean resource from CIB only
*
* \return Standard Pacemaker return code
* \note Event callback will get a reply of type pcmk_controld_reply_resource.
*/
int
pcmk_controld_api_refresh(pcmk_ipc_api_t *api, const char *target_node,
const char *router_node,
const char *rsc_id, const char *rsc_long_id,
const char *standard, const char *provider,
const char *type, bool cib_only)
{
crm_debug("Sending %s IPC request to refresh %s (a.k.a. %s) on %s via %s",
pcmk_ipc_name(api, true), pcmk__s(rsc_id, "unknown resource"),
pcmk__s(rsc_long_id, "no other names"),
pcmk__s(target_node, "unspecified node"),
pcmk__s(router_node, "unspecified node"));
return controller_resource_op(api, CRM_OP_LRM_DELETE, target_node,
router_node, cib_only, rsc_id, rsc_long_id,
standard, provider, type);
}
/*!
* \brief Get the number of IPC replies currently expected from the controller
*
* \param[in] api Controller IPC API connection
*
* \return Number of replies expected
*/
unsigned int
pcmk_controld_api_replies_expected(const pcmk_ipc_api_t *api)
{
struct controld_api_private_s *private = api->api_data;
return private->replies_expected;
}
/*!
* \brief Create XML for a controller IPC "hello" message
*
* \deprecated This function is deprecated as part of the public C API.
*/
// \todo make this static to this file when breaking API backward compatibility
xmlNode *
create_hello_message(const char *uuid, const char *client_name,
const char *major_version, const char *minor_version)
{
xmlNode *hello_node = NULL;
xmlNode *hello = NULL;
if (pcmk__str_empty(uuid) || pcmk__str_empty(client_name)
|| pcmk__str_empty(major_version) || pcmk__str_empty(minor_version)) {
crm_err("Could not create IPC hello message from %s (UUID %s): "
"missing information",
client_name? client_name : "unknown client",
uuid? uuid : "unknown");
return NULL;
}
hello_node = create_xml_node(NULL, PCMK__XE_OPTIONS);
if (hello_node == NULL) {
crm_err("Could not create IPC hello message from %s (UUID %s): "
"Message data creation failed", client_name, uuid);
return NULL;
}
crm_xml_add(hello_node, PCMK__XA_MAJOR_VERSION, major_version);
crm_xml_add(hello_node, PCMK__XA_MINOR_VERSION, minor_version);
crm_xml_add(hello_node, PCMK__XA_CLIENT_NAME, client_name);
// @TODO Nothing uses this. Drop, or keep for debugging?
crm_xml_add(hello_node, PCMK__XA_CLIENT_UUID, uuid);
hello = create_request(CRM_OP_HELLO, hello_node, NULL, NULL, client_name, uuid);
if (hello == NULL) {
crm_err("Could not create IPC hello message from %s (UUID %s): "
"Request creation failed", client_name, uuid);
return NULL;
}
free_xml(hello_node);
crm_trace("Created hello message from %s (UUID %s)", client_name, uuid);
return hello;
}
diff --git a/lib/pacemaker/pcmk_injections.c b/lib/pacemaker/pcmk_injections.c
index b0c335a240..a64d063189 100644
--- a/lib/pacemaker/pcmk_injections.c
+++ b/lib/pacemaker/pcmk_injections.c
@@ -1,796 +1,796 @@
/*
* Copyright 2009-2024 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 <stdio.h>
#include <unistd.h>
#include <stdlib.h>
#include <sys/stat.h>
#include <sys/param.h>
#include <sys/types.h>
#include <dirent.h>
#include <crm/crm.h>
#include <crm/cib.h>
#include <crm/cib/internal.h>
#include <crm/common/util.h>
#include <crm/common/iso8601.h>
#include <crm/common/xml_internal.h>
#include <crm/lrmd_events.h> // lrmd_event_data_t, etc.
#include <crm/lrmd_internal.h>
#include <crm/pengine/status.h>
#include <pacemaker-internal.h>
#include "libpacemaker_private.h"
bool pcmk__simulate_node_config = false;
#define XPATH_NODE_CONFIG "//" PCMK_XE_NODE "[@" PCMK_XA_UNAME "='%s']"
#define XPATH_NODE_STATE "//" PCMK__XE_NODE_STATE "[@" PCMK_XA_UNAME "='%s']"
#define XPATH_NODE_STATE_BY_ID "//" PCMK__XE_NODE_STATE "[@" PCMK_XA_ID "='%s']"
#define XPATH_RSC_HISTORY XPATH_NODE_STATE \
"//" PCMK__XE_LRM_RESOURCE "[@" PCMK_XA_ID "='%s']"
/*!
* \internal
* \brief Inject a fictitious transient node attribute into scheduler input
*
* \param[in,out] out Output object for displaying error messages
* \param[in,out] cib_node \c PCMK__XE_NODE_STATE XML to inject attribute into
* \param[in] name Transient node attribute name to inject
* \param[in] value Transient node attribute value to inject
*/
static void
inject_transient_attr(pcmk__output_t *out, xmlNode *cib_node,
const char *name, const char *value)
{
xmlNode *attrs = NULL;
xmlNode *instance_attrs = NULL;
const char *node_uuid = ID(cib_node);
out->message(out, "inject-attr", name, value, cib_node);
attrs = first_named_child(cib_node, PCMK__XE_TRANSIENT_ATTRIBUTES);
if (attrs == NULL) {
attrs = create_xml_node(cib_node, PCMK__XE_TRANSIENT_ATTRIBUTES);
crm_xml_add(attrs, PCMK_XA_ID, node_uuid);
}
instance_attrs = first_named_child(attrs, PCMK_XE_INSTANCE_ATTRIBUTES);
if (instance_attrs == NULL) {
instance_attrs = create_xml_node(attrs, PCMK_XE_INSTANCE_ATTRIBUTES);
crm_xml_add(instance_attrs, PCMK_XA_ID, node_uuid);
}
crm_create_nvpair_xml(instance_attrs, NULL, name, value);
}
/*!
* \internal
* \brief Inject a fictitious fail count into a scheduler input
*
* \param[in,out] out Output object for displaying error messages
* \param[in,out] cib_node Node state XML to inject into
* \param[in] resource ID of resource for fail count to inject
* \param[in] task Action name for fail count to inject
* \param[in] interval_ms Action interval (in milliseconds) for fail count
* \param[in] rc Action result for fail count to inject (if 0, or
* 7 when interval_ms is 0, inject nothing)
*/
void
pcmk__inject_failcount(pcmk__output_t *out, xmlNode *cib_node,
const char *resource, const char *task,
guint interval_ms, int rc)
{
if (rc == 0) {
return;
} else if ((rc == 7) && (interval_ms == 0)) {
return;
} else {
char *name = NULL;
char *now = pcmk__ttoa(time(NULL));
name = pcmk__failcount_name(resource, task, interval_ms);
inject_transient_attr(out, cib_node, name, "value++");
free(name);
name = pcmk__lastfailure_name(resource, task, interval_ms);
inject_transient_attr(out, cib_node, name, now);
free(name);
free(now);
}
}
/*!
* \internal
* \brief Create a CIB configuration entry for a fictitious node
*
* \param[in,out] cib_conn CIB object to use
* \param[in] node Node name to use
*/
static void
create_node_entry(cib_t *cib_conn, const char *node)
{
int rc = pcmk_ok;
char *xpath = crm_strdup_printf(XPATH_NODE_CONFIG, node);
rc = cib_conn->cmds->query(cib_conn, xpath, NULL,
cib_xpath|cib_sync_call|cib_scope_local);
if (rc == -ENXIO) { // Only add if not already existing
xmlNode *cib_object = create_xml_node(NULL, PCMK_XE_NODE);
crm_xml_add(cib_object, PCMK_XA_ID, node); // Use node name as ID
crm_xml_add(cib_object, PCMK_XA_UNAME, node);
cib_conn->cmds->create(cib_conn, PCMK_XE_NODES, cib_object,
cib_sync_call|cib_scope_local);
/* Not bothering with subsequent query to see if it exists,
we'll bomb out later in the call to query_node_uuid()... */
free_xml(cib_object);
}
free(xpath);
}
/*!
* \internal
* \brief Synthesize a fake executor event for an action
*
* \param[in] cib_resource XML for any existing resource action history
* \param[in] task Name of action to synthesize
* \param[in] interval_ms Interval of action to synthesize
* \param[in] outcome Result of action to synthesize
*
* \return Newly allocated executor event
* \note It is the caller's responsibility to free the result with
* lrmd_free_event().
*/
static lrmd_event_data_t *
create_op(const xmlNode *cib_resource, const char *task, guint interval_ms,
int outcome)
{
lrmd_event_data_t *op = NULL;
xmlNode *xop = NULL;
op = lrmd_new_event(ID(cib_resource), task, interval_ms);
lrmd__set_result(op, outcome, PCMK_EXEC_DONE, "Simulated action result");
op->params = NULL; // Not needed for simulation purposes
op->t_run = (unsigned int) time(NULL);
op->t_rcchange = op->t_run;
// Use a call ID higher than any existing history entries
op->call_id = 0;
for (xop = pcmk__xe_first_child(cib_resource); xop != NULL;
xop = pcmk__xe_next(xop)) {
int tmp = 0;
crm_element_value_int(xop, PCMK__XA_CALL_ID, &tmp);
if (tmp > op->call_id) {
op->call_id = tmp;
}
}
op->call_id++;
return op;
}
/*!
* \internal
* \brief Inject a fictitious resource history entry into a scheduler input
*
* \param[in,out] cib_resource Resource history XML to inject entry into
* \param[in,out] op Action result to inject
* \param[in] target_rc Expected result for action to inject
*
* \return XML of injected resource history entry
*/
xmlNode *
pcmk__inject_action_result(xmlNode *cib_resource, lrmd_event_data_t *op,
int target_rc)
{
return pcmk__create_history_xml(cib_resource, op, CRM_FEATURE_SET,
target_rc, NULL, crm_system_name);
}
/*!
* \internal
* \brief Inject a fictitious node into a scheduler input
*
* \param[in,out] cib_conn Scheduler input CIB to inject node into
* \param[in] node Name of node to inject
* \param[in] uuid UUID of node to inject
*
* \return XML of \c PCMK__XE_NODE_STATE entry for new node
* \note If the global pcmk__simulate_node_config has been set to true, a
* node entry in the configuration section will be added, as well as a
* node state entry in the status section.
*/
xmlNode *
pcmk__inject_node(cib_t *cib_conn, const char *node, const char *uuid)
{
int rc = pcmk_ok;
xmlNode *cib_object = NULL;
char *xpath = crm_strdup_printf(XPATH_NODE_STATE, node);
bool duplicate = false;
char *found_uuid = NULL;
if (pcmk__simulate_node_config) {
create_node_entry(cib_conn, node);
}
rc = cib_conn->cmds->query(cib_conn, xpath, &cib_object,
cib_xpath|cib_sync_call|cib_scope_local);
if ((cib_object != NULL) && (ID(cib_object) == NULL)) {
crm_err("Detected multiple " PCMK__XE_NODE_STATE " entries for "
"xpath=%s, bailing",
xpath);
duplicate = true;
goto done;
}
if (rc == -ENXIO) {
if (uuid == NULL) {
query_node_uuid(cib_conn, node, &found_uuid, NULL);
} else {
found_uuid = strdup(uuid);
}
if (found_uuid) {
char *xpath_by_uuid = crm_strdup_printf(XPATH_NODE_STATE_BY_ID,
found_uuid);
/* It's possible that a PCMK__XE_NODE_STATE entry doesn't have a
* PCMK_XA_UNAME yet
*/
rc = cib_conn->cmds->query(cib_conn, xpath_by_uuid, &cib_object,
cib_xpath|cib_sync_call|cib_scope_local);
if ((cib_object != NULL) && (ID(cib_object) == NULL)) {
crm_err("Can't inject node state for %s because multiple "
"state entries found for ID %s", node, found_uuid);
duplicate = true;
free(xpath_by_uuid);
goto done;
} else if (cib_object != NULL) {
crm_xml_add(cib_object, PCMK_XA_UNAME, node);
rc = cib_conn->cmds->modify(cib_conn, PCMK_XE_STATUS,
cib_object,
cib_sync_call|cib_scope_local);
}
free(xpath_by_uuid);
}
}
if (rc == -ENXIO) {
cib_object = create_xml_node(NULL, PCMK__XE_NODE_STATE);
crm_xml_add(cib_object, PCMK_XA_ID, found_uuid);
crm_xml_add(cib_object, PCMK_XA_UNAME, node);
cib_conn->cmds->create(cib_conn, PCMK_XE_STATUS, cib_object,
cib_sync_call|cib_scope_local);
free_xml(cib_object);
rc = cib_conn->cmds->query(cib_conn, xpath, &cib_object,
cib_xpath|cib_sync_call|cib_scope_local);
crm_trace("Injecting node state for %s (rc=%d)", node, rc);
}
done:
free(found_uuid);
free(xpath);
if (duplicate) {
crm_log_xml_warn(cib_object, "Duplicates");
crm_exit(CRM_EX_SOFTWARE);
return NULL; // not reached, but makes static analysis happy
}
CRM_ASSERT(rc == pcmk_ok);
return cib_object;
}
/*!
* \internal
* \brief Inject a fictitious node state change into a scheduler input
*
* \param[in,out] cib_conn Scheduler input CIB to inject into
* \param[in] node Name of node to inject change for
* \param[in] up If true, change state to online, otherwise offline
*
* \return XML of changed (or added) node state entry
*/
xmlNode *
pcmk__inject_node_state_change(cib_t *cib_conn, const char *node, bool up)
{
xmlNode *cib_node = pcmk__inject_node(cib_conn, node, NULL);
if (up) {
pcmk__xe_set_props(cib_node,
PCMK__XA_IN_CCM, PCMK_VALUE_TRUE,
- PCMK__XA_CRMD, ONLINESTATUS,
+ PCMK_XA_CRMD, ONLINESTATUS,
PCMK__XA_JOIN, CRMD_JOINSTATE_MEMBER,
PCMK_XA_EXPECTED, CRMD_JOINSTATE_MEMBER,
NULL);
} else {
pcmk__xe_set_props(cib_node,
PCMK__XA_IN_CCM, PCMK_VALUE_FALSE,
- PCMK__XA_CRMD, OFFLINESTATUS,
+ PCMK_XA_CRMD, OFFLINESTATUS,
PCMK__XA_JOIN, CRMD_JOINSTATE_DOWN,
PCMK_XA_EXPECTED, CRMD_JOINSTATE_DOWN,
NULL);
}
crm_xml_add(cib_node, PCMK_XA_CRM_DEBUG_ORIGIN, crm_system_name);
return cib_node;
}
/*!
* \internal
* \brief Check whether a node has history for a given resource
*
* \param[in,out] cib_node Node state XML to check
* \param[in] resource Resource name to check for
*
* \return Resource's \c PCMK__XE_LRM_RESOURCE XML entry beneath \p cib_node if
* found, otherwise \c NULL
*/
static xmlNode *
find_resource_xml(xmlNode *cib_node, const char *resource)
{
const char *node = crm_element_value(cib_node, PCMK_XA_UNAME);
char *xpath = crm_strdup_printf(XPATH_RSC_HISTORY, node, resource);
xmlNode *match = get_xpath_object(xpath, cib_node, LOG_TRACE);
free(xpath);
return match;
}
/*!
* \internal
* \brief Inject a resource history element into a scheduler input
*
* \param[in,out] out Output object for displaying error messages
* \param[in,out] cib_node Node state XML to inject resource history entry into
* \param[in] resource ID (in configuration) of resource to inject
* \param[in] lrm_name ID as used in history (could be clone instance)
* \param[in] rclass Resource agent class of resource to inject
* \param[in] rtype Resource agent type of resource to inject
* \param[in] rprovider Resource agent provider of resource to inject
*
* \return XML of injected resource history element
* \note If a history element already exists under either \p resource or
* \p lrm_name, this will return it rather than injecting a new one.
*/
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)
{
xmlNode *lrm = NULL;
xmlNode *container = NULL;
xmlNode *cib_resource = NULL;
cib_resource = find_resource_xml(cib_node, resource);
if (cib_resource != NULL) {
/* If an existing LRM history entry uses the resource name,
* continue using it, even if lrm_name is different.
*/
return cib_resource;
}
// Check for history entry under preferred name
if (strcmp(resource, lrm_name) != 0) {
cib_resource = find_resource_xml(cib_node, lrm_name);
if (cib_resource != NULL) {
return cib_resource;
}
}
if ((rclass == NULL) || (rtype == NULL)) {
// @TODO query configuration for class, provider, type
out->err(out,
"Resource %s not found in the status section of %s "
"(supply class and type to continue)",
resource, ID(cib_node));
return NULL;
} else if (!pcmk__strcase_any_of(rclass,
PCMK_RESOURCE_CLASS_OCF,
PCMK_RESOURCE_CLASS_STONITH,
PCMK_RESOURCE_CLASS_SERVICE,
PCMK_RESOURCE_CLASS_UPSTART,
PCMK_RESOURCE_CLASS_SYSTEMD,
PCMK_RESOURCE_CLASS_LSB, NULL)) {
out->err(out, "Invalid class for %s: %s", resource, rclass);
return NULL;
} else if (pcmk_is_set(pcmk_get_ra_caps(rclass), pcmk_ra_cap_provider)
&& (rprovider == NULL)) {
// @TODO query configuration for provider
out->err(out, "Please specify the provider for resource %s", resource);
return NULL;
}
crm_info("Injecting new resource %s into node state '%s'",
lrm_name, ID(cib_node));
lrm = first_named_child(cib_node, PCMK__XE_LRM);
if (lrm == NULL) {
const char *node_uuid = ID(cib_node);
lrm = create_xml_node(cib_node, PCMK__XE_LRM);
crm_xml_add(lrm, PCMK_XA_ID, node_uuid);
}
container = first_named_child(lrm, PCMK__XE_LRM_RESOURCES);
if (container == NULL) {
container = create_xml_node(lrm, PCMK__XE_LRM_RESOURCES);
}
cib_resource = create_xml_node(container, PCMK__XE_LRM_RESOURCE);
// If we're creating a new entry, use the preferred name
crm_xml_add(cib_resource, PCMK_XA_ID, lrm_name);
crm_xml_add(cib_resource, PCMK_XA_CLASS, rclass);
crm_xml_add(cib_resource, PCMK_XA_PROVIDER, rprovider);
crm_xml_add(cib_resource, PCMK_XA_TYPE, rtype);
return cib_resource;
}
static int
find_ticket_state(pcmk__output_t *out, cib_t *the_cib, const char *ticket_id,
xmlNode **ticket_state_xml)
{
int rc = pcmk_ok;
xmlNode *xml_search = NULL;
GString *xpath = g_string_sized_new(256);
CRM_ASSERT(ticket_state_xml != NULL);
*ticket_state_xml = NULL;
g_string_append(xpath,
"/" PCMK_XE_CIB "/" PCMK_XE_STATUS "/" PCMK_XE_TICKETS);
if (ticket_id) {
pcmk__g_strcat(xpath,
"/" PCMK__XE_TICKET_STATE
"[@" PCMK_XA_ID "=\"", ticket_id, "\"]", NULL);
}
rc = the_cib->cmds->query(the_cib, (const char *) xpath->str, &xml_search,
cib_sync_call|cib_scope_local|cib_xpath);
g_string_free(xpath, TRUE);
if (rc != pcmk_ok) {
return rc;
}
crm_log_xml_debug(xml_search, "Match");
if ((xml_search->children != NULL) && (ticket_id != NULL)) {
out->err(out, "Multiple " PCMK__XE_TICKET_STATE "s match ticket_id=%s",
ticket_id);
}
*ticket_state_xml = xml_search;
return rc;
}
/*!
* \internal
* \brief Inject a ticket attribute into ticket state
*
* \param[in,out] out Output object for displaying error messages
* \param[in] ticket_id Ticket whose state should be changed
* \param[in] attr_name Ticket attribute name to inject
* \param[in] attr_value Boolean value of ticket attribute to inject
* \param[in,out] cib CIB object to use
*
* \return Standard Pacemaker return code
*/
static int
set_ticket_state_attr(pcmk__output_t *out, const char *ticket_id,
const char *attr_name, bool attr_value, cib_t *cib)
{
int rc = pcmk_rc_ok;
xmlNode *xml_top = NULL;
xmlNode *ticket_state_xml = NULL;
// Check for an existing ticket state entry
rc = find_ticket_state(out, cib, ticket_id, &ticket_state_xml);
rc = pcmk_legacy2rc(rc);
if (rc == pcmk_rc_ok) { // Ticket state found, use it
crm_debug("Injecting attribute into existing ticket state %s",
ticket_id);
xml_top = ticket_state_xml;
} else if (rc == ENXIO) { // No ticket state, create it
xmlNode *xml_obj = NULL;
xml_top = create_xml_node(NULL, PCMK_XE_STATUS);
xml_obj = create_xml_node(xml_top, PCMK_XE_TICKETS);
ticket_state_xml = create_xml_node(xml_obj, PCMK__XE_TICKET_STATE);
crm_xml_add(ticket_state_xml, PCMK_XA_ID, ticket_id);
} else { // Error
return rc;
}
// Add the attribute to the ticket state
pcmk__xe_set_bool_attr(ticket_state_xml, attr_name, attr_value);
crm_log_xml_debug(xml_top, "Update");
// Commit the change to the CIB
rc = cib->cmds->modify(cib, PCMK_XE_STATUS, xml_top,
cib_sync_call|cib_scope_local);
rc = pcmk_legacy2rc(rc);
free_xml(xml_top);
return rc;
}
/*!
* \internal
* \brief Inject a fictitious action into the cluster
*
* \param[in,out] out Output object for displaying error messages
* \param[in] spec Action specification to inject
* \param[in,out] cib CIB object for scheduler input
* \param[in] scheduler Scheduler data
*/
static void
inject_action(pcmk__output_t *out, const char *spec, cib_t *cib,
const pcmk_scheduler_t *scheduler)
{
int rc;
int outcome = PCMK_OCF_OK;
guint interval_ms = 0;
char *key = NULL;
char *node = NULL;
char *task = NULL;
char *resource = NULL;
const char *rtype = NULL;
const char *rclass = NULL;
const char *rprovider = NULL;
xmlNode *cib_op = NULL;
xmlNode *cib_node = NULL;
xmlNode *cib_resource = NULL;
const pcmk_resource_t *rsc = NULL;
lrmd_event_data_t *op = NULL;
out->message(out, "inject-spec", spec);
key = calloc(1, strlen(spec) + 1);
node = calloc(1, strlen(spec) + 1);
rc = sscanf(spec, "%[^@]@%[^=]=%d", key, node, &outcome);
if (rc != 3) {
out->err(out, "Invalid operation spec: %s. Only found %d fields",
spec, rc);
goto done;
}
parse_op_key(key, &resource, &task, &interval_ms);
rsc = pe_find_resource(scheduler->resources, resource);
if (rsc == NULL) {
out->err(out, "Invalid resource name: %s", resource);
goto done;
}
rclass = crm_element_value(rsc->xml, PCMK_XA_CLASS);
rtype = crm_element_value(rsc->xml, PCMK_XA_TYPE);
rprovider = crm_element_value(rsc->xml, PCMK_XA_PROVIDER);
cib_node = pcmk__inject_node(cib, node, NULL);
CRM_ASSERT(cib_node != NULL);
pcmk__inject_failcount(out, cib_node, resource, task, interval_ms, outcome);
cib_resource = pcmk__inject_resource_history(out, cib_node,
resource, resource,
rclass, rtype, rprovider);
CRM_ASSERT(cib_resource != NULL);
op = create_op(cib_resource, task, interval_ms, outcome);
CRM_ASSERT(op != NULL);
cib_op = pcmk__inject_action_result(cib_resource, op, 0);
CRM_ASSERT(cib_op != NULL);
lrmd_free_event(op);
rc = cib->cmds->modify(cib, PCMK_XE_STATUS, cib_node,
cib_sync_call|cib_scope_local);
CRM_ASSERT(rc == pcmk_ok);
done:
free(task);
free(node);
free(key);
}
/*!
* \internal
* \brief Inject fictitious scheduler inputs
*
* \param[in,out] scheduler Scheduler data
* \param[in,out] cib CIB object for scheduler input to modify
* \param[in] injections Injections to apply
*/
void
pcmk__inject_scheduler_input(pcmk_scheduler_t *scheduler, cib_t *cib,
const pcmk_injections_t *injections)
{
int rc = pcmk_ok;
const GList *iter = NULL;
xmlNode *cib_node = NULL;
pcmk__output_t *out = scheduler->priv;
out->message(out, "inject-modify-config", injections->quorum,
injections->watchdog);
if (injections->quorum != NULL) {
xmlNode *top = create_xml_node(NULL, PCMK_XE_CIB);
/* crm_xml_add(top, PCMK_XA_DC_UUID, dc_uuid); */
crm_xml_add(top, PCMK_XA_HAVE_QUORUM, injections->quorum);
rc = cib->cmds->modify(cib, NULL, top, cib_sync_call|cib_scope_local);
CRM_ASSERT(rc == pcmk_ok);
}
if (injections->watchdog != NULL) {
rc = cib__update_node_attr(out, cib, cib_sync_call|cib_scope_local,
PCMK_XE_CRM_CONFIG, NULL, NULL, NULL,
NULL, PCMK_OPT_HAVE_WATCHDOG,
injections->watchdog, NULL, NULL);
CRM_ASSERT(rc == pcmk_rc_ok);
}
for (iter = injections->node_up; iter != NULL; iter = iter->next) {
const char *node = (const char *) iter->data;
out->message(out, "inject-modify-node", "Online", node);
cib_node = pcmk__inject_node_state_change(cib, node, true);
CRM_ASSERT(cib_node != NULL);
rc = cib->cmds->modify(cib, PCMK_XE_STATUS, cib_node,
cib_sync_call|cib_scope_local);
CRM_ASSERT(rc == pcmk_ok);
free_xml(cib_node);
}
for (iter = injections->node_down; iter != NULL; iter = iter->next) {
const char *node = (const char *) iter->data;
char *xpath = NULL;
out->message(out, "inject-modify-node", "Offline", node);
cib_node = pcmk__inject_node_state_change(cib, node, false);
CRM_ASSERT(cib_node != NULL);
rc = cib->cmds->modify(cib, PCMK_XE_STATUS, cib_node,
cib_sync_call|cib_scope_local);
CRM_ASSERT(rc == pcmk_ok);
free_xml(cib_node);
xpath = crm_strdup_printf("//" PCMK__XE_NODE_STATE
"[@" PCMK_XA_UNAME "='%s']"
"/" PCMK__XE_LRM,
node);
cib->cmds->remove(cib, xpath, NULL,
cib_xpath|cib_sync_call|cib_scope_local);
free(xpath);
xpath = crm_strdup_printf("//" PCMK__XE_NODE_STATE
"[@" PCMK_XA_UNAME "='%s']"
"/" PCMK__XE_TRANSIENT_ATTRIBUTES,
node);
cib->cmds->remove(cib, xpath, NULL,
cib_xpath|cib_sync_call|cib_scope_local);
free(xpath);
}
for (iter = injections->node_fail; iter != NULL; iter = iter->next) {
const char *node = (const char *) iter->data;
out->message(out, "inject-modify-node", "Failing", node);
cib_node = pcmk__inject_node_state_change(cib, node, true);
crm_xml_add(cib_node, PCMK__XA_IN_CCM, PCMK_VALUE_FALSE);
CRM_ASSERT(cib_node != NULL);
rc = cib->cmds->modify(cib, PCMK_XE_STATUS, cib_node,
cib_sync_call|cib_scope_local);
CRM_ASSERT(rc == pcmk_ok);
free_xml(cib_node);
}
for (iter = injections->ticket_grant; iter != NULL; iter = iter->next) {
const char *ticket_id = (const char *) iter->data;
out->message(out, "inject-modify-ticket", "Granting", ticket_id);
rc = set_ticket_state_attr(out, ticket_id, PCMK__XA_GRANTED, true, cib);
CRM_ASSERT(rc == pcmk_rc_ok);
}
for (iter = injections->ticket_revoke; iter != NULL; iter = iter->next) {
const char *ticket_id = (const char *) iter->data;
out->message(out, "inject-modify-ticket", "Revoking", ticket_id);
rc = set_ticket_state_attr(out, ticket_id, PCMK__XA_GRANTED, false,
cib);
CRM_ASSERT(rc == pcmk_rc_ok);
}
for (iter = injections->ticket_standby; iter != NULL; iter = iter->next) {
const char *ticket_id = (const char *) iter->data;
out->message(out, "inject-modify-ticket", "Standby", ticket_id);
rc = set_ticket_state_attr(out, ticket_id, PCMK_XA_STANDBY, true, cib);
CRM_ASSERT(rc == pcmk_rc_ok);
}
for (iter = injections->ticket_activate; iter != NULL; iter = iter->next) {
const char *ticket_id = (const char *) iter->data;
out->message(out, "inject-modify-ticket", "Activating", ticket_id);
rc = set_ticket_state_attr(out, ticket_id, PCMK_XA_STANDBY, false, cib);
CRM_ASSERT(rc == pcmk_rc_ok);
}
for (iter = injections->op_inject; iter != NULL; iter = iter->next) {
inject_action(out, (const char *) iter->data, cib, scheduler);
}
if (!out->is_quiet(out)) {
out->end_list(out);
}
}
void
pcmk_free_injections(pcmk_injections_t *injections)
{
if (injections == NULL) {
return;
}
g_list_free_full(injections->node_up, g_free);
g_list_free_full(injections->node_down, g_free);
g_list_free_full(injections->node_fail, g_free);
g_list_free_full(injections->op_fail, g_free);
g_list_free_full(injections->op_inject, g_free);
g_list_free_full(injections->ticket_grant, g_free);
g_list_free_full(injections->ticket_revoke, g_free);
g_list_free_full(injections->ticket_standby, g_free);
g_list_free_full(injections->ticket_activate, g_free);
free(injections->quorum);
free(injections->watchdog);
free(injections);
}
diff --git a/lib/pacemaker/pcmk_output.c b/lib/pacemaker/pcmk_output.c
index 993475eda4..0525b94833 100644
--- a/lib/pacemaker/pcmk_output.c
+++ b/lib/pacemaker/pcmk_output.c
@@ -1,2408 +1,2408 @@
/*
* Copyright 2019-2024 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/common/output.h>
#include <crm/common/results.h>
#include <crm/msg_xml.h>
#include <crm/stonith-ng.h>
#include <crm/fencing/internal.h>
#include <crm/pengine/internal.h>
#include <libxml/tree.h>
#include <pacemaker-internal.h>
#include <inttypes.h>
#include <stdint.h>
static char *
colocations_header(pcmk_resource_t *rsc, pcmk__colocation_t *cons,
bool dependents) {
char *retval = NULL;
if (cons->primary_role > pcmk_role_started) {
retval = crm_strdup_printf("%s (score=%s, %s role=%s, id=%s)",
rsc->id, pcmk_readable_score(cons->score),
(dependents? "needs" : "with"),
pcmk_role_text(cons->primary_role),
cons->id);
} else {
retval = crm_strdup_printf("%s (score=%s, id=%s)",
rsc->id, pcmk_readable_score(cons->score),
cons->id);
}
return retval;
}
static void
colocations_xml_node(pcmk__output_t *out, pcmk_resource_t *rsc,
pcmk__colocation_t *cons) {
xmlNodePtr node = NULL;
node = pcmk__output_create_xml_node(out, PCMK_XE_RSC_COLOCATION,
PCMK_XA_ID, cons->id,
PCMK_XA_RSC, cons->dependent->id,
PCMK_XA_WITH_RSC, cons->primary->id,
PCMK_XA_SCORE,
pcmk_readable_score(cons->score),
NULL);
if (cons->node_attribute) {
xmlSetProp(node, (pcmkXmlStr) PCMK_XA_NODE_ATTRIBUTE,
(pcmkXmlStr) cons->node_attribute);
}
if (cons->dependent_role != pcmk_role_unknown) {
xmlSetProp(node, (pcmkXmlStr) PCMK_XA_RSC_ROLE,
(pcmkXmlStr) pcmk_role_text(cons->dependent_role));
}
if (cons->primary_role != pcmk_role_unknown) {
xmlSetProp(node, (pcmkXmlStr) PCMK_XA_WITH_RSC_ROLE,
(pcmkXmlStr) pcmk_role_text(cons->primary_role));
}
}
static int
do_locations_list_xml(pcmk__output_t *out, pcmk_resource_t *rsc,
bool add_header)
{
GList *lpc = NULL;
GList *list = rsc->rsc_location;
int rc = pcmk_rc_no_output;
for (lpc = list; lpc != NULL; lpc = lpc->next) {
pcmk__location_t *cons = lpc->data;
GList *lpc2 = NULL;
for (lpc2 = cons->nodes; lpc2 != NULL; lpc2 = lpc2->next) {
pcmk_node_t *node = (pcmk_node_t *) lpc2->data;
if (add_header) {
PCMK__OUTPUT_LIST_HEADER(out, false, rc, "locations");
}
pcmk__output_create_xml_node(out, PCMK_XE_RSC_LOCATION,
PCMK_XA_NODE, node->details->uname,
PCMK_XA_RSC, rsc->id,
PCMK_XA_ID, cons->id,
PCMK_XA_SCORE,
pcmk_readable_score(node->weight),
NULL);
}
}
if (add_header) {
PCMK__OUTPUT_LIST_FOOTER(out, rc);
}
return rc;
}
PCMK__OUTPUT_ARGS("rsc-action-item", "const char *", "pcmk_resource_t *",
"pcmk_node_t *", "pcmk_node_t *", "pcmk_action_t *",
"pcmk_action_t *")
static int
rsc_action_item(pcmk__output_t *out, va_list args)
{
const char *change = va_arg(args, const char *);
pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *);
pcmk_node_t *origin = va_arg(args, pcmk_node_t *);
pcmk_node_t *destination = va_arg(args, pcmk_node_t *);
pcmk_action_t *action = va_arg(args, pcmk_action_t *);
pcmk_action_t *source = va_arg(args, pcmk_action_t *);
int len = 0;
char *reason = NULL;
char *details = NULL;
bool same_host = false;
bool same_role = false;
bool need_role = false;
static int rsc_width = 5;
static int detail_width = 5;
CRM_ASSERT(action);
CRM_ASSERT(destination != NULL || origin != NULL);
if (source == NULL) {
source = action;
}
len = strlen(rsc->id);
if (len > rsc_width) {
rsc_width = len + 2;
}
if ((rsc->role > pcmk_role_started)
|| (rsc->next_role > pcmk_role_unpromoted)) {
need_role = true;
}
if (pcmk__same_node(origin, destination)) {
same_host = true;
}
if (rsc->role == rsc->next_role) {
same_role = true;
}
if (need_role && (origin == NULL)) {
/* Starting and promoting a promotable clone instance */
details = crm_strdup_printf("%s -> %s %s", pcmk_role_text(rsc->role),
pcmk_role_text(rsc->next_role),
pcmk__node_name(destination));
} else if (origin == NULL) {
/* Starting a resource */
details = crm_strdup_printf("%s", pcmk__node_name(destination));
} else if (need_role && (destination == NULL)) {
/* Stopping a promotable clone instance */
details = crm_strdup_printf("%s %s", pcmk_role_text(rsc->role),
pcmk__node_name(origin));
} else if (destination == NULL) {
/* Stopping a resource */
details = crm_strdup_printf("%s", pcmk__node_name(origin));
} else if (need_role && same_role && same_host) {
/* Recovering, restarting or re-promoting a promotable clone instance */
details = crm_strdup_printf("%s %s", pcmk_role_text(rsc->role),
pcmk__node_name(origin));
} else if (same_role && same_host) {
/* Recovering or Restarting a normal resource */
details = crm_strdup_printf("%s", pcmk__node_name(origin));
} else if (need_role && same_role) {
/* Moving a promotable clone instance */
details = crm_strdup_printf("%s -> %s %s", pcmk__node_name(origin),
pcmk__node_name(destination),
pcmk_role_text(rsc->role));
} else if (same_role) {
/* Moving a normal resource */
details = crm_strdup_printf("%s -> %s", pcmk__node_name(origin),
pcmk__node_name(destination));
} else if (same_host) {
/* Promoting or demoting a promotable clone instance */
details = crm_strdup_printf("%s -> %s %s", pcmk_role_text(rsc->role),
pcmk_role_text(rsc->next_role),
pcmk__node_name(origin));
} else {
/* Moving and promoting/demoting */
details = crm_strdup_printf("%s %s -> %s %s",
pcmk_role_text(rsc->role),
pcmk__node_name(origin),
pcmk_role_text(rsc->next_role),
pcmk__node_name(destination));
}
len = strlen(details);
if (len > detail_width) {
detail_width = len;
}
if ((source->reason != NULL)
&& !pcmk_is_set(action->flags, pcmk_action_runnable)) {
reason = crm_strdup_printf("due to %s (blocked)", source->reason);
} else if (source->reason) {
reason = crm_strdup_printf("due to %s", source->reason);
} else if (!pcmk_is_set(action->flags, pcmk_action_runnable)) {
reason = strdup("blocked");
}
out->list_item(out, NULL, "%-8s %-*s ( %*s )%s%s",
change, rsc_width, rsc->id, detail_width, details,
((reason == NULL)? "" : " "), pcmk__s(reason, ""));
free(details);
free(reason);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("rsc-action-item", "const char *", "pcmk_resource_t *",
"pcmk_node_t *", "pcmk_node_t *", "pcmk_action_t *",
"pcmk_action_t *")
static int
rsc_action_item_xml(pcmk__output_t *out, va_list args)
{
const char *change = va_arg(args, const char *);
pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *);
pcmk_node_t *origin = va_arg(args, pcmk_node_t *);
pcmk_node_t *destination = va_arg(args, pcmk_node_t *);
pcmk_action_t *action = va_arg(args, pcmk_action_t *);
pcmk_action_t *source = va_arg(args, pcmk_action_t *);
char *change_str = NULL;
bool same_host = false;
bool same_role = false;
bool need_role = false;
xmlNode *xml = NULL;
CRM_ASSERT(action);
CRM_ASSERT(destination != NULL || origin != NULL);
if (source == NULL) {
source = action;
}
if ((rsc->role > pcmk_role_started)
|| (rsc->next_role > pcmk_role_unpromoted)) {
need_role = true;
}
if (pcmk__same_node(origin, destination)) {
same_host = true;
}
if (rsc->role == rsc->next_role) {
same_role = true;
}
change_str = g_ascii_strdown(change, -1);
xml = pcmk__output_create_xml_node(out, PCMK_XE_RSC_ACTION,
PCMK_XA_ACTION, change_str,
PCMK_XA_RESOURCE, rsc->id,
NULL);
g_free(change_str);
if (need_role && (origin == NULL)) {
/* Starting and promoting a promotable clone instance */
pcmk__xe_set_props(xml,
PCMK_XA_ROLE, pcmk_role_text(rsc->role),
PCMK_XA_NEXT_ROLE, pcmk_role_text(rsc->next_role),
PCMK_XA_DEST, destination->details->uname,
NULL);
} else if (origin == NULL) {
/* Starting a resource */
crm_xml_add(xml, PCMK_XA_NODE, destination->details->uname);
} else if (need_role && (destination == NULL)) {
/* Stopping a promotable clone instance */
pcmk__xe_set_props(xml,
PCMK_XA_ROLE, pcmk_role_text(rsc->role),
PCMK_XA_NODE, origin->details->uname,
NULL);
} else if (destination == NULL) {
/* Stopping a resource */
crm_xml_add(xml, PCMK_XA_NODE, origin->details->uname);
} else if (need_role && same_role && same_host) {
/* Recovering, restarting or re-promoting a promotable clone instance */
pcmk__xe_set_props(xml,
PCMK_XA_ROLE, pcmk_role_text(rsc->role),
PCMK_XA_SOURCE, origin->details->uname,
NULL);
} else if (same_role && same_host) {
/* Recovering or Restarting a normal resource */
crm_xml_add(xml, PCMK_XA_SOURCE, origin->details->uname);
} else if (need_role && same_role) {
/* Moving a promotable clone instance */
pcmk__xe_set_props(xml,
PCMK_XA_SOURCE, origin->details->uname,
PCMK_XA_DEST, destination->details->uname,
PCMK_XA_ROLE, pcmk_role_text(rsc->role),
NULL);
} else if (same_role) {
/* Moving a normal resource */
pcmk__xe_set_props(xml,
PCMK_XA_SOURCE, origin->details->uname,
PCMK_XA_DEST, destination->details->uname,
NULL);
} else if (same_host) {
/* Promoting or demoting a promotable clone instance */
pcmk__xe_set_props(xml,
PCMK_XA_ROLE, pcmk_role_text(rsc->role),
PCMK_XA_NEXT_ROLE, pcmk_role_text(rsc->next_role),
PCMK_XA_SOURCE, origin->details->uname,
NULL);
} else {
/* Moving and promoting/demoting */
pcmk__xe_set_props(xml,
PCMK_XA_ROLE, pcmk_role_text(rsc->role),
PCMK_XA_SOURCE, origin->details->uname,
PCMK_XA_NEXT_ROLE, pcmk_role_text(rsc->next_role),
PCMK_XA_DEST, destination->details->uname,
NULL);
}
if ((source->reason != NULL)
&& !pcmk_is_set(action->flags, pcmk_action_runnable)) {
pcmk__xe_set_props(xml,
PCMK_XA_REASON, source->reason,
PCMK_XA_BLOCKED, PCMK_VALUE_TRUE,
NULL);
} else if (source->reason != NULL) {
crm_xml_add(xml, PCMK_XA_REASON, source->reason);
} else if (!pcmk_is_set(action->flags, pcmk_action_runnable)) {
pcmk__xe_set_bool_attr(xml, PCMK_XA_BLOCKED, true);
}
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("rsc-is-colocated-with-list", "pcmk_resource_t *", "bool")
static int
rsc_is_colocated_with_list(pcmk__output_t *out, va_list args) {
pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *);
bool recursive = va_arg(args, int);
int rc = pcmk_rc_no_output;
if (pcmk_is_set(rsc->flags, pcmk_rsc_detect_loop)) {
return rc;
}
/* We're listing constraints explicitly involving rsc, so use rsc->rsc_cons
* directly rather than rsc->cmds->this_with_colocations().
*/
pcmk__set_rsc_flags(rsc, pcmk_rsc_detect_loop);
for (GList *lpc = rsc->rsc_cons; lpc != NULL; lpc = lpc->next) {
pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data;
char *hdr = NULL;
PCMK__OUTPUT_LIST_HEADER(out, false, rc,
"Resources %s is colocated with", rsc->id);
if (pcmk_is_set(cons->primary->flags, pcmk_rsc_detect_loop)) {
out->list_item(out, NULL, "%s (id=%s - loop)",
cons->primary->id, cons->id);
continue;
}
hdr = colocations_header(cons->primary, cons, false);
out->list_item(out, NULL, "%s", hdr);
free(hdr);
// Empty list header for indentation of information about this resource
out->begin_list(out, NULL, NULL, NULL);
out->message(out, "locations-list", cons->primary);
if (recursive) {
out->message(out, "rsc-is-colocated-with-list",
cons->primary, recursive);
}
out->end_list(out);
}
PCMK__OUTPUT_LIST_FOOTER(out, rc);
return rc;
}
PCMK__OUTPUT_ARGS("rsc-is-colocated-with-list", "pcmk_resource_t *", "bool")
static int
rsc_is_colocated_with_list_xml(pcmk__output_t *out, va_list args) {
pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *);
bool recursive = va_arg(args, int);
int rc = pcmk_rc_no_output;
if (pcmk_is_set(rsc->flags, pcmk_rsc_detect_loop)) {
return rc;
}
/* We're listing constraints explicitly involving rsc, so use rsc->rsc_cons
* directly rather than rsc->cmds->this_with_colocations().
*/
pcmk__set_rsc_flags(rsc, pcmk_rsc_detect_loop);
for (GList *lpc = rsc->rsc_cons; lpc != NULL; lpc = lpc->next) {
pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data;
if (pcmk_is_set(cons->primary->flags, pcmk_rsc_detect_loop)) {
colocations_xml_node(out, cons->primary, cons);
continue;
}
colocations_xml_node(out, cons->primary, cons);
do_locations_list_xml(out, cons->primary, false);
if (recursive) {
out->message(out, "rsc-is-colocated-with-list",
cons->primary, recursive);
}
}
return rc;
}
PCMK__OUTPUT_ARGS("rscs-colocated-with-list", "pcmk_resource_t *", "bool")
static int
rscs_colocated_with_list(pcmk__output_t *out, va_list args) {
pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *);
bool recursive = va_arg(args, int);
int rc = pcmk_rc_no_output;
if (pcmk_is_set(rsc->flags, pcmk_rsc_detect_loop)) {
return rc;
}
/* We're listing constraints explicitly involving rsc, so use
* rsc->rsc_cons_lhs directly rather than
* rsc->cmds->with_this_colocations().
*/
pcmk__set_rsc_flags(rsc, pcmk_rsc_detect_loop);
for (GList *lpc = rsc->rsc_cons_lhs; lpc != NULL; lpc = lpc->next) {
pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data;
char *hdr = NULL;
PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Resources colocated with %s",
rsc->id);
if (pcmk_is_set(cons->dependent->flags, pcmk_rsc_detect_loop)) {
out->list_item(out, NULL, "%s (id=%s - loop)",
cons->dependent->id, cons->id);
continue;
}
hdr = colocations_header(cons->dependent, cons, true);
out->list_item(out, NULL, "%s", hdr);
free(hdr);
// Empty list header for indentation of information about this resource
out->begin_list(out, NULL, NULL, NULL);
out->message(out, "locations-list", cons->dependent);
if (recursive) {
out->message(out, "rscs-colocated-with-list",
cons->dependent, recursive);
}
out->end_list(out);
}
PCMK__OUTPUT_LIST_FOOTER(out, rc);
return rc;
}
PCMK__OUTPUT_ARGS("rscs-colocated-with-list", "pcmk_resource_t *", "bool")
static int
rscs_colocated_with_list_xml(pcmk__output_t *out, va_list args) {
pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *);
bool recursive = va_arg(args, int);
int rc = pcmk_rc_no_output;
if (pcmk_is_set(rsc->flags, pcmk_rsc_detect_loop)) {
return rc;
}
/* We're listing constraints explicitly involving rsc, so use
* rsc->rsc_cons_lhs directly rather than
* rsc->cmds->with_this_colocations().
*/
pcmk__set_rsc_flags(rsc, pcmk_rsc_detect_loop);
for (GList *lpc = rsc->rsc_cons_lhs; lpc != NULL; lpc = lpc->next) {
pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data;
if (pcmk_is_set(cons->dependent->flags, pcmk_rsc_detect_loop)) {
colocations_xml_node(out, cons->dependent, cons);
continue;
}
colocations_xml_node(out, cons->dependent, cons);
do_locations_list_xml(out, cons->dependent, false);
if (recursive) {
out->message(out, "rscs-colocated-with-list",
cons->dependent, recursive);
}
}
return rc;
}
PCMK__OUTPUT_ARGS("locations-list", "pcmk_resource_t *")
static int
locations_list(pcmk__output_t *out, va_list args) {
pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *);
GList *lpc = NULL;
GList *list = rsc->rsc_location;
int rc = pcmk_rc_no_output;
for (lpc = list; lpc != NULL; lpc = lpc->next) {
pcmk__location_t *cons = lpc->data;
GList *lpc2 = NULL;
for (lpc2 = cons->nodes; lpc2 != NULL; lpc2 = lpc2->next) {
pcmk_node_t *node = (pcmk_node_t *) lpc2->data;
PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Locations");
out->list_item(out, NULL, "Node %s (score=%s, id=%s, rsc=%s)",
pcmk__node_name(node),
pcmk_readable_score(node->weight), cons->id,
rsc->id);
}
}
PCMK__OUTPUT_LIST_FOOTER(out, rc);
return rc;
}
PCMK__OUTPUT_ARGS("locations-list", "pcmk_resource_t *")
static int
locations_list_xml(pcmk__output_t *out, va_list args) {
pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *);
return do_locations_list_xml(out, rsc, true);
}
PCMK__OUTPUT_ARGS("locations-and-colocations", "pcmk_resource_t *",
"bool", "bool")
static int
locations_and_colocations(pcmk__output_t *out, va_list args)
{
pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *);
bool recursive = va_arg(args, int);
bool force = va_arg(args, int);
pcmk__unpack_constraints(rsc->cluster);
// Constraints apply to group/clone, not member/instance
if (!force) {
rsc = uber_parent(rsc);
}
out->message(out, "locations-list", rsc);
pe__clear_resource_flags_on_all(rsc->cluster, pcmk_rsc_detect_loop);
out->message(out, "rscs-colocated-with-list", rsc, recursive);
pe__clear_resource_flags_on_all(rsc->cluster, pcmk_rsc_detect_loop);
out->message(out, "rsc-is-colocated-with-list", rsc, recursive);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("locations-and-colocations", "pcmk_resource_t *",
"bool", "bool")
static int
locations_and_colocations_xml(pcmk__output_t *out, va_list args)
{
pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *);
bool recursive = va_arg(args, int);
bool force = va_arg(args, int);
pcmk__unpack_constraints(rsc->cluster);
// Constraints apply to group/clone, not member/instance
if (!force) {
rsc = uber_parent(rsc);
}
pcmk__output_xml_create_parent(out, PCMK_XE_CONSTRAINTS, NULL);
do_locations_list_xml(out, rsc, false);
pe__clear_resource_flags_on_all(rsc->cluster, pcmk_rsc_detect_loop);
out->message(out, "rscs-colocated-with-list", rsc, recursive);
pe__clear_resource_flags_on_all(rsc->cluster, pcmk_rsc_detect_loop);
out->message(out, "rsc-is-colocated-with-list", rsc, recursive);
pcmk__output_xml_pop_parent(out);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("health", "const char *", "const char *", "const char *",
"const char *")
static int
health(pcmk__output_t *out, va_list args)
{
const char *sys_from G_GNUC_UNUSED = va_arg(args, const char *);
const char *host_from = va_arg(args, const char *);
const char *fsa_state = va_arg(args, const char *);
const char *result = va_arg(args, const char *);
return out->info(out, "Controller on %s in state %s: %s",
pcmk__s(host_from, "unknown node"),
pcmk__s(fsa_state, "unknown"),
pcmk__s(result, "unknown result"));
}
PCMK__OUTPUT_ARGS("health", "const char *", "const char *", "const char *",
"const char *")
static int
health_text(pcmk__output_t *out, va_list args)
{
if (!out->is_quiet(out)) {
return health(out, args);
} else {
const char *sys_from G_GNUC_UNUSED = va_arg(args, const char *);
const char *host_from G_GNUC_UNUSED = va_arg(args, const char *);
const char *fsa_state = va_arg(args, const char *);
const char *result G_GNUC_UNUSED = va_arg(args, const char *);
if (fsa_state != NULL) {
pcmk__formatted_printf(out, "%s\n", fsa_state);
return pcmk_rc_ok;
}
}
return pcmk_rc_no_output;
}
PCMK__OUTPUT_ARGS("health", "const char *", "const char *", "const char *",
"const char *")
static int
health_xml(pcmk__output_t *out, va_list args)
{
const char *sys_from = va_arg(args, const char *);
const char *host_from = va_arg(args, const char *);
const char *fsa_state = va_arg(args, const char *);
const char *result = va_arg(args, const char *);
pcmk__output_create_xml_node(out, pcmk__s(sys_from, ""),
"node_name", pcmk__s(host_from, ""),
"state", pcmk__s(fsa_state, ""),
PCMK_XA_RESULT, pcmk__s(result, ""),
NULL);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("pacemakerd-health", "const char *",
"enum pcmk_pacemakerd_state", "const char *", "time_t")
static int
pacemakerd_health(pcmk__output_t *out, va_list args)
{
const char *sys_from = va_arg(args, const char *);
enum pcmk_pacemakerd_state state =
(enum pcmk_pacemakerd_state) va_arg(args, int);
const char *state_s = va_arg(args, const char *);
time_t last_updated = va_arg(args, time_t);
char *last_updated_s = NULL;
int rc = pcmk_rc_ok;
if (sys_from == NULL) {
if (state == pcmk_pacemakerd_state_remote) {
sys_from = "pacemaker-remoted";
} else {
sys_from = CRM_SYSTEM_MCP;
}
}
if (state_s == NULL) {
state_s = pcmk__pcmkd_state_enum2friendly(state);
}
if (last_updated != 0) {
last_updated_s = pcmk__epoch2str(&last_updated,
crm_time_log_date
|crm_time_log_timeofday
|crm_time_log_with_timezone);
}
rc = out->info(out, "Status of %s: '%s' (last updated %s)",
sys_from, state_s,
pcmk__s(last_updated_s, "at unknown time"));
free(last_updated_s);
return rc;
}
PCMK__OUTPUT_ARGS("pacemakerd-health", "const char *",
"enum pcmk_pacemakerd_state", "const char *", "time_t")
static int
pacemakerd_health_html(pcmk__output_t *out, va_list args)
{
const char *sys_from = va_arg(args, const char *);
enum pcmk_pacemakerd_state state =
(enum pcmk_pacemakerd_state) va_arg(args, int);
const char *state_s = va_arg(args, const char *);
time_t last_updated = va_arg(args, time_t);
char *last_updated_s = NULL;
char *msg = NULL;
if (sys_from == NULL) {
if (state == pcmk_pacemakerd_state_remote) {
sys_from = "pacemaker-remoted";
} else {
sys_from = CRM_SYSTEM_MCP;
}
}
if (state_s == NULL) {
state_s = pcmk__pcmkd_state_enum2friendly(state);
}
if (last_updated != 0) {
last_updated_s = pcmk__epoch2str(&last_updated,
crm_time_log_date
|crm_time_log_timeofday
|crm_time_log_with_timezone);
}
msg = crm_strdup_printf("Status of %s: '%s' (last updated %s)",
sys_from, state_s,
pcmk__s(last_updated_s, "at unknown time"));
pcmk__output_create_html_node(out, "li", NULL, NULL, msg);
free(msg);
free(last_updated_s);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("pacemakerd-health", "const char *",
"enum pcmk_pacemakerd_state", "const char *", "time_t")
static int
pacemakerd_health_text(pcmk__output_t *out, va_list args)
{
if (!out->is_quiet(out)) {
return pacemakerd_health(out, args);
} else {
const char *sys_from G_GNUC_UNUSED = va_arg(args, const char *);
enum pcmk_pacemakerd_state state =
(enum pcmk_pacemakerd_state) va_arg(args, int);
const char *state_s = va_arg(args, const char *);
time_t last_updated G_GNUC_UNUSED = va_arg(args, time_t);
if (state_s == NULL) {
state_s = pcmk_pacemakerd_api_daemon_state_enum2text(state);
}
pcmk__formatted_printf(out, "%s\n", state_s);
return pcmk_rc_ok;
}
}
PCMK__OUTPUT_ARGS("pacemakerd-health", "const char *",
"enum pcmk_pacemakerd_state", "const char *", "time_t")
static int
pacemakerd_health_xml(pcmk__output_t *out, va_list args)
{
const char *sys_from = va_arg(args, const char *);
enum pcmk_pacemakerd_state state =
(enum pcmk_pacemakerd_state) va_arg(args, int);
const char *state_s = va_arg(args, const char *);
time_t last_updated = va_arg(args, time_t);
char *last_updated_s = NULL;
if (sys_from == NULL) {
if (state == pcmk_pacemakerd_state_remote) {
sys_from = "pacemaker-remoted";
} else {
sys_from = CRM_SYSTEM_MCP;
}
}
if (state_s == NULL) {
state_s = pcmk_pacemakerd_api_daemon_state_enum2text(state);
}
if (last_updated != 0) {
last_updated_s = pcmk__epoch2str(&last_updated,
crm_time_log_date
|crm_time_log_timeofday
|crm_time_log_with_timezone);
}
pcmk__output_create_xml_node(out, "pacemakerd",
"sys_from", sys_from,
"state", state_s,
PCMK_XA_LAST_UPDATED, last_updated_s,
NULL);
free(last_updated_s);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("profile", "const char *", "clock_t", "clock_t")
static int
profile_default(pcmk__output_t *out, va_list args) {
const char *xml_file = va_arg(args, const char *);
clock_t start = va_arg(args, clock_t);
clock_t end = va_arg(args, clock_t);
out->list_item(out, NULL, "Testing %s ... %.2f secs", xml_file,
(end - start) / (float) CLOCKS_PER_SEC);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("profile", "const char *", "clock_t", "clock_t")
static int
profile_xml(pcmk__output_t *out, va_list args) {
const char *xml_file = va_arg(args, const char *);
clock_t start = va_arg(args, clock_t);
clock_t end = va_arg(args, clock_t);
char *duration = pcmk__ftoa((end - start) / (float) CLOCKS_PER_SEC);
pcmk__output_create_xml_node(out, "timing",
PCMK_XA_FILE, xml_file,
PCMK_XA_DURATION, duration,
NULL);
free(duration);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("dc", "const char *")
static int
dc(pcmk__output_t *out, va_list args)
{
const char *dc = va_arg(args, const char *);
return out->info(out, "Designated Controller is: %s",
pcmk__s(dc, "not yet elected"));
}
PCMK__OUTPUT_ARGS("dc", "const char *")
static int
dc_text(pcmk__output_t *out, va_list args)
{
if (!out->is_quiet(out)) {
return dc(out, args);
} else {
const char *dc = va_arg(args, const char *);
if (dc != NULL) {
pcmk__formatted_printf(out, "%s\n", pcmk__s(dc, ""));
return pcmk_rc_ok;
}
}
return pcmk_rc_no_output;
}
PCMK__OUTPUT_ARGS("dc", "const char *")
static int
dc_xml(pcmk__output_t *out, va_list args)
{
const char *dc = va_arg(args, const char *);
pcmk__output_create_xml_node(out, PCMK_XE_DC,
"node_name", pcmk__s(dc, ""),
NULL);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("crmadmin-node", "const char *", "const char *",
"const char *", "bool")
static int
crmadmin_node(pcmk__output_t *out, va_list args)
{
const char *type = va_arg(args, const char *);
const char *name = va_arg(args, const char *);
const char *id = va_arg(args, const char *);
bool bash_export = va_arg(args, int);
if (bash_export) {
return out->info(out, "export %s=%s",
pcmk__s(name, "<null>"), pcmk__s(id, ""));
} else {
return out->info(out, "%s node: %s (%s)", type ? type : "cluster",
pcmk__s(name, "<null>"), pcmk__s(id, "<null>"));
}
}
PCMK__OUTPUT_ARGS("crmadmin-node", "const char *", "const char *",
"const char *", "bool")
static int
crmadmin_node_text(pcmk__output_t *out, va_list args)
{
if (!out->is_quiet(out)) {
return crmadmin_node(out, args);
} else {
const char *type G_GNUC_UNUSED = va_arg(args, const char *);
const char *name = va_arg(args, const char *);
const char *id G_GNUC_UNUSED = va_arg(args, const char *);
bool bash_export G_GNUC_UNUSED = va_arg(args, int);
pcmk__formatted_printf(out, "%s\n", pcmk__s(name, "<null>"));
return pcmk_rc_ok;
}
}
PCMK__OUTPUT_ARGS("crmadmin-node", "const char *", "const char *",
"const char *", "bool")
static int
crmadmin_node_xml(pcmk__output_t *out, va_list args)
{
const char *type = va_arg(args, const char *);
const char *name = va_arg(args, const char *);
const char *id = va_arg(args, const char *);
bool bash_export G_GNUC_UNUSED = va_arg(args, int);
pcmk__output_create_xml_node(out, PCMK_XE_NODE,
PCMK_XA_TYPE, pcmk__s(type, "cluster"),
PCMK_XA_NAME, pcmk__s(name, ""),
PCMK_XA_ID, pcmk__s(id, ""),
NULL);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("digests", "const pcmk_resource_t *", "const pcmk_node_t *",
"const char *", "guint", "const pcmk__op_digest_t *")
static int
digests_text(pcmk__output_t *out, va_list args)
{
const pcmk_resource_t *rsc = va_arg(args, const pcmk_resource_t *);
const pcmk_node_t *node = va_arg(args, const pcmk_node_t *);
const char *task = va_arg(args, const char *);
guint interval_ms = va_arg(args, guint);
const pcmk__op_digest_t *digests = va_arg(args, const pcmk__op_digest_t *);
char *action_desc = NULL;
const char *rsc_desc = "unknown resource";
const char *node_desc = "unknown node";
if (interval_ms != 0) {
action_desc = crm_strdup_printf("%ums-interval %s action", interval_ms,
((task == NULL)? "unknown" : task));
} else if (pcmk__str_eq(task, PCMK_ACTION_MONITOR, pcmk__str_none)) {
action_desc = strdup("probe action");
} else {
action_desc = crm_strdup_printf("%s action",
((task == NULL)? "unknown" : task));
}
if ((rsc != NULL) && (rsc->id != NULL)) {
rsc_desc = rsc->id;
}
if ((node != NULL) && (node->details->uname != NULL)) {
node_desc = node->details->uname;
}
out->begin_list(out, NULL, NULL, "Digests for %s %s on %s",
rsc_desc, action_desc, node_desc);
free(action_desc);
if (digests == NULL) {
out->list_item(out, NULL, "none");
out->end_list(out);
return pcmk_rc_ok;
}
if (digests->digest_all_calc != NULL) {
out->list_item(out, NULL, "%s (all parameters)",
digests->digest_all_calc);
}
if (digests->digest_secure_calc != NULL) {
out->list_item(out, NULL, "%s (non-private parameters)",
digests->digest_secure_calc);
}
if (digests->digest_restart_calc != NULL) {
out->list_item(out, NULL, "%s (non-reloadable parameters)",
digests->digest_restart_calc);
}
out->end_list(out);
return pcmk_rc_ok;
}
static void
add_digest_xml(xmlNode *parent, const char *type, const char *digest,
xmlNode *digest_source)
{
if (digest != NULL) {
xmlNodePtr digest_xml = create_xml_node(parent, PCMK_XE_DIGEST);
crm_xml_add(digest_xml, PCMK_XA_TYPE, pcmk__s(type, "unspecified"));
crm_xml_add(digest_xml, PCMK_XA_HASH, digest);
if (digest_source != NULL) {
add_node_copy(digest_xml, digest_source);
}
}
}
PCMK__OUTPUT_ARGS("digests", "const pcmk_resource_t *", "const pcmk_node_t *",
"const char *", "guint", "const pcmk__op_digest_t *")
static int
digests_xml(pcmk__output_t *out, va_list args)
{
const pcmk_resource_t *rsc = va_arg(args, const pcmk_resource_t *);
const pcmk_node_t *node = va_arg(args, const pcmk_node_t *);
const char *task = va_arg(args, const char *);
guint interval_ms = va_arg(args, guint);
const pcmk__op_digest_t *digests = va_arg(args, const pcmk__op_digest_t *);
char *interval_s = crm_strdup_printf("%ums", interval_ms);
xmlNode *xml = NULL;
xml = pcmk__output_create_xml_node(out, PCMK_XE_DIGESTS,
PCMK_XA_RESOURCE, pcmk__s(rsc->id, ""),
PCMK_XA_NODE,
pcmk__s(node->details->uname, ""),
"task", pcmk__s(task, ""),
PCMK_XA_INTERVAL, interval_s,
NULL);
free(interval_s);
if (digests != NULL) {
add_digest_xml(xml, "all", digests->digest_all_calc,
digests->params_all);
add_digest_xml(xml, "nonprivate", digests->digest_secure_calc,
digests->params_secure);
add_digest_xml(xml, "nonreloadable", digests->digest_restart_calc,
digests->params_restart);
}
return pcmk_rc_ok;
}
#define STOP_SANITY_ASSERT(lineno) do { \
if ((current != NULL) && current->details->unclean) { \
/* It will be a pseudo op */ \
} else if (stop == NULL) { \
crm_err("%s:%d: No stop action exists for %s", \
__func__, lineno, rsc->id); \
CRM_ASSERT(stop != NULL); \
} else if (pcmk_is_set(stop->flags, pcmk_action_optional)) { \
crm_err("%s:%d: Action %s is still optional", \
__func__, lineno, stop->uuid); \
CRM_ASSERT(!pcmk_is_set(stop->flags, pcmk_action_optional));\
} \
} while (0)
PCMK__OUTPUT_ARGS("rsc-action", "pcmk_resource_t *", "pcmk_node_t *",
"pcmk_node_t *")
static int
rsc_action_default(pcmk__output_t *out, va_list args)
{
pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *);
pcmk_node_t *current = va_arg(args, pcmk_node_t *);
pcmk_node_t *next = va_arg(args, pcmk_node_t *);
GList *possible_matches = NULL;
char *key = NULL;
int rc = pcmk_rc_no_output;
bool moving = false;
pcmk_node_t *start_node = NULL;
pcmk_action_t *start = NULL;
pcmk_action_t *stop = NULL;
pcmk_action_t *promote = NULL;
pcmk_action_t *demote = NULL;
pcmk_action_t *reason_op = NULL;
if (!pcmk_is_set(rsc->flags, pcmk_rsc_managed)
|| (current == NULL && next == NULL)) {
const bool managed = pcmk_is_set(rsc->flags, pcmk_rsc_managed);
pcmk__rsc_info(rsc, "Leave %s\t(%s%s)",
rsc->id, pcmk_role_text(rsc->role),
(managed? "" : " unmanaged"));
return rc;
}
moving = (current != NULL) && (next != NULL)
&& !pcmk__same_node(current, next);
possible_matches = pe__resource_actions(rsc, next, PCMK_ACTION_START,
false);
if (possible_matches) {
start = possible_matches->data;
g_list_free(possible_matches);
}
if ((start == NULL)
|| !pcmk_is_set(start->flags, pcmk_action_runnable)) {
start_node = NULL;
} else {
start_node = current;
}
possible_matches = pe__resource_actions(rsc, start_node, PCMK_ACTION_STOP,
false);
if (possible_matches) {
stop = possible_matches->data;
g_list_free(possible_matches);
} else if (pcmk_is_set(rsc->flags, pcmk_rsc_stop_unexpected)) {
/* The resource is multiply active with PCMK_META_MULTIPLE_ACTIVE set to
* stop_unexpected, and not stopping on its current node, but it should
* be stopping elsewhere.
*/
possible_matches = pe__resource_actions(rsc, NULL, PCMK_ACTION_STOP,
false);
if (possible_matches != NULL) {
stop = possible_matches->data;
g_list_free(possible_matches);
}
}
possible_matches = pe__resource_actions(rsc, next, PCMK_ACTION_PROMOTE,
false);
if (possible_matches) {
promote = possible_matches->data;
g_list_free(possible_matches);
}
possible_matches = pe__resource_actions(rsc, next, PCMK_ACTION_DEMOTE,
false);
if (possible_matches) {
demote = possible_matches->data;
g_list_free(possible_matches);
}
if (rsc->role == rsc->next_role) {
pcmk_action_t *migrate_op = NULL;
CRM_CHECK(next != NULL, return rc);
possible_matches = pe__resource_actions(rsc, next,
PCMK_ACTION_MIGRATE_FROM,
false);
if (possible_matches) {
migrate_op = possible_matches->data;
}
if ((migrate_op != NULL) && (current != NULL)
&& pcmk_is_set(migrate_op->flags, pcmk_action_runnable)) {
rc = out->message(out, "rsc-action-item", "Migrate", rsc, current,
next, start, NULL);
} else if (pcmk_is_set(rsc->flags, pcmk_rsc_reload)) {
rc = out->message(out, "rsc-action-item", "Reload", rsc, current,
next, start, NULL);
} else if ((start == NULL)
|| pcmk_is_set(start->flags, pcmk_action_optional)) {
if ((demote != NULL) && (promote != NULL)
&& !pcmk_is_set(demote->flags, pcmk_action_optional)
&& !pcmk_is_set(promote->flags, pcmk_action_optional)) {
rc = out->message(out, "rsc-action-item", "Re-promote", rsc,
current, next, promote, demote);
} else {
pcmk__rsc_info(rsc, "Leave %s\t(%s %s)", rsc->id,
pcmk_role_text(rsc->role),
pcmk__node_name(next));
}
} else if (!pcmk_is_set(start->flags, pcmk_action_runnable)) {
if ((stop == NULL) || (stop->reason == NULL)) {
reason_op = start;
} else {
reason_op = stop;
}
rc = out->message(out, "rsc-action-item", "Stop", rsc, current,
NULL, stop, reason_op);
STOP_SANITY_ASSERT(__LINE__);
} else if (moving && current) {
const bool failed = pcmk_is_set(rsc->flags, pcmk_rsc_failed);
rc = out->message(out, "rsc-action-item",
(failed? "Recover" : "Move"), rsc, current, next,
stop, NULL);
} else if (pcmk_is_set(rsc->flags, pcmk_rsc_failed)) {
rc = out->message(out, "rsc-action-item", "Recover", rsc, current,
NULL, stop, NULL);
STOP_SANITY_ASSERT(__LINE__);
} else {
rc = out->message(out, "rsc-action-item", "Restart", rsc, current,
next, start, NULL);
#if 0
/* @TODO This can be reached in situations that should really be
* "Start" (see for example the migrate-fail-7 regression test)
*/
STOP_SANITY_ASSERT(__LINE__);
#endif
}
g_list_free(possible_matches);
return rc;
}
if ((stop != NULL)
&& ((rsc->next_role == pcmk_role_stopped)
|| ((start != NULL)
&& !pcmk_is_set(start->flags, pcmk_action_runnable)))) {
key = stop_key(rsc);
for (GList *iter = rsc->running_on; iter != NULL; iter = iter->next) {
pcmk_node_t *node = iter->data;
pcmk_action_t *stop_op = NULL;
reason_op = start;
possible_matches = find_actions(rsc->actions, key, node);
if (possible_matches) {
stop_op = possible_matches->data;
g_list_free(possible_matches);
}
if (stop_op != NULL) {
if (pcmk_is_set(stop_op->flags, pcmk_action_runnable)) {
STOP_SANITY_ASSERT(__LINE__);
}
if (stop_op->reason != NULL) {
reason_op = stop_op;
}
}
if (out->message(out, "rsc-action-item", "Stop", rsc, node, NULL,
stop_op, reason_op) == pcmk_rc_ok) {
rc = pcmk_rc_ok;
}
}
free(key);
} else if ((stop != NULL)
&& pcmk_all_flags_set(rsc->flags,
pcmk_rsc_failed|pcmk_rsc_stop_if_failed)) {
/* 'stop' may be NULL if the failure was ignored */
rc = out->message(out, "rsc-action-item", "Recover", rsc, current,
next, stop, start);
STOP_SANITY_ASSERT(__LINE__);
} else if (moving) {
rc = out->message(out, "rsc-action-item", "Move", rsc, current, next,
stop, NULL);
STOP_SANITY_ASSERT(__LINE__);
} else if (pcmk_is_set(rsc->flags, pcmk_rsc_reload)) {
rc = out->message(out, "rsc-action-item", "Reload", rsc, current, next,
start, NULL);
} else if ((stop != NULL)
&& !pcmk_is_set(stop->flags, pcmk_action_optional)) {
rc = out->message(out, "rsc-action-item", "Restart", rsc, current,
next, start, NULL);
STOP_SANITY_ASSERT(__LINE__);
} else if (rsc->role == pcmk_role_promoted) {
CRM_LOG_ASSERT(current != NULL);
rc = out->message(out, "rsc-action-item", "Demote", rsc, current,
next, demote, NULL);
} else if (rsc->next_role == pcmk_role_promoted) {
CRM_LOG_ASSERT(next);
rc = out->message(out, "rsc-action-item", "Promote", rsc, current,
next, promote, NULL);
} else if ((rsc->role == pcmk_role_stopped)
&& (rsc->next_role > pcmk_role_stopped)) {
rc = out->message(out, "rsc-action-item", "Start", rsc, current, next,
start, NULL);
}
return rc;
}
PCMK__OUTPUT_ARGS("node-action", "const char *", "const char *", "const char *")
static int
node_action(pcmk__output_t *out, va_list args)
{
const char *task = va_arg(args, const char *);
const char *node_name = va_arg(args, const char *);
const char *reason = va_arg(args, const char *);
if (task == NULL) {
return pcmk_rc_no_output;
} else if (reason) {
out->list_item(out, NULL, "%s %s '%s'", task, node_name, reason);
} else {
crm_notice(" * %s %s", task, node_name);
}
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("node-action", "const char *", "const char *", "const char *")
static int
node_action_xml(pcmk__output_t *out, va_list args)
{
const char *task = va_arg(args, const char *);
const char *node_name = va_arg(args, const char *);
const char *reason = va_arg(args, const char *);
if (task == NULL) {
return pcmk_rc_no_output;
} else if (reason) {
pcmk__output_create_xml_node(out, PCMK_XE_NODE_ACTION,
"task", task,
PCMK_XA_NODE, node_name,
PCMK_XA_REASON, reason,
NULL);
} else {
crm_notice(" * %s %s", task, node_name);
}
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("node-info", "uint32_t", "const char *", "const char *",
"const char *", "bool", "bool")
static int
node_info_default(pcmk__output_t *out, va_list args)
{
uint32_t node_id = va_arg(args, uint32_t);
const char *node_name = va_arg(args, const char *);
const char *uuid = va_arg(args, const char *);
const char *state = va_arg(args, const char *);
bool have_quorum = (bool) va_arg(args, int);
bool is_remote = (bool) va_arg(args, int);
return out->info(out,
"Node %" PRIu32 ": %s "
"(uuid=%s, state=%s, have_quorum=%s, is_remote=%s)",
node_id, pcmk__s(node_name, "unknown"),
pcmk__s(uuid, "unknown"), pcmk__s(state, "unknown"),
pcmk__btoa(have_quorum), pcmk__btoa(is_remote));
}
PCMK__OUTPUT_ARGS("node-info", "uint32_t", "const char *", "const char *",
"const char *", "bool", "bool")
static int
node_info_xml(pcmk__output_t *out, va_list args)
{
uint32_t node_id = va_arg(args, uint32_t);
const char *node_name = va_arg(args, const char *);
const char *uuid = va_arg(args, const char *);
const char *state = va_arg(args, const char *);
bool have_quorum = (bool) va_arg(args, int);
bool is_remote = (bool) va_arg(args, int);
char *id_s = crm_strdup_printf("%" PRIu32, node_id);
pcmk__output_create_xml_node(out, PCMK_XE_NODE_INFO,
PCMK_XA_NODEID, id_s,
PCMK_XA_UNAME, node_name,
PCMK_XA_ID, uuid,
- PCMK__XA_CRMD, state,
+ PCMK_XA_CRMD, state,
PCMK_XA_HAVE_QUORUM, pcmk__btoa(have_quorum),
PCMK__XA_REMOTE_NODE, pcmk__btoa(is_remote),
NULL);
free(id_s);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("inject-cluster-action", "const char *", "const char *",
"xmlNodePtr")
static int
inject_cluster_action(pcmk__output_t *out, va_list args)
{
const char *node = va_arg(args, const char *);
const char *task = va_arg(args, const char *);
xmlNodePtr rsc = va_arg(args, xmlNodePtr);
if (out->is_quiet(out)) {
return pcmk_rc_no_output;
}
if (rsc != NULL) {
out->list_item(out, NULL, "Cluster action: %s for %s on %s",
task, ID(rsc), node);
} else {
out->list_item(out, NULL, "Cluster action: %s on %s", task, node);
}
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("inject-cluster-action", "const char *", "const char *",
"xmlNodePtr")
static int
inject_cluster_action_xml(pcmk__output_t *out, va_list args)
{
const char *node = va_arg(args, const char *);
const char *task = va_arg(args, const char *);
xmlNodePtr rsc = va_arg(args, xmlNodePtr);
xmlNodePtr xml_node = NULL;
if (out->is_quiet(out)) {
return pcmk_rc_no_output;
}
xml_node = pcmk__output_create_xml_node(out, PCMK_XE_CLUSTER_ACTION,
"task", task,
PCMK_XA_NODE, node,
NULL);
if (rsc) {
crm_xml_add(xml_node, PCMK_XA_ID, ID(rsc));
}
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("inject-fencing-action", "const char *", "const char *")
static int
inject_fencing_action(pcmk__output_t *out, va_list args)
{
const char *target = va_arg(args, const char *);
const char *op = va_arg(args, const char *);
if (out->is_quiet(out)) {
return pcmk_rc_no_output;
}
out->list_item(out, NULL, "Fencing %s (%s)", target, op);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("inject-fencing-action", "const char *", "const char *")
static int
inject_fencing_action_xml(pcmk__output_t *out, va_list args)
{
const char *target = va_arg(args, const char *);
const char *op = va_arg(args, const char *);
if (out->is_quiet(out)) {
return pcmk_rc_no_output;
}
pcmk__output_create_xml_node(out, PCMK_XE_FENCING_ACTION,
PCMK_XA_TARGET, target,
PCMK_XA_OP, op,
NULL);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("inject-attr", "const char *", "const char *", "xmlNodePtr")
static int
inject_attr(pcmk__output_t *out, va_list args)
{
const char *name = va_arg(args, const char *);
const char *value = va_arg(args, const char *);
xmlNodePtr cib_node = va_arg(args, xmlNodePtr);
xmlChar *node_path = NULL;
if (out->is_quiet(out)) {
return pcmk_rc_no_output;
}
node_path = xmlGetNodePath(cib_node);
out->list_item(out, NULL, "Injecting attribute %s=%s into %s '%s'",
name, value, node_path, ID(cib_node));
free(node_path);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("inject-attr", "const char *", "const char *", "xmlNodePtr")
static int
inject_attr_xml(pcmk__output_t *out, va_list args)
{
const char *name = va_arg(args, const char *);
const char *value = va_arg(args, const char *);
xmlNodePtr cib_node = va_arg(args, xmlNodePtr);
xmlChar *node_path = NULL;
if (out->is_quiet(out)) {
return pcmk_rc_no_output;
}
node_path = xmlGetNodePath(cib_node);
pcmk__output_create_xml_node(out, PCMK_XE_INJECT_ATTR,
PCMK_XA_NAME, name,
PCMK_XA_VALUE, value,
"node_path", node_path,
PCMK_XA_CIB_NODE, ID(cib_node),
NULL);
free(node_path);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("inject-spec", "const char *")
static int
inject_spec(pcmk__output_t *out, va_list args)
{
const char *spec = va_arg(args, const char *);
if (out->is_quiet(out)) {
return pcmk_rc_no_output;
}
out->list_item(out, NULL, "Injecting %s into the configuration", spec);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("inject-spec", "const char *")
static int
inject_spec_xml(pcmk__output_t *out, va_list args)
{
const char *spec = va_arg(args, const char *);
if (out->is_quiet(out)) {
return pcmk_rc_no_output;
}
pcmk__output_create_xml_node(out, PCMK_XE_INJECT_SPEC,
"spec", spec,
NULL);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("inject-modify-config", "const char *", "const char *")
static int
inject_modify_config(pcmk__output_t *out, va_list args)
{
const char *quorum = va_arg(args, const char *);
const char *watchdog = va_arg(args, const char *);
if (out->is_quiet(out)) {
return pcmk_rc_no_output;
}
out->begin_list(out, NULL, NULL, "Performing Requested Modifications");
if (quorum) {
out->list_item(out, NULL, "Setting quorum: %s", quorum);
}
if (watchdog) {
out->list_item(out, NULL, "Setting watchdog: %s", watchdog);
}
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("inject-modify-config", "const char *", "const char *")
static int
inject_modify_config_xml(pcmk__output_t *out, va_list args)
{
const char *quorum = va_arg(args, const char *);
const char *watchdog = va_arg(args, const char *);
xmlNodePtr node = NULL;
if (out->is_quiet(out)) {
return pcmk_rc_no_output;
}
node = pcmk__output_xml_create_parent(out, PCMK_XE_MODIFICATIONS, NULL);
if (quorum) {
crm_xml_add(node, "quorum", quorum);
}
if (watchdog) {
crm_xml_add(node, "watchdog", watchdog);
}
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("inject-modify-node", "const char *", "const char *")
static int
inject_modify_node(pcmk__output_t *out, va_list args)
{
const char *action = va_arg(args, const char *);
const char *node = va_arg(args, const char *);
if (out->is_quiet(out)) {
return pcmk_rc_no_output;
}
if (pcmk__str_eq(action, "Online", pcmk__str_none)) {
out->list_item(out, NULL, "Bringing node %s online", node);
return pcmk_rc_ok;
} else if (pcmk__str_eq(action, "Offline", pcmk__str_none)) {
out->list_item(out, NULL, "Taking node %s offline", node);
return pcmk_rc_ok;
} else if (pcmk__str_eq(action, "Failing", pcmk__str_none)) {
out->list_item(out, NULL, "Failing node %s", node);
return pcmk_rc_ok;
}
return pcmk_rc_no_output;
}
PCMK__OUTPUT_ARGS("inject-modify-node", "const char *", "const char *")
static int
inject_modify_node_xml(pcmk__output_t *out, va_list args)
{
const char *action = va_arg(args, const char *);
const char *node = va_arg(args, const char *);
if (out->is_quiet(out)) {
return pcmk_rc_no_output;
}
pcmk__output_create_xml_node(out, PCMK_XE_MODIFY_NODE,
PCMK_XA_ACTION, action,
PCMK_XA_NODE, node,
NULL);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("inject-modify-ticket", "const char *", "const char *")
static int
inject_modify_ticket(pcmk__output_t *out, va_list args)
{
const char *action = va_arg(args, const char *);
const char *ticket = va_arg(args, const char *);
if (out->is_quiet(out)) {
return pcmk_rc_no_output;
}
if (pcmk__str_eq(action, "Standby", pcmk__str_none)) {
out->list_item(out, NULL, "Making ticket %s standby", ticket);
} else {
out->list_item(out, NULL, "%s ticket %s", action, ticket);
}
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("inject-modify-ticket", "const char *", "const char *")
static int
inject_modify_ticket_xml(pcmk__output_t *out, va_list args)
{
const char *action = va_arg(args, const char *);
const char *ticket = va_arg(args, const char *);
if (out->is_quiet(out)) {
return pcmk_rc_no_output;
}
pcmk__output_create_xml_node(out, PCMK_XE_MODIFY_TICKET,
PCMK_XA_ACTION, action,
PCMK_XA_TICKET, ticket,
NULL);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("inject-pseudo-action", "const char *", "const char *")
static int
inject_pseudo_action(pcmk__output_t *out, va_list args)
{
const char *node = va_arg(args, const char *);
const char *task = va_arg(args, const char *);
if (out->is_quiet(out)) {
return pcmk_rc_no_output;
}
out->list_item(out, NULL, "Pseudo action: %s%s%s",
task, ((node == NULL)? "" : " on "), pcmk__s(node, ""));
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("inject-pseudo-action", "const char *", "const char *")
static int
inject_pseudo_action_xml(pcmk__output_t *out, va_list args)
{
const char *node = va_arg(args, const char *);
const char *task = va_arg(args, const char *);
xmlNodePtr xml_node = NULL;
if (out->is_quiet(out)) {
return pcmk_rc_no_output;
}
xml_node = pcmk__output_create_xml_node(out, "pseudo_action",
"task", task,
NULL);
if (node) {
crm_xml_add(xml_node, PCMK_XA_NODE, node);
}
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("inject-rsc-action", "const char *", "const char *",
"const char *", "guint")
static int
inject_rsc_action(pcmk__output_t *out, va_list args)
{
const char *rsc = va_arg(args, const char *);
const char *operation = va_arg(args, const char *);
const char *node = va_arg(args, const char *);
guint interval_ms = va_arg(args, guint);
if (out->is_quiet(out)) {
return pcmk_rc_no_output;
}
if (interval_ms) {
out->list_item(out, NULL, "Resource action: %-15s %s=%u on %s",
rsc, operation, interval_ms, node);
} else {
out->list_item(out, NULL, "Resource action: %-15s %s on %s",
rsc, operation, node);
}
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("inject-rsc-action", "const char *", "const char *",
"const char *", "guint")
static int
inject_rsc_action_xml(pcmk__output_t *out, va_list args)
{
const char *rsc = va_arg(args, const char *);
const char *operation = va_arg(args, const char *);
const char *node = va_arg(args, const char *);
guint interval_ms = va_arg(args, guint);
xmlNodePtr xml_node = NULL;
if (out->is_quiet(out)) {
return pcmk_rc_no_output;
}
xml_node = pcmk__output_create_xml_node(out, PCMK_XE_RSC_ACTION,
PCMK_XA_RESOURCE, rsc,
PCMK_XA_OP, operation,
PCMK_XA_NODE, node,
NULL);
if (interval_ms) {
char *interval_s = pcmk__itoa(interval_ms);
crm_xml_add(xml_node, PCMK_XA_INTERVAL, interval_s);
free(interval_s);
}
return pcmk_rc_ok;
}
#define CHECK_RC(retcode, retval) \
if (retval == pcmk_rc_ok) { \
retcode = pcmk_rc_ok; \
}
PCMK__OUTPUT_ARGS("cluster-status", "pcmk_scheduler_t *",
"enum pcmk_pacemakerd_state", "crm_exit_t",
"stonith_history_t *", "enum pcmk__fence_history", "uint32_t",
"uint32_t", "const char *", "GList *", "GList *")
int
pcmk__cluster_status_text(pcmk__output_t *out, va_list args)
{
pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *);
enum pcmk_pacemakerd_state pcmkd_state =
(enum pcmk_pacemakerd_state) va_arg(args, int);
crm_exit_t history_rc = va_arg(args, crm_exit_t);
stonith_history_t *stonith_history = va_arg(args, stonith_history_t *);
enum pcmk__fence_history fence_history = va_arg(args, int);
uint32_t section_opts = va_arg(args, uint32_t);
uint32_t show_opts = va_arg(args, uint32_t);
const char *prefix = va_arg(args, const char *);
GList *unames = va_arg(args, GList *);
GList *resources = va_arg(args, GList *);
int rc = pcmk_rc_no_output;
bool already_printed_failure = false;
CHECK_RC(rc, out->message(out, "cluster-summary", scheduler, pcmkd_state,
section_opts, show_opts));
if (pcmk_is_set(section_opts, pcmk_section_nodes) && unames) {
CHECK_RC(rc, out->message(out, "node-list", scheduler->nodes, unames,
resources, show_opts, rc == pcmk_rc_ok));
}
/* Print resources section, if needed */
if (pcmk_is_set(section_opts, pcmk_section_resources)) {
CHECK_RC(rc, out->message(out, "resource-list", scheduler, show_opts,
true, unames, resources, rc == pcmk_rc_ok));
}
/* print Node Attributes section if requested */
if (pcmk_is_set(section_opts, pcmk_section_attributes)) {
CHECK_RC(rc, out->message(out, "node-attribute-list", scheduler,
show_opts, (rc == pcmk_rc_ok), unames,
resources));
}
/* If requested, print resource operations (which includes failcounts)
* or just failcounts
*/
if (pcmk_any_flags_set(section_opts,
pcmk_section_operations|pcmk_section_failcounts)) {
CHECK_RC(rc, out->message(out, "node-summary", scheduler, unames,
resources, section_opts, show_opts,
(rc == pcmk_rc_ok)));
}
/* If there were any failed actions, print them */
if (pcmk_is_set(section_opts, pcmk_section_failures)
&& (scheduler->failed != NULL)
&& (scheduler->failed->children != NULL)) {
CHECK_RC(rc, out->message(out, "failed-action-list", scheduler, unames,
resources, show_opts, rc == pcmk_rc_ok));
}
/* Print failed stonith actions */
if (pcmk_is_set(section_opts, pcmk_section_fence_failed) &&
fence_history != pcmk__fence_history_none) {
if (history_rc == 0) {
stonith_history_t *hp = NULL;
hp = stonith__first_matching_event(stonith_history,
stonith__event_state_eq,
GINT_TO_POINTER(st_failed));
if (hp) {
CHECK_RC(rc, out->message(out, "failed-fencing-list",
stonith_history, unames, section_opts,
show_opts, rc == pcmk_rc_ok));
}
} else {
PCMK__OUTPUT_SPACER_IF(out, rc == pcmk_rc_ok);
out->begin_list(out, NULL, NULL, "Failed Fencing Actions");
out->list_item(out, NULL, "Failed to get fencing history: %s",
crm_exit_str(history_rc));
out->end_list(out);
already_printed_failure = true;
}
}
/* Print tickets if requested */
if (pcmk_is_set(section_opts, pcmk_section_tickets)) {
CHECK_RC(rc, out->message(out, "ticket-list", scheduler,
(rc == pcmk_rc_ok)));
}
/* Print negative location constraints if requested */
if (pcmk_is_set(section_opts, pcmk_section_bans)) {
CHECK_RC(rc, out->message(out, "ban-list", scheduler, prefix, resources,
show_opts, rc == pcmk_rc_ok));
}
/* Print stonith history */
if (pcmk_any_flags_set(section_opts, pcmk_section_fencing_all) &&
fence_history != pcmk__fence_history_none) {
if (history_rc != 0) {
if (!already_printed_failure) {
PCMK__OUTPUT_SPACER_IF(out, rc == pcmk_rc_ok);
out->begin_list(out, NULL, NULL, "Failed Fencing Actions");
out->list_item(out, NULL, "Failed to get fencing history: %s",
crm_exit_str(history_rc));
out->end_list(out);
}
} else if (pcmk_is_set(section_opts, pcmk_section_fence_worked)) {
stonith_history_t *hp = NULL;
hp = stonith__first_matching_event(stonith_history,
stonith__event_state_neq,
GINT_TO_POINTER(st_failed));
if (hp) {
CHECK_RC(rc, out->message(out, "fencing-list", hp, unames,
section_opts, show_opts,
rc == pcmk_rc_ok));
}
} else if (pcmk_is_set(section_opts, pcmk_section_fence_pending)) {
stonith_history_t *hp = NULL;
hp = stonith__first_matching_event(stonith_history,
stonith__event_state_pending,
NULL);
if (hp) {
CHECK_RC(rc, out->message(out, "pending-fencing-list", hp,
unames, section_opts, show_opts,
rc == pcmk_rc_ok));
}
}
}
return rc;
}
PCMK__OUTPUT_ARGS("cluster-status", "pcmk_scheduler_t *",
"enum pcmk_pacemakerd_state", "crm_exit_t",
"stonith_history_t *", "enum pcmk__fence_history", "uint32_t",
"uint32_t", "const char *", "GList *", "GList *")
static int
cluster_status_xml(pcmk__output_t *out, va_list args)
{
pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *);
enum pcmk_pacemakerd_state pcmkd_state =
(enum pcmk_pacemakerd_state) va_arg(args, int);
crm_exit_t history_rc = va_arg(args, crm_exit_t);
stonith_history_t *stonith_history = va_arg(args, stonith_history_t *);
enum pcmk__fence_history fence_history = va_arg(args, int);
uint32_t section_opts = va_arg(args, uint32_t);
uint32_t show_opts = va_arg(args, uint32_t);
const char *prefix = va_arg(args, const char *);
GList *unames = va_arg(args, GList *);
GList *resources = va_arg(args, GList *);
out->message(out, "cluster-summary", scheduler, pcmkd_state, section_opts,
show_opts);
/*** NODES ***/
if (pcmk_is_set(section_opts, pcmk_section_nodes)) {
out->message(out, "node-list", scheduler->nodes, unames, resources,
show_opts, false);
}
/* Print resources section, if needed */
if (pcmk_is_set(section_opts, pcmk_section_resources)) {
/* XML output always displays full details. */
uint32_t full_show_opts = show_opts & ~pcmk_show_brief;
out->message(out, "resource-list", scheduler, full_show_opts,
false, unames, resources, false);
}
/* print Node Attributes section if requested */
if (pcmk_is_set(section_opts, pcmk_section_attributes)) {
out->message(out, "node-attribute-list", scheduler, show_opts, false,
unames, resources);
}
/* If requested, print resource operations (which includes failcounts)
* or just failcounts
*/
if (pcmk_any_flags_set(section_opts,
pcmk_section_operations|pcmk_section_failcounts)) {
out->message(out, "node-summary", scheduler, unames,
resources, section_opts, show_opts, false);
}
/* If there were any failed actions, print them */
if (pcmk_is_set(section_opts, pcmk_section_failures)
&& (scheduler->failed != NULL)
&& (scheduler->failed->children != NULL)) {
out->message(out, "failed-action-list", scheduler, unames, resources,
show_opts, false);
}
/* Print stonith history */
if (pcmk_is_set(section_opts, pcmk_section_fencing_all) &&
fence_history != pcmk__fence_history_none) {
out->message(out, "full-fencing-list", history_rc, stonith_history,
unames, section_opts, show_opts, false);
}
/* Print tickets if requested */
if (pcmk_is_set(section_opts, pcmk_section_tickets)) {
out->message(out, "ticket-list", scheduler, false);
}
/* Print negative location constraints if requested */
if (pcmk_is_set(section_opts, pcmk_section_bans)) {
out->message(out, "ban-list", scheduler, prefix, resources, show_opts,
false);
}
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("cluster-status", "pcmk_scheduler_t *",
"enum pcmk_pacemakerd_state", "crm_exit_t",
"stonith_history_t *", "enum pcmk__fence_history", "uint32_t",
"uint32_t", "const char *", "GList *", "GList *")
static int
cluster_status_html(pcmk__output_t *out, va_list args)
{
pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *);
enum pcmk_pacemakerd_state pcmkd_state =
(enum pcmk_pacemakerd_state) va_arg(args, int);
crm_exit_t history_rc = va_arg(args, crm_exit_t);
stonith_history_t *stonith_history = va_arg(args, stonith_history_t *);
enum pcmk__fence_history fence_history = va_arg(args, int);
uint32_t section_opts = va_arg(args, uint32_t);
uint32_t show_opts = va_arg(args, uint32_t);
const char *prefix = va_arg(args, const char *);
GList *unames = va_arg(args, GList *);
GList *resources = va_arg(args, GList *);
bool already_printed_failure = false;
out->message(out, "cluster-summary", scheduler, pcmkd_state, section_opts,
show_opts);
/*** NODE LIST ***/
if (pcmk_is_set(section_opts, pcmk_section_nodes) && unames) {
out->message(out, "node-list", scheduler->nodes, unames, resources,
show_opts, false);
}
/* Print resources section, if needed */
if (pcmk_is_set(section_opts, pcmk_section_resources)) {
out->message(out, "resource-list", scheduler, show_opts, true, unames,
resources, false);
}
/* print Node Attributes section if requested */
if (pcmk_is_set(section_opts, pcmk_section_attributes)) {
out->message(out, "node-attribute-list", scheduler, show_opts, false,
unames, resources);
}
/* If requested, print resource operations (which includes failcounts)
* or just failcounts
*/
if (pcmk_any_flags_set(section_opts,
pcmk_section_operations|pcmk_section_failcounts)) {
out->message(out, "node-summary", scheduler, unames,
resources, section_opts, show_opts, false);
}
/* If there were any failed actions, print them */
if (pcmk_is_set(section_opts, pcmk_section_failures)
&& (scheduler->failed != NULL)
&& (scheduler->failed->children != NULL)) {
out->message(out, "failed-action-list", scheduler, unames, resources,
show_opts, false);
}
/* Print failed stonith actions */
if (pcmk_is_set(section_opts, pcmk_section_fence_failed) &&
fence_history != pcmk__fence_history_none) {
if (history_rc == 0) {
stonith_history_t *hp = NULL;
hp = stonith__first_matching_event(stonith_history,
stonith__event_state_eq,
GINT_TO_POINTER(st_failed));
if (hp) {
out->message(out, "failed-fencing-list", stonith_history,
unames, section_opts, show_opts, false);
}
} else {
out->begin_list(out, NULL, NULL, "Failed Fencing Actions");
out->list_item(out, NULL, "Failed to get fencing history: %s",
crm_exit_str(history_rc));
out->end_list(out);
}
}
/* Print stonith history */
if (pcmk_any_flags_set(section_opts, pcmk_section_fencing_all) &&
fence_history != pcmk__fence_history_none) {
if (history_rc != 0) {
if (!already_printed_failure) {
out->begin_list(out, NULL, NULL, "Failed Fencing Actions");
out->list_item(out, NULL, "Failed to get fencing history: %s",
crm_exit_str(history_rc));
out->end_list(out);
}
} else if (pcmk_is_set(section_opts, pcmk_section_fence_worked)) {
stonith_history_t *hp = NULL;
hp = stonith__first_matching_event(stonith_history,
stonith__event_state_neq,
GINT_TO_POINTER(st_failed));
if (hp) {
out->message(out, "fencing-list", hp, unames, section_opts,
show_opts, false);
}
} else if (pcmk_is_set(section_opts, pcmk_section_fence_pending)) {
stonith_history_t *hp = NULL;
hp = stonith__first_matching_event(stonith_history,
stonith__event_state_pending,
NULL);
if (hp) {
out->message(out, "pending-fencing-list", hp, unames,
section_opts, show_opts, false);
}
}
}
/* Print tickets if requested */
if (pcmk_is_set(section_opts, pcmk_section_tickets)) {
out->message(out, "ticket-list", scheduler, false);
}
/* Print negative location constraints if requested */
if (pcmk_is_set(section_opts, pcmk_section_bans)) {
out->message(out, "ban-list", scheduler, prefix, resources, show_opts,
false);
}
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("attribute", "const char *", "const char *", "const char *",
"const char *", "const char *")
static int
attribute_default(pcmk__output_t *out, va_list args)
{
const char *scope = va_arg(args, const char *);
const char *instance = va_arg(args, const char *);
const char *name = va_arg(args, const char *);
const char *value = va_arg(args, const char *);
const char *host = va_arg(args, const char *);
GString *s = g_string_sized_new(50);
if (!pcmk__str_empty(scope)) {
pcmk__g_strcat(s, PCMK_XA_SCOPE "=\"", scope, "\" ", NULL);
}
if (!pcmk__str_empty(instance)) {
pcmk__g_strcat(s, PCMK_XA_ID "=\"", instance, "\" ", NULL);
}
pcmk__g_strcat(s, PCMK_XA_NAME "=\"", pcmk__s(name, ""), "\" ", NULL);
if (!pcmk__str_empty(host)) {
pcmk__g_strcat(s, PCMK_XA_HOST "=\"", host, "\" ", NULL);
}
pcmk__g_strcat(s, PCMK_XA_VALUE "=\"", pcmk__s(value, ""), "\"", NULL);
out->info(out, "%s", s->str);
g_string_free(s, TRUE);
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("attribute", "const char *", "const char *", "const char *",
"const char *", "const char *")
static int
attribute_xml(pcmk__output_t *out, va_list args)
{
const char *scope = va_arg(args, const char *);
const char *instance = va_arg(args, const char *);
const char *name = va_arg(args, const char *);
const char *value = va_arg(args, const char *);
const char *host = va_arg(args, const char *);
xmlNodePtr node = NULL;
node = pcmk__output_create_xml_node(out, PCMK_XE_ATTRIBUTE,
PCMK_XA_NAME, name,
PCMK_XA_VALUE, pcmk__s(value, ""),
NULL);
if (!pcmk__str_empty(scope)) {
crm_xml_add(node, PCMK_XA_SCOPE, scope);
}
if (!pcmk__str_empty(instance)) {
crm_xml_add(node, PCMK_XA_ID, instance);
}
if (!pcmk__str_empty(host)) {
crm_xml_add(node, PCMK_XA_HOST, host);
}
return pcmk_rc_ok;
}
PCMK__OUTPUT_ARGS("rule-check", "const char *", "int", "const char *")
static int
rule_check_default(pcmk__output_t *out, va_list args)
{
const char *rule_id = va_arg(args, const char *);
int result = va_arg(args, int);
const char *error = va_arg(args, const char *);
switch (result) {
case pcmk_rc_within_range:
return out->info(out, "Rule %s is still in effect", rule_id);
case pcmk_rc_ok:
return out->info(out, "Rule %s satisfies conditions", rule_id);
case pcmk_rc_after_range:
return out->info(out, "Rule %s is expired", rule_id);
case pcmk_rc_before_range:
return out->info(out, "Rule %s has not yet taken effect", rule_id);
case pcmk_rc_op_unsatisfied:
return out->info(out, "Rule %s does not satisfy conditions",
rule_id);
default:
out->err(out,
"Could not determine whether rule %s is in effect: %s",
rule_id, ((error != NULL)? error : "unexpected error"));
return pcmk_rc_ok;
}
}
PCMK__OUTPUT_ARGS("rule-check", "const char *", "int", "const char *")
static int
rule_check_xml(pcmk__output_t *out, va_list args)
{
const char *rule_id = va_arg(args, const char *);
int result = va_arg(args, int);
const char *error = va_arg(args, const char *);
char *rc_str = pcmk__itoa(pcmk_rc2exitc(result));
pcmk__output_create_xml_node(out, "rule-check",
"rule-id", rule_id,
"rc", rc_str,
NULL);
free(rc_str);
switch (result) {
case pcmk_rc_within_range:
case pcmk_rc_ok:
case pcmk_rc_after_range:
case pcmk_rc_before_range:
case pcmk_rc_op_unsatisfied:
return pcmk_rc_ok;
default:
out->err(out,
"Could not determine whether rule %s is in effect: %s",
rule_id, ((error != NULL)? error : "unexpected error"));
return pcmk_rc_ok;
}
}
PCMK__OUTPUT_ARGS("result-code", "int", "const char *", "const char *")
static int
result_code_none(pcmk__output_t *out, va_list args)
{
return pcmk_rc_no_output;
}
PCMK__OUTPUT_ARGS("result-code", "int", "const char *", "const char *")
static int
result_code_text(pcmk__output_t *out, va_list args)
{
int code = va_arg(args, int);
const char *name = va_arg(args, const char *);
const char *desc = va_arg(args, const char *);
static int code_width = 0;
if (out->is_quiet(out)) {
/* If out->is_quiet(), don't print the code. Print name and/or desc in a
* compact format for text output, or print nothing at all for none-type
* output.
*/
if ((name != NULL) && (desc != NULL)) {
pcmk__formatted_printf(out, "%s - %s\n", name, desc);
} else if ((name != NULL) || (desc != NULL)) {
pcmk__formatted_printf(out, "%s\n", ((name != NULL)? name : desc));
}
return pcmk_rc_ok;
}
/* Get length of longest (most negative) standard Pacemaker return code
* This should be longer than all the values of any other type of return
* code.
*/
if (code_width == 0) {
long long most_negative = pcmk_rc_error - (long long) pcmk__n_rc + 1;
code_width = (int) snprintf(NULL, 0, "%lld", most_negative);
}
if ((name != NULL) && (desc != NULL)) {
static int name_width = 0;
if (name_width == 0) {
// Get length of longest standard Pacemaker return code name
for (int lpc = 0; lpc < pcmk__n_rc; lpc++) {
int len = (int) strlen(pcmk_rc_name(pcmk_rc_error - lpc));
name_width = QB_MAX(name_width, len);
}
}
return out->info(out, "% *d: %-*s %s", code_width, code, name_width,
name, desc);
}
if ((name != NULL) || (desc != NULL)) {
return out->info(out, "% *d: %s", code_width, code,
((name != NULL)? name : desc));
}
return out->info(out, "% *d", code_width, code);
}
PCMK__OUTPUT_ARGS("result-code", "int", "const char *", "const char *")
static int
result_code_xml(pcmk__output_t *out, va_list args)
{
int code = va_arg(args, int);
const char *name = va_arg(args, const char *);
const char *desc = va_arg(args, const char *);
char *code_str = pcmk__itoa(code);
pcmk__output_create_xml_node(out, PCMK_XE_RESULT_CODE,
PCMK_XA_CODE, code_str,
PCMK_XA_NAME, name,
PCMK_XA_DESCRIPTION, desc,
NULL);
free(code_str);
return pcmk_rc_ok;
}
static pcmk__message_entry_t fmt_functions[] = {
{ "attribute", "default", attribute_default },
{ "attribute", "xml", attribute_xml },
{ "cluster-status", "default", pcmk__cluster_status_text },
{ "cluster-status", "html", cluster_status_html },
{ "cluster-status", "xml", cluster_status_xml },
{ "crmadmin-node", "default", crmadmin_node },
{ "crmadmin-node", "text", crmadmin_node_text },
{ "crmadmin-node", "xml", crmadmin_node_xml },
{ "dc", "default", dc },
{ "dc", "text", dc_text },
{ "dc", "xml", dc_xml },
{ "digests", "default", digests_text },
{ "digests", "xml", digests_xml },
{ "health", "default", health },
{ "health", "text", health_text },
{ "health", "xml", health_xml },
{ "inject-attr", "default", inject_attr },
{ "inject-attr", "xml", inject_attr_xml },
{ "inject-cluster-action", "default", inject_cluster_action },
{ "inject-cluster-action", "xml", inject_cluster_action_xml },
{ "inject-fencing-action", "default", inject_fencing_action },
{ "inject-fencing-action", "xml", inject_fencing_action_xml },
{ "inject-modify-config", "default", inject_modify_config },
{ "inject-modify-config", "xml", inject_modify_config_xml },
{ "inject-modify-node", "default", inject_modify_node },
{ "inject-modify-node", "xml", inject_modify_node_xml },
{ "inject-modify-ticket", "default", inject_modify_ticket },
{ "inject-modify-ticket", "xml", inject_modify_ticket_xml },
{ "inject-pseudo-action", "default", inject_pseudo_action },
{ "inject-pseudo-action", "xml", inject_pseudo_action_xml },
{ "inject-rsc-action", "default", inject_rsc_action },
{ "inject-rsc-action", "xml", inject_rsc_action_xml },
{ "inject-spec", "default", inject_spec },
{ "inject-spec", "xml", inject_spec_xml },
{ "locations-list", "default", locations_list },
{ "locations-list", "xml", locations_list_xml },
{ "node-action", "default", node_action },
{ "node-action", "xml", node_action_xml },
{ "node-info", "default", node_info_default },
{ "node-info", "xml", node_info_xml },
{ "pacemakerd-health", "default", pacemakerd_health },
{ "pacemakerd-health", "html", pacemakerd_health_html },
{ "pacemakerd-health", "text", pacemakerd_health_text },
{ "pacemakerd-health", "xml", pacemakerd_health_xml },
{ "profile", "default", profile_default, },
{ "profile", "xml", profile_xml },
{ "result-code", "none", result_code_none },
{ "result-code", "text", result_code_text },
{ "result-code", "xml", result_code_xml },
{ "rsc-action", "default", rsc_action_default },
{ "rsc-action-item", "default", rsc_action_item },
{ "rsc-action-item", "xml", rsc_action_item_xml },
{ "rsc-is-colocated-with-list", "default", rsc_is_colocated_with_list },
{ "rsc-is-colocated-with-list", "xml", rsc_is_colocated_with_list_xml },
{ "rscs-colocated-with-list", "default", rscs_colocated_with_list },
{ "rscs-colocated-with-list", "xml", rscs_colocated_with_list_xml },
{ "rule-check", "default", rule_check_default },
{ "rule-check", "xml", rule_check_xml },
{ "locations-and-colocations", "default", locations_and_colocations },
{ "locations-and-colocations", "xml", locations_and_colocations_xml },
{ NULL, NULL, NULL }
};
void
pcmk__register_lib_messages(pcmk__output_t *out) {
pcmk__register_messages(out, fmt_functions);
}
diff --git a/lib/pengine/unpack.c b/lib/pengine/unpack.c
index 744648d092..46f9c08bea 100644
--- a/lib/pengine/unpack.c
+++ b/lib/pengine/unpack.c
@@ -1,5173 +1,5173 @@
/*
* Copyright 2004-2024 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.
*/
#include <crm_internal.h>
#include <stdio.h>
#include <string.h>
#include <glib.h>
#include <time.h>
#include <crm/crm.h>
#include <crm/services.h>
#include <crm/msg_xml.h>
#include <crm/common/xml.h>
#include <crm/common/xml_internal.h>
#include <crm/common/util.h>
#include <crm/pengine/rules.h>
#include <crm/pengine/internal.h>
#include <pe_status_private.h>
CRM_TRACE_INIT_DATA(pe_status);
// A (parsed) resource action history entry
struct action_history {
pcmk_resource_t *rsc; // Resource that history is for
pcmk_node_t *node; // Node that history is for
xmlNode *xml; // History entry XML
// Parsed from entry XML
const char *id; // XML ID of history entry
const char *key; // Operation key of action
const char *task; // Action name
const char *exit_reason; // Exit reason given for result
guint interval_ms; // Action interval
int call_id; // Call ID of action
int expected_exit_status; // Expected exit status of action
int exit_status; // Actual exit status of action
int execution_status; // Execution status of action
};
/* This uses pcmk__set_flags_as()/pcmk__clear_flags_as() directly rather than
* use pcmk__set_scheduler_flags()/pcmk__clear_scheduler_flags() so that the
* flag is stringified more readably in log messages.
*/
#define set_config_flag(scheduler, option, flag) do { \
GHashTable *config_hash = (scheduler)->config_hash; \
const char *scf_value = pcmk__cluster_option(config_hash, (option)); \
\
if (scf_value != NULL) { \
if (crm_is_true(scf_value)) { \
(scheduler)->flags = pcmk__set_flags_as(__func__, __LINE__, \
LOG_TRACE, "Scheduler", \
crm_system_name, (scheduler)->flags, \
(flag), #flag); \
} else { \
(scheduler)->flags = pcmk__clear_flags_as(__func__, __LINE__, \
LOG_TRACE, "Scheduler", \
crm_system_name, (scheduler)->flags, \
(flag), #flag); \
} \
} \
} while(0)
static void unpack_rsc_op(pcmk_resource_t *rsc, pcmk_node_t *node,
xmlNode *xml_op, xmlNode **last_failure,
enum action_fail_response *failed);
static void determine_remote_online_status(pcmk_scheduler_t *scheduler,
pcmk_node_t *this_node);
static void add_node_attrs(const xmlNode *xml_obj, pcmk_node_t *node,
bool overwrite, pcmk_scheduler_t *scheduler);
static void determine_online_status(const xmlNode *node_state,
pcmk_node_t *this_node,
pcmk_scheduler_t *scheduler);
static void unpack_node_lrm(pcmk_node_t *node, const xmlNode *xml,
pcmk_scheduler_t *scheduler);
static gboolean
is_dangling_guest_node(pcmk_node_t *node)
{
/* we are looking for a remote-node that was supposed to be mapped to a
* container resource, but all traces of that container have disappeared
* from both the config and the status section. */
if (pe__is_guest_or_remote_node(node) &&
node->details->remote_rsc &&
node->details->remote_rsc->container == NULL &&
pcmk_is_set(node->details->remote_rsc->flags,
pcmk_rsc_removed_filler)) {
return TRUE;
}
return FALSE;
}
/*!
* \brief Schedule a fence action for a node
*
* \param[in,out] scheduler Scheduler data
* \param[in,out] node Node to fence
* \param[in] reason Text description of why fencing is needed
* \param[in] priority_delay Whether to consider
* \c PCMK_OPT_PRIORITY_FENCING_DELAY
*/
void
pe_fence_node(pcmk_scheduler_t *scheduler, pcmk_node_t *node,
const char *reason, bool priority_delay)
{
CRM_CHECK(node, return);
/* A guest node is fenced by marking its container as failed */
if (pe__is_guest_node(node)) {
pcmk_resource_t *rsc = node->details->remote_rsc->container;
if (!pcmk_is_set(rsc->flags, pcmk_rsc_failed)) {
if (!pcmk_is_set(rsc->flags, pcmk_rsc_managed)) {
crm_notice("Not fencing guest node %s "
"(otherwise would because %s): "
"its guest resource %s is unmanaged",
pcmk__node_name(node), reason, rsc->id);
} else {
pcmk__sched_warn("Guest node %s will be fenced "
"(by recovering its guest resource %s): %s",
pcmk__node_name(node), rsc->id, reason);
/* We don't mark the node as unclean because that would prevent the
* node from running resources. We want to allow it to run resources
* in this transition if the recovery succeeds.
*/
node->details->remote_requires_reset = TRUE;
pcmk__set_rsc_flags(rsc,
pcmk_rsc_failed|pcmk_rsc_stop_if_failed);
}
}
} else if (is_dangling_guest_node(node)) {
crm_info("Cleaning up dangling connection for guest node %s: "
"fencing was already done because %s, "
"and guest resource no longer exists",
pcmk__node_name(node), reason);
pcmk__set_rsc_flags(node->details->remote_rsc,
pcmk_rsc_failed|pcmk_rsc_stop_if_failed);
} else if (pe__is_remote_node(node)) {
pcmk_resource_t *rsc = node->details->remote_rsc;
if ((rsc != NULL) && !pcmk_is_set(rsc->flags, pcmk_rsc_managed)) {
crm_notice("Not fencing remote node %s "
"(otherwise would because %s): connection is unmanaged",
pcmk__node_name(node), reason);
} else if(node->details->remote_requires_reset == FALSE) {
node->details->remote_requires_reset = TRUE;
pcmk__sched_warn("Remote node %s %s: %s",
pcmk__node_name(node),
pe_can_fence(scheduler, node)? "will be fenced" : "is unclean",
reason);
}
node->details->unclean = TRUE;
// No need to apply PCMK_OPT_PRIORITY_FENCING_DELAY for remote nodes
pe_fence_op(node, NULL, TRUE, reason, FALSE, scheduler);
} else if (node->details->unclean) {
crm_trace("Cluster node %s %s because %s",
pcmk__node_name(node),
pe_can_fence(scheduler, node)? "would also be fenced" : "also is unclean",
reason);
} else {
pcmk__sched_warn("Cluster node %s %s: %s",
pcmk__node_name(node),
pe_can_fence(scheduler, node)? "will be fenced" : "is unclean",
reason);
node->details->unclean = TRUE;
pe_fence_op(node, NULL, TRUE, reason, priority_delay, scheduler);
}
}
// @TODO xpaths can't handle templates, rules, or id-refs
// nvpair with provides or requires set to unfencing
#define XPATH_UNFENCING_NVPAIR PCMK_XE_NVPAIR \
"[(@" PCMK_XA_NAME "='" PCMK_STONITH_PROVIDES "'" \
"or @" PCMK_XA_NAME "='" PCMK_META_REQUIRES "') " \
"and @" PCMK_XA_VALUE "='" PCMK__VALUE_UNFENCING "']"
// unfencing in rsc_defaults or any resource
#define XPATH_ENABLE_UNFENCING \
"/" PCMK_XE_CIB "/" PCMK_XE_CONFIGURATION "/" PCMK_XE_RESOURCES \
"//" PCMK_XE_META_ATTRIBUTES "/" XPATH_UNFENCING_NVPAIR \
"|/" PCMK_XE_CIB "/" PCMK_XE_CONFIGURATION "/" PCMK_XE_RSC_DEFAULTS \
"/" PCMK_XE_META_ATTRIBUTES "/" XPATH_UNFENCING_NVPAIR
static void
set_if_xpath(uint64_t flag, const char *xpath, pcmk_scheduler_t *scheduler)
{
xmlXPathObjectPtr result = NULL;
if (!pcmk_is_set(scheduler->flags, flag)) {
result = xpath_search(scheduler->input, xpath);
if (result && (numXpathResults(result) > 0)) {
pcmk__set_scheduler_flags(scheduler, flag);
}
freeXpathObject(result);
}
}
gboolean
unpack_config(xmlNode *config, pcmk_scheduler_t *scheduler)
{
const char *value = NULL;
guint interval_ms = 0U;
GHashTable *config_hash = pcmk__strkey_table(free, free);
pe_rule_eval_data_t rule_data = {
.node_hash = NULL,
.role = pcmk_role_unknown,
.now = scheduler->now,
.match_data = NULL,
.rsc_data = NULL,
.op_data = NULL
};
scheduler->config_hash = config_hash;
pe__unpack_dataset_nvpairs(config, PCMK_XE_CLUSTER_PROPERTY_SET, &rule_data,
config_hash, PCMK_VALUE_CIB_BOOTSTRAP_OPTIONS,
FALSE, scheduler);
pcmk__validate_cluster_options(config_hash);
set_config_flag(scheduler, PCMK_OPT_ENABLE_STARTUP_PROBES,
pcmk_sched_probe_resources);
if (!pcmk_is_set(scheduler->flags, pcmk_sched_probe_resources)) {
crm_info("Startup probes: disabled (dangerous)");
}
value = pcmk__cluster_option(config_hash, PCMK_OPT_HAVE_WATCHDOG);
if (value && crm_is_true(value)) {
crm_info("Watchdog-based self-fencing will be performed via SBD if "
"fencing is required and " PCMK_OPT_STONITH_WATCHDOG_TIMEOUT
" is nonzero");
pcmk__set_scheduler_flags(scheduler, pcmk_sched_have_fencing);
}
/* Set certain flags via xpath here, so they can be used before the relevant
* configuration sections are unpacked.
*/
set_if_xpath(pcmk_sched_enable_unfencing, XPATH_ENABLE_UNFENCING,
scheduler);
value = pcmk__cluster_option(config_hash, PCMK_OPT_STONITH_TIMEOUT);
pcmk_parse_interval_spec(value, &interval_ms);
if (interval_ms >= INT_MAX) {
scheduler->stonith_timeout = INT_MAX;
} else {
scheduler->stonith_timeout = (int) interval_ms;
}
crm_debug("STONITH timeout: %d", scheduler->stonith_timeout);
set_config_flag(scheduler, PCMK_OPT_STONITH_ENABLED,
pcmk_sched_fencing_enabled);
if (pcmk_is_set(scheduler->flags, pcmk_sched_fencing_enabled)) {
crm_debug("STONITH of failed nodes is enabled");
} else {
crm_debug("STONITH of failed nodes is disabled");
}
scheduler->stonith_action = pcmk__cluster_option(config_hash,
PCMK_OPT_STONITH_ACTION);
if (!strcmp(scheduler->stonith_action, "poweroff")) {
pcmk__warn_once(pcmk__wo_poweroff,
"Support for " PCMK_OPT_STONITH_ACTION " of "
"'poweroff' is deprecated and will be removed in a "
"future release (use 'off' instead)");
scheduler->stonith_action = PCMK_ACTION_OFF;
}
crm_trace("STONITH will %s nodes", scheduler->stonith_action);
set_config_flag(scheduler, PCMK_OPT_CONCURRENT_FENCING,
pcmk_sched_concurrent_fencing);
if (pcmk_is_set(scheduler->flags, pcmk_sched_concurrent_fencing)) {
crm_debug("Concurrent fencing is enabled");
} else {
crm_debug("Concurrent fencing is disabled");
}
value = pcmk__cluster_option(config_hash, PCMK_OPT_PRIORITY_FENCING_DELAY);
if (value) {
pcmk_parse_interval_spec(value, &interval_ms);
scheduler->priority_fencing_delay = (int) (interval_ms / 1000);
crm_trace("Priority fencing delay is %ds",
scheduler->priority_fencing_delay);
}
set_config_flag(scheduler, PCMK_OPT_STOP_ALL_RESOURCES,
pcmk_sched_stop_all);
crm_debug("Stop all active resources: %s",
pcmk__flag_text(scheduler->flags, pcmk_sched_stop_all));
set_config_flag(scheduler, PCMK_OPT_SYMMETRIC_CLUSTER,
pcmk_sched_symmetric_cluster);
if (pcmk_is_set(scheduler->flags, pcmk_sched_symmetric_cluster)) {
crm_debug("Cluster is symmetric" " - resources can run anywhere by default");
}
value = pcmk__cluster_option(config_hash, PCMK_OPT_NO_QUORUM_POLICY);
if (pcmk__str_eq(value, PCMK_VALUE_IGNORE, pcmk__str_casei)) {
scheduler->no_quorum_policy = pcmk_no_quorum_ignore;
} else if (pcmk__str_eq(value, PCMK_VALUE_FREEZE, pcmk__str_casei)) {
scheduler->no_quorum_policy = pcmk_no_quorum_freeze;
} else if (pcmk__str_eq(value, PCMK_VALUE_DEMOTE, pcmk__str_casei)) {
scheduler->no_quorum_policy = pcmk_no_quorum_demote;
} else if (pcmk__str_eq(value, PCMK_VALUE_FENCE_LEGACY, pcmk__str_casei)) {
if (pcmk_is_set(scheduler->flags, pcmk_sched_fencing_enabled)) {
int do_panic = 0;
crm_element_value_int(scheduler->input, PCMK_XA_NO_QUORUM_PANIC,
&do_panic);
if (do_panic || pcmk_is_set(scheduler->flags, pcmk_sched_quorate)) {
scheduler->no_quorum_policy = pcmk_no_quorum_fence;
} else {
crm_notice("Resetting " PCMK_OPT_NO_QUORUM_POLICY
" to 'stop': cluster has never had quorum");
scheduler->no_quorum_policy = pcmk_no_quorum_stop;
}
} else {
pcmk__config_err("Resetting " PCMK_OPT_NO_QUORUM_POLICY
" to 'stop' because fencing is disabled");
scheduler->no_quorum_policy = pcmk_no_quorum_stop;
}
} else {
scheduler->no_quorum_policy = pcmk_no_quorum_stop;
}
switch (scheduler->no_quorum_policy) {
case pcmk_no_quorum_freeze:
crm_debug("On loss of quorum: Freeze resources");
break;
case pcmk_no_quorum_stop:
crm_debug("On loss of quorum: Stop ALL resources");
break;
case pcmk_no_quorum_demote:
crm_debug("On loss of quorum: "
"Demote promotable resources and stop other resources");
break;
case pcmk_no_quorum_fence:
crm_notice("On loss of quorum: Fence all remaining nodes");
break;
case pcmk_no_quorum_ignore:
crm_notice("On loss of quorum: Ignore");
break;
}
set_config_flag(scheduler, PCMK_OPT_STOP_ORPHAN_RESOURCES,
pcmk_sched_stop_removed_resources);
if (pcmk_is_set(scheduler->flags, pcmk_sched_stop_removed_resources)) {
crm_trace("Orphan resources are stopped");
} else {
crm_trace("Orphan resources are ignored");
}
set_config_flag(scheduler, PCMK_OPT_STOP_ORPHAN_ACTIONS,
pcmk_sched_cancel_removed_actions);
if (pcmk_is_set(scheduler->flags, pcmk_sched_cancel_removed_actions)) {
crm_trace("Orphan resource actions are stopped");
} else {
crm_trace("Orphan resource actions are ignored");
}
value = pcmk__cluster_option(config_hash, PCMK__OPT_REMOVE_AFTER_STOP);
if (value != NULL) {
if (crm_is_true(value)) {
pcmk__set_scheduler_flags(scheduler, pcmk_sched_remove_after_stop);
#ifndef PCMK__COMPAT_2_0
pcmk__warn_once(pcmk__wo_remove_after,
"Support for the " PCMK__OPT_REMOVE_AFTER_STOP
" cluster property is deprecated and will be "
"removed in a future release");
#endif
} else {
pcmk__clear_scheduler_flags(scheduler,
pcmk_sched_remove_after_stop);
}
}
set_config_flag(scheduler, PCMK_OPT_MAINTENANCE_MODE,
pcmk_sched_in_maintenance);
crm_trace("Maintenance mode: %s",
pcmk__flag_text(scheduler->flags, pcmk_sched_in_maintenance));
set_config_flag(scheduler, PCMK_OPT_START_FAILURE_IS_FATAL,
pcmk_sched_start_failure_fatal);
if (pcmk_is_set(scheduler->flags, pcmk_sched_start_failure_fatal)) {
crm_trace("Start failures are always fatal");
} else {
crm_trace("Start failures are handled by failcount");
}
if (pcmk_is_set(scheduler->flags, pcmk_sched_fencing_enabled)) {
set_config_flag(scheduler, PCMK_OPT_STARTUP_FENCING,
pcmk_sched_startup_fencing);
}
if (pcmk_is_set(scheduler->flags, pcmk_sched_startup_fencing)) {
crm_trace("Unseen nodes will be fenced");
} else {
pcmk__warn_once(pcmk__wo_blind,
"Blind faith: not fencing unseen nodes");
}
pe__unpack_node_health_scores(scheduler);
scheduler->placement_strategy =
pcmk__cluster_option(config_hash, PCMK_OPT_PLACEMENT_STRATEGY);
crm_trace("Placement strategy: %s", scheduler->placement_strategy);
set_config_flag(scheduler, PCMK_OPT_SHUTDOWN_LOCK,
pcmk_sched_shutdown_lock);
if (pcmk_is_set(scheduler->flags, pcmk_sched_shutdown_lock)) {
value = pcmk__cluster_option(config_hash, PCMK_OPT_SHUTDOWN_LOCK_LIMIT);
pcmk_parse_interval_spec(value, &(scheduler->shutdown_lock));
scheduler->shutdown_lock /= 1000;
crm_trace("Resources will be locked to nodes that were cleanly "
"shut down (locks expire after %s)",
pcmk__readable_interval(scheduler->shutdown_lock));
} else {
crm_trace("Resources will not be locked to nodes that were cleanly "
"shut down");
}
value = pcmk__cluster_option(config_hash, PCMK_OPT_NODE_PENDING_TIMEOUT);
pcmk_parse_interval_spec(value, &(scheduler->node_pending_timeout));
scheduler->node_pending_timeout /= 1000;
if (scheduler->node_pending_timeout == 0) {
crm_trace("Do not fence pending nodes");
} else {
crm_trace("Fence pending nodes after %s",
pcmk__readable_interval(scheduler->node_pending_timeout
* 1000));
}
return TRUE;
}
pcmk_node_t *
pe_create_node(const char *id, const char *uname, const char *type,
const char *score, pcmk_scheduler_t *scheduler)
{
pcmk_node_t *new_node = NULL;
if (pe_find_node(scheduler->nodes, uname) != NULL) {
pcmk__config_warn("More than one node entry has name '%s'", uname);
}
new_node = calloc(1, sizeof(pcmk_node_t));
if (new_node == NULL) {
pcmk__sched_err("Could not allocate memory for node %s", uname);
return NULL;
}
new_node->weight = char2score(score);
new_node->details = calloc(1, sizeof(struct pe_node_shared_s));
if (new_node->details == NULL) {
free(new_node);
pcmk__sched_err("Could not allocate memory for node %s", uname);
return NULL;
}
crm_trace("Creating node for entry %s/%s", uname, id);
new_node->details->id = id;
new_node->details->uname = uname;
new_node->details->online = FALSE;
new_node->details->shutdown = FALSE;
new_node->details->rsc_discovery_enabled = TRUE;
new_node->details->running_rsc = NULL;
new_node->details->data_set = scheduler;
if (pcmk__str_eq(type, PCMK_VALUE_MEMBER,
pcmk__str_null_matches|pcmk__str_casei)) {
new_node->details->type = pcmk_node_variant_cluster;
} else if (pcmk__str_eq(type, PCMK_VALUE_REMOTE, pcmk__str_casei)) {
new_node->details->type = pcmk_node_variant_remote;
pcmk__set_scheduler_flags(scheduler, pcmk_sched_have_remote_nodes);
} else {
/* @COMPAT 'ping' is the default for backward compatibility, but it
* should be changed to 'member' at a compatibility break
*/
if (!pcmk__str_eq(type, PCMK__VALUE_PING, pcmk__str_casei)) {
pcmk__config_warn("Node %s has unrecognized type '%s', "
"assuming '" PCMK__VALUE_PING "'",
pcmk__s(uname, "without name"), type);
}
pcmk__warn_once(pcmk__wo_ping_node,
"Support for nodes of type '" PCMK__VALUE_PING "' "
"(such as %s) is deprecated and will be removed in a "
"future release",
pcmk__s(uname, "unnamed node"));
new_node->details->type = node_ping;
}
new_node->details->attrs = pcmk__strkey_table(free, free);
if (pe__is_guest_or_remote_node(new_node)) {
g_hash_table_insert(new_node->details->attrs, strdup(CRM_ATTR_KIND),
strdup("remote"));
} else {
g_hash_table_insert(new_node->details->attrs, strdup(CRM_ATTR_KIND),
strdup("cluster"));
}
new_node->details->utilization = pcmk__strkey_table(free, free);
new_node->details->digest_cache = pcmk__strkey_table(free,
pe__free_digests);
scheduler->nodes = g_list_insert_sorted(scheduler->nodes, new_node,
pe__cmp_node_name);
return new_node;
}
static const char *
expand_remote_rsc_meta(xmlNode *xml_obj, xmlNode *parent, pcmk_scheduler_t *data)
{
xmlNode *attr_set = NULL;
xmlNode *attr = NULL;
const char *container_id = ID(xml_obj);
const char *remote_name = NULL;
const char *remote_server = NULL;
const char *remote_port = NULL;
const char *connect_timeout = "60s";
const char *remote_allow_migrate=NULL;
const char *is_managed = NULL;
for (attr_set = pcmk__xe_first_child(xml_obj); attr_set != NULL;
attr_set = pcmk__xe_next(attr_set)) {
if (!pcmk__str_eq((const char *) attr_set->name,
PCMK_XE_META_ATTRIBUTES, pcmk__str_casei)) {
continue;
}
for (attr = pcmk__xe_first_child(attr_set); attr != NULL;
attr = pcmk__xe_next(attr)) {
const char *value = crm_element_value(attr, PCMK_XA_VALUE);
const char *name = crm_element_value(attr, PCMK_XA_NAME);
if (name == NULL) { // Sanity
continue;
}
if (strcmp(name, PCMK_META_REMOTE_NODE) == 0) {
remote_name = value;
} else if (strcmp(name, PCMK_META_REMOTE_ADDR) == 0) {
remote_server = value;
} else if (strcmp(name, PCMK_META_REMOTE_PORT) == 0) {
remote_port = value;
} else if (strcmp(name, PCMK_META_REMOTE_CONNECT_TIMEOUT) == 0) {
connect_timeout = value;
} else if (strcmp(name, PCMK_META_REMOTE_ALLOW_MIGRATE) == 0) {
remote_allow_migrate = value;
} else if (strcmp(name, PCMK_META_IS_MANAGED) == 0) {
is_managed = value;
}
}
}
if (remote_name == NULL) {
return NULL;
}
if (pe_find_resource(data->resources, remote_name) != NULL) {
return NULL;
}
pe_create_remote_xml(parent, remote_name, container_id,
remote_allow_migrate, is_managed,
connect_timeout, remote_server, remote_port);
return remote_name;
}
static void
handle_startup_fencing(pcmk_scheduler_t *scheduler, pcmk_node_t *new_node)
{
if ((new_node->details->type == pcmk_node_variant_remote)
&& (new_node->details->remote_rsc == NULL)) {
/* Ignore fencing for remote nodes that don't have a connection resource
* associated with them. This happens when remote node entries get left
* in the nodes section after the connection resource is removed.
*/
return;
}
if (pcmk_is_set(scheduler->flags, pcmk_sched_startup_fencing)) {
// All nodes are unclean until we've seen their status entry
new_node->details->unclean = TRUE;
} else {
// Blind faith ...
new_node->details->unclean = FALSE;
}
/* We need to be able to determine if a node's status section
* exists or not separate from whether the node is unclean. */
new_node->details->unseen = TRUE;
}
gboolean
unpack_nodes(xmlNode *xml_nodes, pcmk_scheduler_t *scheduler)
{
xmlNode *xml_obj = NULL;
pcmk_node_t *new_node = NULL;
const char *id = NULL;
const char *uname = NULL;
const char *type = NULL;
const char *score = NULL;
for (xml_obj = pcmk__xe_first_child(xml_nodes); xml_obj != NULL;
xml_obj = pcmk__xe_next(xml_obj)) {
if (pcmk__str_eq((const char *) xml_obj->name, PCMK_XE_NODE,
pcmk__str_none)) {
new_node = NULL;
id = crm_element_value(xml_obj, PCMK_XA_ID);
uname = crm_element_value(xml_obj, PCMK_XA_UNAME);
type = crm_element_value(xml_obj, PCMK_XA_TYPE);
score = crm_element_value(xml_obj, PCMK_XA_SCORE);
crm_trace("Processing node %s/%s", uname, id);
if (id == NULL) {
pcmk__config_err("Ignoring <" PCMK_XE_NODE
"> entry in configuration without id");
continue;
}
new_node = pe_create_node(id, uname, type, score, scheduler);
if (new_node == NULL) {
return FALSE;
}
handle_startup_fencing(scheduler, new_node);
add_node_attrs(xml_obj, new_node, FALSE, scheduler);
crm_trace("Done with node %s",
crm_element_value(xml_obj, PCMK_XA_UNAME));
}
}
if (scheduler->localhost
&& (pe_find_node(scheduler->nodes, scheduler->localhost) == NULL)) {
crm_info("Creating a fake local node");
pe_create_node(scheduler->localhost, scheduler->localhost, NULL, 0,
scheduler);
}
return TRUE;
}
static void
setup_container(pcmk_resource_t *rsc, pcmk_scheduler_t *scheduler)
{
const char *container_id = NULL;
if (rsc->children) {
g_list_foreach(rsc->children, (GFunc) setup_container, scheduler);
return;
}
container_id = g_hash_table_lookup(rsc->meta, PCMK__META_CONTAINER);
if (container_id && !pcmk__str_eq(container_id, rsc->id, pcmk__str_casei)) {
pcmk_resource_t *container = pe_find_resource(scheduler->resources,
container_id);
if (container) {
rsc->container = container;
pcmk__set_rsc_flags(container, pcmk_rsc_has_filler);
container->fillers = g_list_append(container->fillers, rsc);
pcmk__rsc_trace(rsc, "Resource %s's container is %s",
rsc->id, container_id);
} else {
pcmk__config_err("Resource %s: Unknown resource container (%s)",
rsc->id, container_id);
}
}
}
gboolean
unpack_remote_nodes(xmlNode *xml_resources, pcmk_scheduler_t *scheduler)
{
xmlNode *xml_obj = NULL;
/* Create remote nodes and guest nodes from the resource configuration
* before unpacking resources.
*/
for (xml_obj = pcmk__xe_first_child(xml_resources); xml_obj != NULL;
xml_obj = pcmk__xe_next(xml_obj)) {
const char *new_node_id = NULL;
/* Check for remote nodes, which are defined by ocf:pacemaker:remote
* primitives.
*/
if (xml_contains_remote_node(xml_obj)) {
new_node_id = ID(xml_obj);
/* The "pe_find_node" check is here to make sure we don't iterate over
* an expanded node that has already been added to the node list. */
if (new_node_id
&& (pe_find_node(scheduler->nodes, new_node_id) == NULL)) {
crm_trace("Found remote node %s defined by resource %s",
new_node_id, ID(xml_obj));
pe_create_node(new_node_id, new_node_id, PCMK_VALUE_REMOTE,
NULL, scheduler);
}
continue;
}
/* Check for guest nodes, which are defined by special meta-attributes
* of a primitive of any type (for example, VirtualDomain or Xen).
*/
if (pcmk__str_eq((const char *) xml_obj->name, PCMK_XE_PRIMITIVE,
pcmk__str_none)) {
/* This will add an ocf:pacemaker:remote primitive to the
* configuration for the guest node's connection, to be unpacked
* later.
*/
new_node_id = expand_remote_rsc_meta(xml_obj, xml_resources,
scheduler);
if (new_node_id
&& (pe_find_node(scheduler->nodes, new_node_id) == NULL)) {
crm_trace("Found guest node %s in resource %s",
new_node_id, ID(xml_obj));
pe_create_node(new_node_id, new_node_id, PCMK_VALUE_REMOTE,
NULL, scheduler);
}
continue;
}
/* Check for guest nodes inside a group. Clones are currently not
* supported as guest nodes.
*/
if (pcmk__str_eq((const char *) xml_obj->name, PCMK_XE_GROUP,
pcmk__str_none)) {
xmlNode *xml_obj2 = NULL;
for (xml_obj2 = pcmk__xe_first_child(xml_obj); xml_obj2 != NULL;
xml_obj2 = pcmk__xe_next(xml_obj2)) {
new_node_id = expand_remote_rsc_meta(xml_obj2, xml_resources,
scheduler);
if (new_node_id
&& (pe_find_node(scheduler->nodes, new_node_id) == NULL)) {
crm_trace("Found guest node %s in resource %s inside group %s",
new_node_id, ID(xml_obj2), ID(xml_obj));
pe_create_node(new_node_id, new_node_id, PCMK_VALUE_REMOTE,
NULL, scheduler);
}
}
}
}
return TRUE;
}
/* Call this after all the nodes and resources have been
* unpacked, but before the status section is read.
*
* A remote node's online status is reflected by the state
* of the remote node's connection resource. We need to link
* the remote node to this connection resource so we can have
* easy access to the connection resource during the scheduler calculations.
*/
static void
link_rsc2remotenode(pcmk_scheduler_t *scheduler, pcmk_resource_t *new_rsc)
{
pcmk_node_t *remote_node = NULL;
if (new_rsc->is_remote_node == FALSE) {
return;
}
if (pcmk_is_set(scheduler->flags, pcmk_sched_location_only)) {
/* remote_nodes and remote_resources are not linked in quick location calculations */
return;
}
remote_node = pe_find_node(scheduler->nodes, new_rsc->id);
CRM_CHECK(remote_node != NULL, return);
pcmk__rsc_trace(new_rsc, "Linking remote connection resource %s to %s",
new_rsc->id, pcmk__node_name(remote_node));
remote_node->details->remote_rsc = new_rsc;
if (new_rsc->container == NULL) {
/* Handle start-up fencing for remote nodes (as opposed to guest nodes)
* the same as is done for cluster nodes.
*/
handle_startup_fencing(scheduler, remote_node);
} else {
/* pe_create_node() marks the new node as "remote" or "cluster"; now
* that we know the node is a guest node, update it correctly.
*/
g_hash_table_replace(remote_node->details->attrs, strdup(CRM_ATTR_KIND),
strdup("container"));
}
}
static void
destroy_tag(gpointer data)
{
pcmk_tag_t *tag = data;
if (tag) {
free(tag->id);
g_list_free_full(tag->refs, free);
free(tag);
}
}
/*!
* \internal
* \brief Parse configuration XML for resource information
*
* \param[in] xml_resources Top of resource configuration XML
* \param[in,out] scheduler Scheduler data
*
* \return TRUE
*
* \note unpack_remote_nodes() MUST be called before this, so that the nodes can
* be used when pe__unpack_resource() calls resource_location()
*/
gboolean
unpack_resources(const xmlNode *xml_resources, pcmk_scheduler_t *scheduler)
{
xmlNode *xml_obj = NULL;
GList *gIter = NULL;
scheduler->template_rsc_sets = pcmk__strkey_table(free, destroy_tag);
for (xml_obj = pcmk__xe_first_child(xml_resources); xml_obj != NULL;
xml_obj = pcmk__xe_next(xml_obj)) {
pcmk_resource_t *new_rsc = NULL;
const char *id = ID(xml_obj);
if (pcmk__str_empty(id)) {
pcmk__config_err("Ignoring <%s> resource without ID",
xml_obj->name);
continue;
}
if (pcmk__str_eq((const char *) xml_obj->name, PCMK_XE_TEMPLATE,
pcmk__str_none)) {
if (g_hash_table_lookup_extended(scheduler->template_rsc_sets, id,
NULL, NULL) == FALSE) {
/* Record the template's ID for the knowledge of its existence anyway. */
g_hash_table_insert(scheduler->template_rsc_sets, strdup(id),
NULL);
}
continue;
}
crm_trace("Unpacking <%s " PCMK_XA_ID "='%s'>", xml_obj->name, id);
if (pe__unpack_resource(xml_obj, &new_rsc, NULL,
scheduler) == pcmk_rc_ok) {
scheduler->resources = g_list_append(scheduler->resources, new_rsc);
pcmk__rsc_trace(new_rsc, "Added resource %s", new_rsc->id);
} else {
pcmk__config_err("Ignoring <%s> resource '%s' "
"because configuration is invalid",
xml_obj->name, id);
}
}
for (gIter = scheduler->resources; gIter != NULL; gIter = gIter->next) {
pcmk_resource_t *rsc = (pcmk_resource_t *) gIter->data;
setup_container(rsc, scheduler);
link_rsc2remotenode(scheduler, rsc);
}
scheduler->resources = g_list_sort(scheduler->resources,
pe__cmp_rsc_priority);
if (pcmk_is_set(scheduler->flags, pcmk_sched_location_only)) {
/* Ignore */
} else if (pcmk_is_set(scheduler->flags, pcmk_sched_fencing_enabled)
&& !pcmk_is_set(scheduler->flags, pcmk_sched_have_fencing)) {
pcmk__config_err("Resource start-up disabled since no STONITH resources have been defined");
pcmk__config_err("Either configure some or disable STONITH with the "
PCMK_OPT_STONITH_ENABLED " option");
pcmk__config_err("NOTE: Clusters with shared data need STONITH to ensure data integrity");
}
return TRUE;
}
gboolean
unpack_tags(xmlNode *xml_tags, pcmk_scheduler_t *scheduler)
{
xmlNode *xml_tag = NULL;
scheduler->tags = pcmk__strkey_table(free, destroy_tag);
for (xml_tag = pcmk__xe_first_child(xml_tags); xml_tag != NULL;
xml_tag = pcmk__xe_next(xml_tag)) {
xmlNode *xml_obj_ref = NULL;
const char *tag_id = ID(xml_tag);
if (!pcmk__str_eq((const char *) xml_tag->name, PCMK_XE_TAG,
pcmk__str_none)) {
continue;
}
if (tag_id == NULL) {
pcmk__config_err("Ignoring <%s> without " PCMK_XA_ID,
(const char *) xml_tag->name);
continue;
}
for (xml_obj_ref = pcmk__xe_first_child(xml_tag); xml_obj_ref != NULL;
xml_obj_ref = pcmk__xe_next(xml_obj_ref)) {
const char *obj_ref = ID(xml_obj_ref);
if (!pcmk__str_eq((const char *) xml_obj_ref->name, PCMK_XE_OBJ_REF,
pcmk__str_none)) {
continue;
}
if (obj_ref == NULL) {
pcmk__config_err("Ignoring <%s> for tag '%s' without " PCMK_XA_ID,
xml_obj_ref->name, tag_id);
continue;
}
if (add_tag_ref(scheduler->tags, tag_id, obj_ref) == FALSE) {
return FALSE;
}
}
}
return TRUE;
}
/* The ticket state section:
* "/cib/status/tickets/ticket_state" */
static gboolean
unpack_ticket_state(xmlNode *xml_ticket, pcmk_scheduler_t *scheduler)
{
const char *ticket_id = NULL;
const char *granted = NULL;
const char *last_granted = NULL;
const char *standby = NULL;
xmlAttrPtr xIter = NULL;
pcmk_ticket_t *ticket = NULL;
ticket_id = ID(xml_ticket);
if (pcmk__str_empty(ticket_id)) {
return FALSE;
}
crm_trace("Processing ticket state for %s", ticket_id);
ticket = g_hash_table_lookup(scheduler->tickets, ticket_id);
if (ticket == NULL) {
ticket = ticket_new(ticket_id, scheduler);
if (ticket == NULL) {
return FALSE;
}
}
for (xIter = xml_ticket->properties; xIter; xIter = xIter->next) {
const char *prop_name = (const char *)xIter->name;
const char *prop_value = pcmk__xml_attr_value(xIter);
if (pcmk__str_eq(prop_name, PCMK_XA_ID, pcmk__str_none)) {
continue;
}
g_hash_table_replace(ticket->state, strdup(prop_name), strdup(prop_value));
}
granted = g_hash_table_lookup(ticket->state, PCMK__XA_GRANTED);
if (granted && crm_is_true(granted)) {
ticket->granted = TRUE;
crm_info("We have ticket '%s'", ticket->id);
} else {
ticket->granted = FALSE;
crm_info("We do not have ticket '%s'", ticket->id);
}
last_granted = g_hash_table_lookup(ticket->state, PCMK_XA_LAST_GRANTED);
if (last_granted) {
long long last_granted_ll;
pcmk__scan_ll(last_granted, &last_granted_ll, 0LL);
ticket->last_granted = (time_t) last_granted_ll;
}
standby = g_hash_table_lookup(ticket->state, PCMK_XA_STANDBY);
if (standby && crm_is_true(standby)) {
ticket->standby = TRUE;
if (ticket->granted) {
crm_info("Granted ticket '%s' is in standby-mode", ticket->id);
}
} else {
ticket->standby = FALSE;
}
crm_trace("Done with ticket state for %s", ticket_id);
return TRUE;
}
static gboolean
unpack_tickets_state(xmlNode *xml_tickets, pcmk_scheduler_t *scheduler)
{
xmlNode *xml_obj = NULL;
for (xml_obj = pcmk__xe_first_child(xml_tickets); xml_obj != NULL;
xml_obj = pcmk__xe_next(xml_obj)) {
if (!pcmk__str_eq((const char *) xml_obj->name, PCMK__XE_TICKET_STATE,
pcmk__str_none)) {
continue;
}
unpack_ticket_state(xml_obj, scheduler);
}
return TRUE;
}
static void
unpack_handle_remote_attrs(pcmk_node_t *this_node, const xmlNode *state,
pcmk_scheduler_t *scheduler)
{
const char *discovery = NULL;
const xmlNode *attrs = NULL;
pcmk_resource_t *rsc = NULL;
if (!pcmk__str_eq((const char *) state->name, PCMK__XE_NODE_STATE,
pcmk__str_none)) {
return;
}
if ((this_node == NULL) || !pe__is_guest_or_remote_node(this_node)) {
return;
}
crm_trace("Processing Pacemaker Remote node %s",
pcmk__node_name(this_node));
pcmk__scan_min_int(crm_element_value(state, PCMK__XA_NODE_IN_MAINTENANCE),
&(this_node->details->remote_maintenance), 0);
rsc = this_node->details->remote_rsc;
if (this_node->details->remote_requires_reset == FALSE) {
this_node->details->unclean = FALSE;
this_node->details->unseen = FALSE;
}
attrs = find_xml_node(state, PCMK__XE_TRANSIENT_ATTRIBUTES, FALSE);
add_node_attrs(attrs, this_node, TRUE, scheduler);
if (pe__shutdown_requested(this_node)) {
crm_info("%s is shutting down", pcmk__node_name(this_node));
this_node->details->shutdown = TRUE;
}
if (crm_is_true(pe_node_attribute_raw(this_node, "standby"))) {
crm_info("%s is in standby mode", pcmk__node_name(this_node));
this_node->details->standby = TRUE;
}
if (crm_is_true(pe_node_attribute_raw(this_node,
PCMK_NODE_ATTR_MAINTENANCE))
|| ((rsc != NULL) && !pcmk_is_set(rsc->flags, pcmk_rsc_managed))) {
crm_info("%s is in maintenance mode", pcmk__node_name(this_node));
this_node->details->maintenance = TRUE;
}
discovery =
pe_node_attribute_raw(this_node,
PCMK__NODE_ATTR_RESOURCE_DISCOVERY_ENABLED);
if ((discovery != NULL) && !crm_is_true(discovery)) {
pcmk__warn_once(pcmk__wo_rdisc_enabled,
"Support for the "
PCMK__NODE_ATTR_RESOURCE_DISCOVERY_ENABLED
" node attribute is deprecated and will be removed"
" (and behave as 'true') in a future release.");
if (pe__is_remote_node(this_node)
&& !pcmk_is_set(scheduler->flags, pcmk_sched_fencing_enabled)) {
pcmk__config_warn("Ignoring "
PCMK__NODE_ATTR_RESOURCE_DISCOVERY_ENABLED
" attribute on Pacemaker Remote node %s"
" because fencing is disabled",
pcmk__node_name(this_node));
} else {
/* This is either a remote node with fencing enabled, or a guest
* node. We don't care whether fencing is enabled when fencing guest
* nodes, because they are "fenced" by recovering their containing
* resource.
*/
crm_info("%s has resource discovery disabled",
pcmk__node_name(this_node));
this_node->details->rsc_discovery_enabled = FALSE;
}
}
}
/*!
* \internal
* \brief Unpack a cluster node's transient attributes
*
* \param[in] state CIB node state XML
* \param[in,out] node Cluster node whose attributes are being unpacked
* \param[in,out] scheduler Scheduler data
*/
static void
unpack_transient_attributes(const xmlNode *state, pcmk_node_t *node,
pcmk_scheduler_t *scheduler)
{
const char *discovery = NULL;
const xmlNode *attrs = find_xml_node(state, PCMK__XE_TRANSIENT_ATTRIBUTES,
FALSE);
add_node_attrs(attrs, node, TRUE, scheduler);
if (crm_is_true(pe_node_attribute_raw(node, "standby"))) {
crm_info("%s is in standby mode", pcmk__node_name(node));
node->details->standby = TRUE;
}
if (crm_is_true(pe_node_attribute_raw(node, PCMK_NODE_ATTR_MAINTENANCE))) {
crm_info("%s is in maintenance mode", pcmk__node_name(node));
node->details->maintenance = TRUE;
}
discovery
= pe_node_attribute_raw(node,
PCMK__NODE_ATTR_RESOURCE_DISCOVERY_ENABLED);
if ((discovery != NULL) && !crm_is_true(discovery)) {
pcmk__config_warn("Ignoring "
PCMK__NODE_ATTR_RESOURCE_DISCOVERY_ENABLED
" attribute for %s because disabling resource"
" discovery is not allowed for cluster nodes",
pcmk__node_name(node));
}
}
/*!
* \internal
* \brief Unpack a node state entry (first pass)
*
* Unpack one node state entry from status. This unpacks information from the
* \C PCMK__XE_NODE_STATE element itself and node attributes inside it, but not
* the resource history inside it. Multiple passes through the status are needed
* to fully unpack everything.
*
* \param[in] state CIB node state XML
* \param[in,out] scheduler Scheduler data
*/
static void
unpack_node_state(const xmlNode *state, pcmk_scheduler_t *scheduler)
{
const char *id = NULL;
const char *uname = NULL;
pcmk_node_t *this_node = NULL;
id = crm_element_value(state, PCMK_XA_ID);
if (id == NULL) {
pcmk__config_err("Ignoring invalid " PCMK__XE_NODE_STATE " entry without "
PCMK_XA_ID);
crm_log_xml_info(state, "missing-id");
return;
}
uname = crm_element_value(state, PCMK_XA_UNAME);
if (uname == NULL) {
/* If a joining peer makes the cluster acquire the quorum from corosync
* meanwhile it has not joined CPG membership of pacemaker-controld yet,
* it's possible that the created PCMK__XE_NODE_STATE entry doesn't have
* a PCMK_XA_UNAME yet. We should recognize the node as `pending` and
* wait for it to join CPG.
*/
crm_trace("Handling " PCMK__XE_NODE_STATE " entry with id=\"%s\" "
"without " PCMK_XA_UNAME,
id);
}
this_node = pe_find_node_any(scheduler->nodes, id, uname);
if (this_node == NULL) {
pcmk__config_warn("Ignoring recorded node state for id=\"%s\" (%s) "
"because it is no longer in the configuration",
id, pcmk__s(uname, "uname unknown"));
return;
}
if (pe__is_guest_or_remote_node(this_node)) {
/* We can't determine the online status of Pacemaker Remote nodes until
* after all resource history has been unpacked. In this first pass, we
* do need to mark whether the node has been fenced, as this plays a
* role during unpacking cluster node resource state.
*/
pcmk__scan_min_int(crm_element_value(state, PCMK__XA_NODE_FENCED),
&(this_node->details->remote_was_fenced), 0);
return;
}
unpack_transient_attributes(state, this_node, scheduler);
/* Provisionally mark this cluster node as clean. We have at least seen it
* in the current cluster's lifetime.
*/
this_node->details->unclean = FALSE;
this_node->details->unseen = FALSE;
crm_trace("Determining online status of cluster node %s (id %s)",
pcmk__node_name(this_node), id);
determine_online_status(state, this_node, scheduler);
if (!pcmk_is_set(scheduler->flags, pcmk_sched_quorate)
&& this_node->details->online
&& (scheduler->no_quorum_policy == pcmk_no_quorum_fence)) {
/* Everything else should flow from this automatically
* (at least until the scheduler becomes able to migrate off
* healthy resources)
*/
pe_fence_node(scheduler, this_node, "cluster does not have quorum",
FALSE);
}
}
/*!
* \internal
* \brief Unpack nodes' resource history as much as possible
*
* Unpack as many nodes' resource history as possible in one pass through the
* status. We need to process Pacemaker Remote nodes' connections/containers
* before unpacking their history; the connection/container history will be
* in another node's history, so it might take multiple passes to unpack
* everything.
*
* \param[in] status CIB XML status section
* \param[in] fence If true, treat any not-yet-unpacked nodes as unseen
* \param[in,out] scheduler Scheduler data
*
* \return Standard Pacemaker return code (specifically pcmk_rc_ok if done,
* or EAGAIN if more unpacking remains to be done)
*/
static int
unpack_node_history(const xmlNode *status, bool fence,
pcmk_scheduler_t *scheduler)
{
int rc = pcmk_rc_ok;
// Loop through all PCMK__XE_NODE_STATE entries in CIB status
for (const xmlNode *state = first_named_child(status, PCMK__XE_NODE_STATE);
state != NULL; state = crm_next_same_xml(state)) {
const char *id = ID(state);
const char *uname = crm_element_value(state, PCMK_XA_UNAME);
pcmk_node_t *this_node = NULL;
if ((id == NULL) || (uname == NULL)) {
// Warning already logged in first pass through status section
crm_trace("Not unpacking resource history from malformed "
PCMK__XE_NODE_STATE " without id and/or uname");
continue;
}
this_node = pe_find_node_any(scheduler->nodes, id, uname);
if (this_node == NULL) {
// Warning already logged in first pass through status section
crm_trace("Not unpacking resource history for node %s because "
"no longer in configuration", id);
continue;
}
if (this_node->details->unpacked) {
crm_trace("Not unpacking resource history for node %s because "
"already unpacked", id);
continue;
}
if (fence) {
// We're processing all remaining nodes
} else if (pe__is_guest_node(this_node)) {
/* We can unpack a guest node's history only after we've unpacked
* other resource history to the point that we know that the node's
* connection and containing resource are both up.
*/
pcmk_resource_t *rsc = this_node->details->remote_rsc;
if ((rsc == NULL) || (rsc->role != pcmk_role_started)
|| (rsc->container->role != pcmk_role_started)) {
crm_trace("Not unpacking resource history for guest node %s "
"because container and connection are not known to "
"be up", id);
continue;
}
} else if (pe__is_remote_node(this_node)) {
/* We can unpack a remote node's history only after we've unpacked
* other resource history to the point that we know that the node's
* connection is up, with the exception of when shutdown locks are
* in use.
*/
pcmk_resource_t *rsc = this_node->details->remote_rsc;
if ((rsc == NULL)
|| (!pcmk_is_set(scheduler->flags, pcmk_sched_shutdown_lock)
&& (rsc->role != pcmk_role_started))) {
crm_trace("Not unpacking resource history for remote node %s "
"because connection is not known to be up", id);
continue;
}
/* If fencing and shutdown locks are disabled and we're not processing
* unseen nodes, then we don't want to unpack offline nodes until online
* nodes have been unpacked. This allows us to number active clone
* instances first.
*/
} else if (!pcmk_any_flags_set(scheduler->flags,
pcmk_sched_fencing_enabled
|pcmk_sched_shutdown_lock)
&& !this_node->details->online) {
crm_trace("Not unpacking resource history for offline "
"cluster node %s", id);
continue;
}
if (pe__is_guest_or_remote_node(this_node)) {
determine_remote_online_status(scheduler, this_node);
unpack_handle_remote_attrs(this_node, state, scheduler);
}
crm_trace("Unpacking resource history for %snode %s",
(fence? "unseen " : ""), id);
this_node->details->unpacked = TRUE;
unpack_node_lrm(this_node, state, scheduler);
rc = EAGAIN; // Other node histories might depend on this one
}
return rc;
}
/* remove nodes that are down, stopping */
/* create positive rsc_to_node constraints between resources and the nodes they are running on */
/* anything else? */
gboolean
unpack_status(xmlNode *status, pcmk_scheduler_t *scheduler)
{
xmlNode *state = NULL;
crm_trace("Beginning unpack");
if (scheduler->tickets == NULL) {
scheduler->tickets = pcmk__strkey_table(free, destroy_ticket);
}
for (state = pcmk__xe_first_child(status); state != NULL;
state = pcmk__xe_next(state)) {
if (pcmk__str_eq((const char *) state->name, PCMK_XE_TICKETS,
pcmk__str_none)) {
unpack_tickets_state((xmlNode *) state, scheduler);
} else if (pcmk__str_eq((const char *) state->name, PCMK__XE_NODE_STATE,
pcmk__str_none)) {
unpack_node_state(state, scheduler);
}
}
while (unpack_node_history(status, FALSE, scheduler) == EAGAIN) {
crm_trace("Another pass through node resource histories is needed");
}
// Now catch any nodes we didn't see
unpack_node_history(status,
pcmk_is_set(scheduler->flags,
pcmk_sched_fencing_enabled),
scheduler);
/* Now that we know where resources are, we can schedule stops of containers
* with failed bundle connections
*/
if (scheduler->stop_needed != NULL) {
for (GList *item = scheduler->stop_needed; item; item = item->next) {
pcmk_resource_t *container = item->data;
pcmk_node_t *node = pcmk__current_node(container);
if (node) {
stop_action(container, node, FALSE);
}
}
g_list_free(scheduler->stop_needed);
scheduler->stop_needed = NULL;
}
/* Now that we know status of all Pacemaker Remote connections and nodes,
* we can stop connections for node shutdowns, and check the online status
* of remote/guest nodes that didn't have any node history to unpack.
*/
for (GList *gIter = scheduler->nodes; gIter != NULL; gIter = gIter->next) {
pcmk_node_t *this_node = gIter->data;
if (!pe__is_guest_or_remote_node(this_node)) {
continue;
}
if (this_node->details->shutdown
&& (this_node->details->remote_rsc != NULL)) {
pe__set_next_role(this_node->details->remote_rsc, pcmk_role_stopped,
"remote shutdown");
}
if (!this_node->details->unpacked) {
determine_remote_online_status(scheduler, this_node);
}
}
return TRUE;
}
/*!
* \internal
* \brief Unpack node's time when it became a member at the cluster layer
*
* \param[in] node_state Node's \c PCMK__XE_NODE_STATE entry
* \param[in,out] scheduler Scheduler data
*
* \return Epoch time when node became a cluster member
* (or scheduler effective time for legacy entries) if a member,
* 0 if not a member, or -1 if no valid information available
*/
static long long
unpack_node_member(const xmlNode *node_state, pcmk_scheduler_t *scheduler)
{
const char *member_time = crm_element_value(node_state, PCMK__XA_IN_CCM);
int member = 0;
if (member_time == NULL) {
return -1LL;
} else if (crm_str_to_boolean(member_time, &member) == 1) {
/* If in_ccm=0, we'll return 0 here. If in_ccm=1, either the entry was
* recorded as a boolean for a DC < 2.1.7, or the node is pending
* shutdown and has left the CPG, in which case it was set to 1 to avoid
* fencing for PCMK_OPT_NODE_PENDING_TIMEOUT.
*
* We return the effective time for in_ccm=1 because what's important to
* avoid fencing is that effective time minus this value is less than
* the pending node timeout.
*/
return member? (long long) get_effective_time(scheduler) : 0LL;
} else {
long long when_member = 0LL;
if ((pcmk__scan_ll(member_time, &when_member,
0LL) != pcmk_rc_ok) || (when_member < 0LL)) {
crm_warn("Unrecognized value '%s' for " PCMK__XA_IN_CCM
" in " PCMK__XE_NODE_STATE " entry", member_time);
return -1LL;
}
return when_member;
}
}
/*!
* \internal
* \brief Unpack node's time when it became online in process group
*
* \param[in] node_state Node's \c PCMK__XE_NODE_STATE entry
*
* \return Epoch time when node became online in process group (or 0 if not
* online, or 1 for legacy online entries)
*/
static long long
unpack_node_online(const xmlNode *node_state)
{
- const char *peer_time = crm_element_value(node_state, PCMK__XA_CRMD);
+ const char *peer_time = crm_element_value(node_state, PCMK_XA_CRMD);
// @COMPAT Entries recorded for DCs < 2.1.7 have "online" or "offline"
if (pcmk__str_eq(peer_time, OFFLINESTATUS,
pcmk__str_casei|pcmk__str_null_matches)) {
return 0LL;
} else if (pcmk__str_eq(peer_time, ONLINESTATUS, pcmk__str_casei)) {
return 1LL;
} else {
long long when_online = 0LL;
if ((pcmk__scan_ll(peer_time, &when_online, 0LL) != pcmk_rc_ok)
|| (when_online < 0)) {
- crm_warn("Unrecognized value '%s' for " PCMK__XA_CRMD " in "
+ crm_warn("Unrecognized value '%s' for " PCMK_XA_CRMD " in "
PCMK__XE_NODE_STATE " entry, assuming offline", peer_time);
return 0LL;
}
return when_online;
}
}
/*!
* \internal
* \brief Unpack node attribute for user-requested fencing
*
* \param[in] node Node to check
* \param[in] node_state Node's \c PCMK__XE_NODE_STATE entry in CIB status
*
* \return \c true if fencing has been requested for \p node, otherwise \c false
*/
static bool
unpack_node_terminate(const pcmk_node_t *node, const xmlNode *node_state)
{
long long value = 0LL;
int value_i = 0;
const char *value_s = pe_node_attribute_raw(node, PCMK_NODE_ATTR_TERMINATE);
// Value may be boolean or an epoch time
if (crm_str_to_boolean(value_s, &value_i) == 1) {
return (value_i != 0);
}
if (pcmk__scan_ll(value_s, &value, 0LL) == pcmk_rc_ok) {
return (value > 0);
}
crm_warn("Ignoring unrecognized value '%s' for " PCMK_NODE_ATTR_TERMINATE
"node attribute for %s", value_s, pcmk__node_name(node));
return false;
}
static gboolean
determine_online_status_no_fencing(pcmk_scheduler_t *scheduler,
const xmlNode *node_state,
pcmk_node_t *this_node)
{
gboolean online = FALSE;
const char *join = crm_element_value(node_state, PCMK__XA_JOIN);
const char *exp_state = crm_element_value(node_state, PCMK_XA_EXPECTED);
long long when_member = unpack_node_member(node_state, scheduler);
long long when_online = unpack_node_online(node_state);
if (when_member <= 0) {
crm_trace("Node %s is %sdown", pcmk__node_name(this_node),
((when_member < 0)? "presumed " : ""));
} else if (when_online > 0) {
if (pcmk__str_eq(join, CRMD_JOINSTATE_MEMBER, pcmk__str_casei)) {
online = TRUE;
} else {
crm_debug("Node %s is not ready to run resources: %s",
pcmk__node_name(this_node), join);
}
} else if (this_node->details->expected_up == FALSE) {
crm_trace("Node %s controller is down: "
"member@%lld online@%lld join=%s expected=%s",
pcmk__node_name(this_node), when_member, when_online,
pcmk__s(join, "<null>"), pcmk__s(exp_state, "<null>"));
} else {
/* mark it unclean */
pe_fence_node(scheduler, this_node, "peer is unexpectedly down", FALSE);
crm_info("Node %s member@%lld online@%lld join=%s expected=%s",
pcmk__node_name(this_node), when_member, when_online,
pcmk__s(join, "<null>"), pcmk__s(exp_state, "<null>"));
}
return online;
}
/*!
* \internal
* \brief Check whether a node has taken too long to join controller group
*
* \param[in,out] scheduler Scheduler data
* \param[in] node Node to check
* \param[in] when_member Epoch time when node became a cluster member
* \param[in] when_online Epoch time when node joined controller group
*
* \return true if node has been pending (on the way up) longer than
* \c PCMK_OPT_NODE_PENDING_TIMEOUT, otherwise false
* \note This will also update the cluster's recheck time if appropriate.
*/
static inline bool
pending_too_long(pcmk_scheduler_t *scheduler, const pcmk_node_t *node,
long long when_member, long long when_online)
{
if ((scheduler->node_pending_timeout > 0)
&& (when_member > 0) && (when_online <= 0)) {
// There is a timeout on pending nodes, and node is pending
time_t timeout = when_member + scheduler->node_pending_timeout;
if (get_effective_time(node->details->data_set) >= timeout) {
return true; // Node has timed out
}
// Node is pending, but still has time
pe__update_recheck_time(timeout, scheduler, "pending node timeout");
}
return false;
}
static bool
determine_online_status_fencing(pcmk_scheduler_t *scheduler,
const xmlNode *node_state,
pcmk_node_t *this_node)
{
bool termination_requested = unpack_node_terminate(this_node, node_state);
const char *join = crm_element_value(node_state, PCMK__XA_JOIN);
const char *exp_state = crm_element_value(node_state, PCMK_XA_EXPECTED);
long long when_member = unpack_node_member(node_state, scheduler);
long long when_online = unpack_node_online(node_state);
/*
- PCMK__XA_JOIN ::= member|down|pending|banned
- PCMK_XA_EXPECTED ::= member|down
@COMPAT with entries recorded for DCs < 2.1.7
- PCMK__XA_IN_CCM ::= true|false
- - PCMK__XA_CRMD ::= online|offline
+ - PCMK_XA_CRMD ::= online|offline
Since crm_feature_set 3.18.0 (pacemaker-2.1.7):
- PCMK__XA_IN_CCM ::= <timestamp>|0
Since when node has been a cluster member. A value 0 of means the node is not
a cluster member.
- - PCMK__XA_CRMD ::= <timestamp>|0
+ - PCMK_XA_CRMD ::= <timestamp>|0
Since when peer has been online in CPG. A value 0 means the peer is offline
in CPG.
*/
crm_trace("Node %s member@%lld online@%lld join=%s expected=%s%s",
pcmk__node_name(this_node), when_member, when_online,
pcmk__s(join, "<null>"), pcmk__s(exp_state, "<null>"),
(termination_requested? " (termination requested)" : ""));
if (this_node->details->shutdown) {
crm_debug("%s is shutting down", pcmk__node_name(this_node));
/* Slightly different criteria since we can't shut down a dead peer */
return (when_online > 0);
}
if (when_member < 0) {
pe_fence_node(scheduler, this_node,
"peer has not been seen by the cluster", FALSE);
return false;
}
if (pcmk__str_eq(join, CRMD_JOINSTATE_NACK, pcmk__str_none)) {
pe_fence_node(scheduler, this_node,
"peer failed Pacemaker membership criteria", FALSE);
} else if (termination_requested) {
if ((when_member <= 0) && (when_online <= 0)
&& pcmk__str_eq(join, CRMD_JOINSTATE_DOWN, pcmk__str_none)) {
crm_info("%s was fenced as requested", pcmk__node_name(this_node));
return false;
}
pe_fence_node(scheduler, this_node, "fencing was requested", false);
} else if (pcmk__str_eq(exp_state, CRMD_JOINSTATE_DOWN,
pcmk__str_null_matches)) {
if (pending_too_long(scheduler, this_node, when_member, when_online)) {
pe_fence_node(scheduler, this_node,
"peer pending timed out on joining the process group",
FALSE);
} else if ((when_member > 0) || (when_online > 0)) {
crm_info("- %s is not ready to run resources",
pcmk__node_name(this_node));
this_node->details->standby = TRUE;
this_node->details->pending = TRUE;
} else {
crm_trace("%s is down or still coming up",
pcmk__node_name(this_node));
}
} else if (when_member <= 0) {
// Consider PCMK_OPT_PRIORITY_FENCING_DELAY for lost nodes
pe_fence_node(scheduler, this_node,
"peer is no longer part of the cluster", TRUE);
} else if (when_online <= 0) {
pe_fence_node(scheduler, this_node,
"peer process is no longer available", FALSE);
/* Everything is running at this point, now check join state */
} else if (pcmk__str_eq(join, CRMD_JOINSTATE_MEMBER, pcmk__str_none)) {
crm_info("%s is active", pcmk__node_name(this_node));
} else if (pcmk__str_any_of(join, CRMD_JOINSTATE_PENDING,
CRMD_JOINSTATE_DOWN, NULL)) {
crm_info("%s is not ready to run resources",
pcmk__node_name(this_node));
this_node->details->standby = TRUE;
this_node->details->pending = TRUE;
} else {
pe_fence_node(scheduler, this_node, "peer was in an unknown state",
FALSE);
}
return (when_member > 0);
}
static void
determine_remote_online_status(pcmk_scheduler_t *scheduler,
pcmk_node_t *this_node)
{
pcmk_resource_t *rsc = this_node->details->remote_rsc;
pcmk_resource_t *container = NULL;
pcmk_node_t *host = NULL;
/* If there is a node state entry for a (former) Pacemaker Remote node
* but no resource creating that node, the node's connection resource will
* be NULL. Consider it an offline remote node in that case.
*/
if (rsc == NULL) {
this_node->details->online = FALSE;
goto remote_online_done;
}
container = rsc->container;
if (container && pcmk__list_of_1(rsc->running_on)) {
host = rsc->running_on->data;
}
/* If the resource is currently started, mark it online. */
if (rsc->role == pcmk_role_started) {
crm_trace("%s node %s presumed ONLINE because connection resource is started",
(container? "Guest" : "Remote"), this_node->details->id);
this_node->details->online = TRUE;
}
/* consider this node shutting down if transitioning start->stop */
if ((rsc->role == pcmk_role_started)
&& (rsc->next_role == pcmk_role_stopped)) {
crm_trace("%s node %s shutting down because connection resource is stopping",
(container? "Guest" : "Remote"), this_node->details->id);
this_node->details->shutdown = TRUE;
}
/* Now check all the failure conditions. */
if(container && pcmk_is_set(container->flags, pcmk_rsc_failed)) {
crm_trace("Guest node %s UNCLEAN because guest resource failed",
this_node->details->id);
this_node->details->online = FALSE;
this_node->details->remote_requires_reset = TRUE;
} else if (pcmk_is_set(rsc->flags, pcmk_rsc_failed)) {
crm_trace("%s node %s OFFLINE because connection resource failed",
(container? "Guest" : "Remote"), this_node->details->id);
this_node->details->online = FALSE;
} else if ((rsc->role == pcmk_role_stopped)
|| ((container != NULL)
&& (container->role == pcmk_role_stopped))) {
crm_trace("%s node %s OFFLINE because its resource is stopped",
(container? "Guest" : "Remote"), this_node->details->id);
this_node->details->online = FALSE;
this_node->details->remote_requires_reset = FALSE;
} else if (host && (host->details->online == FALSE)
&& host->details->unclean) {
crm_trace("Guest node %s UNCLEAN because host is unclean",
this_node->details->id);
this_node->details->online = FALSE;
this_node->details->remote_requires_reset = TRUE;
}
remote_online_done:
crm_trace("Remote node %s online=%s",
this_node->details->id, this_node->details->online ? "TRUE" : "FALSE");
}
static void
determine_online_status(const xmlNode *node_state, pcmk_node_t *this_node,
pcmk_scheduler_t *scheduler)
{
gboolean online = FALSE;
const char *exp_state = crm_element_value(node_state, PCMK_XA_EXPECTED);
CRM_CHECK(this_node != NULL, return);
this_node->details->shutdown = FALSE;
this_node->details->expected_up = FALSE;
if (pe__shutdown_requested(this_node)) {
this_node->details->shutdown = TRUE;
} else if (pcmk__str_eq(exp_state, CRMD_JOINSTATE_MEMBER, pcmk__str_casei)) {
this_node->details->expected_up = TRUE;
}
if (this_node->details->type == node_ping) {
this_node->details->unclean = FALSE;
online = FALSE; /* As far as resource management is concerned,
* the node is safely offline.
* Anyone caught abusing this logic will be shot
*/
} else if (!pcmk_is_set(scheduler->flags, pcmk_sched_fencing_enabled)) {
online = determine_online_status_no_fencing(scheduler, node_state,
this_node);
} else {
online = determine_online_status_fencing(scheduler, node_state,
this_node);
}
if (online) {
this_node->details->online = TRUE;
} else {
/* remove node from contention */
this_node->fixed = TRUE; // @COMPAT deprecated and unused
this_node->weight = -INFINITY;
}
if (online && this_node->details->shutdown) {
/* don't run resources here */
this_node->fixed = TRUE; // @COMPAT deprecated and unused
this_node->weight = -INFINITY;
}
if (this_node->details->type == node_ping) {
crm_info("%s is not a Pacemaker node", pcmk__node_name(this_node));
} else if (this_node->details->unclean) {
pcmk__sched_warn("%s is unclean", pcmk__node_name(this_node));
} else if (this_node->details->online) {
crm_info("%s is %s", pcmk__node_name(this_node),
this_node->details->shutdown ? "shutting down" :
this_node->details->pending ? "pending" :
this_node->details->standby ? "standby" :
this_node->details->maintenance ? "maintenance" : "online");
} else {
crm_trace("%s is offline", pcmk__node_name(this_node));
}
}
/*!
* \internal
* \brief Find the end of a resource's name, excluding any clone suffix
*
* \param[in] id Resource ID to check
*
* \return Pointer to last character of resource's base name
*/
const char *
pe_base_name_end(const char *id)
{
if (!pcmk__str_empty(id)) {
const char *end = id + strlen(id) - 1;
for (const char *s = end; s > id; --s) {
switch (*s) {
case '0':
case '1':
case '2':
case '3':
case '4':
case '5':
case '6':
case '7':
case '8':
case '9':
break;
case ':':
return (s == end)? s : (s - 1);
default:
return end;
}
}
return end;
}
return NULL;
}
/*!
* \internal
* \brief Get a resource name excluding any clone suffix
*
* \param[in] last_rsc_id Resource ID to check
*
* \return Pointer to newly allocated string with resource's base name
* \note It is the caller's responsibility to free() the result.
* This asserts on error, so callers can assume result is not NULL.
*/
char *
clone_strip(const char *last_rsc_id)
{
const char *end = pe_base_name_end(last_rsc_id);
char *basename = NULL;
CRM_ASSERT(end);
basename = strndup(last_rsc_id, end - last_rsc_id + 1);
CRM_ASSERT(basename);
return basename;
}
/*!
* \internal
* \brief Get the name of the first instance of a cloned resource
*
* \param[in] last_rsc_id Resource ID to check
*
* \return Pointer to newly allocated string with resource's base name plus :0
* \note It is the caller's responsibility to free() the result.
* This asserts on error, so callers can assume result is not NULL.
*/
char *
clone_zero(const char *last_rsc_id)
{
const char *end = pe_base_name_end(last_rsc_id);
size_t base_name_len = end - last_rsc_id + 1;
char *zero = NULL;
CRM_ASSERT(end);
zero = calloc(base_name_len + 3, sizeof(char));
CRM_ASSERT(zero);
memcpy(zero, last_rsc_id, base_name_len);
zero[base_name_len] = ':';
zero[base_name_len + 1] = '0';
return zero;
}
static pcmk_resource_t *
create_fake_resource(const char *rsc_id, const xmlNode *rsc_entry,
pcmk_scheduler_t *scheduler)
{
pcmk_resource_t *rsc = NULL;
xmlNode *xml_rsc = create_xml_node(NULL, PCMK_XE_PRIMITIVE);
copy_in_properties(xml_rsc, rsc_entry);
crm_xml_add(xml_rsc, PCMK_XA_ID, rsc_id);
crm_log_xml_debug(xml_rsc, "Orphan resource");
if (pe__unpack_resource(xml_rsc, &rsc, NULL, scheduler) != pcmk_rc_ok) {
return NULL;
}
if (xml_contains_remote_node(xml_rsc)) {
pcmk_node_t *node;
crm_debug("Detected orphaned remote node %s", rsc_id);
node = pe_find_node(scheduler->nodes, rsc_id);
if (node == NULL) {
node = pe_create_node(rsc_id, rsc_id, PCMK_VALUE_REMOTE, NULL,
scheduler);
}
link_rsc2remotenode(scheduler, rsc);
if (node) {
crm_trace("Setting node %s as shutting down due to orphaned connection resource", rsc_id);
node->details->shutdown = TRUE;
}
}
if (crm_element_value(rsc_entry, PCMK__META_CONTAINER)) {
/* This orphaned rsc needs to be mapped to a container. */
crm_trace("Detected orphaned container filler %s", rsc_id);
pcmk__set_rsc_flags(rsc, pcmk_rsc_removed_filler);
}
pcmk__set_rsc_flags(rsc, pcmk_rsc_removed);
scheduler->resources = g_list_append(scheduler->resources, rsc);
return rsc;
}
/*!
* \internal
* \brief Create orphan instance for anonymous clone resource history
*
* \param[in,out] parent Clone resource that orphan will be added to
* \param[in] rsc_id Orphan's resource ID
* \param[in] node Where orphan is active (for logging only)
* \param[in,out] scheduler Scheduler data
*
* \return Newly added orphaned instance of \p parent
*/
static pcmk_resource_t *
create_anonymous_orphan(pcmk_resource_t *parent, const char *rsc_id,
const pcmk_node_t *node, pcmk_scheduler_t *scheduler)
{
pcmk_resource_t *top = pe__create_clone_child(parent, scheduler);
// find_rsc() because we might be a cloned group
pcmk_resource_t *orphan = top->fns->find_rsc(top, rsc_id, NULL,
pcmk_rsc_match_clone_only);
pcmk__rsc_debug(parent, "Created orphan %s for %s: %s on %s",
top->id, parent->id, rsc_id, pcmk__node_name(node));
return orphan;
}
/*!
* \internal
* \brief Check a node for an instance of an anonymous clone
*
* Return a child instance of the specified anonymous clone, in order of
* preference: (1) the instance running on the specified node, if any;
* (2) an inactive instance (i.e. within the total of \c PCMK_META_CLONE_MAX
* instances); (3) a newly created orphan (that is, \c PCMK_META_CLONE_MAX
* instances are already active).
*
* \param[in,out] scheduler Scheduler data
* \param[in] node Node on which to check for instance
* \param[in,out] parent Clone to check
* \param[in] rsc_id Name of cloned resource in history (no instance)
*/
static pcmk_resource_t *
find_anonymous_clone(pcmk_scheduler_t *scheduler, const pcmk_node_t *node,
pcmk_resource_t *parent, const char *rsc_id)
{
GList *rIter = NULL;
pcmk_resource_t *rsc = NULL;
pcmk_resource_t *inactive_instance = NULL;
gboolean skip_inactive = FALSE;
CRM_ASSERT(parent != NULL);
CRM_ASSERT(pe_rsc_is_clone(parent));
CRM_ASSERT(!pcmk_is_set(parent->flags, pcmk_rsc_unique));
// Check for active (or partially active, for cloned groups) instance
pcmk__rsc_trace(parent, "Looking for %s on %s in %s",
rsc_id, pcmk__node_name(node), parent->id);
for (rIter = parent->children; rsc == NULL && rIter; rIter = rIter->next) {
GList *locations = NULL;
pcmk_resource_t *child = rIter->data;
/* Check whether this instance is already known to be active or pending
* anywhere, at this stage of unpacking. Because this function is called
* for a resource before the resource's individual operation history
* entries are unpacked, locations will generally not contain the
* desired node.
*
* However, there are three exceptions:
* (1) when child is a cloned group and we have already unpacked the
* history of another member of the group on the same node;
* (2) when we've already unpacked the history of another numbered
* instance on the same node (which can happen if
* PCMK_META_GLOBALLY_UNIQUE was flipped from true to false); and
* (3) when we re-run calculations on the same scheduler data as part of
* a simulation.
*/
child->fns->location(child, &locations, 2);
if (locations) {
/* We should never associate the same numbered anonymous clone
* instance with multiple nodes, and clone instances can't migrate,
* so there must be only one location, regardless of history.
*/
CRM_LOG_ASSERT(locations->next == NULL);
if (pcmk__same_node((pcmk_node_t *) locations->data, node)) {
/* This child instance is active on the requested node, so check
* for a corresponding configured resource. We use find_rsc()
* instead of child because child may be a cloned group, and we
* need the particular member corresponding to rsc_id.
*
* If the history entry is orphaned, rsc will be NULL.
*/
rsc = parent->fns->find_rsc(child, rsc_id, NULL,
pcmk_rsc_match_clone_only);
if (rsc) {
/* If there are multiple instance history entries for an
* anonymous clone in a single node's history (which can
* happen if PCMK_META_GLOBALLY_UNIQUE is switched from true
* to false), we want to consider the instances beyond the
* first as orphans, even if there are inactive instance
* numbers available.
*/
if (rsc->running_on) {
crm_notice("Active (now-)anonymous clone %s has "
"multiple (orphan) instance histories on %s",
parent->id, pcmk__node_name(node));
skip_inactive = TRUE;
rsc = NULL;
} else {
pcmk__rsc_trace(parent, "Resource %s, active", rsc->id);
}
}
}
g_list_free(locations);
} else {
pcmk__rsc_trace(parent, "Resource %s, skip inactive", child->id);
if (!skip_inactive && !inactive_instance
&& !pcmk_is_set(child->flags, pcmk_rsc_blocked)) {
// Remember one inactive instance in case we don't find active
inactive_instance = parent->fns->find_rsc(child, rsc_id, NULL,
pcmk_rsc_match_clone_only);
/* ... but don't use it if it was already associated with a
* pending action on another node
*/
if ((inactive_instance != NULL) &&
(inactive_instance->pending_node != NULL) &&
!pcmk__same_node(inactive_instance->pending_node, node)) {
inactive_instance = NULL;
}
}
}
}
if ((rsc == NULL) && !skip_inactive && (inactive_instance != NULL)) {
pcmk__rsc_trace(parent, "Resource %s, empty slot",
inactive_instance->id);
rsc = inactive_instance;
}
/* If the resource has "requires" set to "quorum" or "nothing", and we don't
* have a clone instance for every node, we don't want to consume a valid
* instance number for unclean nodes. Such instances may appear to be active
* according to the history, but should be considered inactive, so we can
* start an instance elsewhere. Treat such instances as orphans.
*
* An exception is instances running on guest nodes -- since guest node
* "fencing" is actually just a resource stop, requires shouldn't apply.
*
* @TODO Ideally, we'd use an inactive instance number if it is not needed
* for any clean instances. However, we don't know that at this point.
*/
if ((rsc != NULL) && !pcmk_is_set(rsc->flags, pcmk_rsc_needs_fencing)
&& (!node->details->online || node->details->unclean)
&& !pe__is_guest_node(node)
&& !pe__is_universal_clone(parent, scheduler)) {
rsc = NULL;
}
if (rsc == NULL) {
rsc = create_anonymous_orphan(parent, rsc_id, node, scheduler);
pcmk__rsc_trace(parent, "Resource %s, orphan", rsc->id);
}
return rsc;
}
static pcmk_resource_t *
unpack_find_resource(pcmk_scheduler_t *scheduler, const pcmk_node_t *node,
const char *rsc_id)
{
pcmk_resource_t *rsc = NULL;
pcmk_resource_t *parent = NULL;
crm_trace("looking for %s", rsc_id);
rsc = pe_find_resource(scheduler->resources, rsc_id);
if (rsc == NULL) {
/* If we didn't find the resource by its name in the operation history,
* check it again as a clone instance. Even when PCMK_META_CLONE_MAX=0,
* we create a single :0 orphan to match against here.
*/
char *clone0_id = clone_zero(rsc_id);
pcmk_resource_t *clone0 = pe_find_resource(scheduler->resources,
clone0_id);
if (clone0 && !pcmk_is_set(clone0->flags, pcmk_rsc_unique)) {
rsc = clone0;
parent = uber_parent(clone0);
crm_trace("%s found as %s (%s)", rsc_id, clone0_id, parent->id);
} else {
crm_trace("%s is not known as %s either (orphan)",
rsc_id, clone0_id);
}
free(clone0_id);
} else if (rsc->variant > pcmk_rsc_variant_primitive) {
crm_trace("Resource history for %s is orphaned because it is no longer primitive",
rsc_id);
return NULL;
} else {
parent = uber_parent(rsc);
}
if (pe_rsc_is_anon_clone(parent)) {
if (pe_rsc_is_bundled(parent)) {
rsc = pe__find_bundle_replica(parent->parent, node);
} else {
char *base = clone_strip(rsc_id);
rsc = find_anonymous_clone(scheduler, node, parent, base);
free(base);
CRM_ASSERT(rsc != NULL);
}
}
if (rsc && !pcmk__str_eq(rsc_id, rsc->id, pcmk__str_casei)
&& !pcmk__str_eq(rsc_id, rsc->clone_name, pcmk__str_casei)) {
pcmk__str_update(&rsc->clone_name, rsc_id);
pcmk__rsc_debug(rsc, "Internally renamed %s on %s to %s%s",
rsc_id, pcmk__node_name(node), rsc->id,
pcmk_is_set(rsc->flags, pcmk_rsc_removed)? " (ORPHAN)" : "");
}
return rsc;
}
static pcmk_resource_t *
process_orphan_resource(const xmlNode *rsc_entry, const pcmk_node_t *node,
pcmk_scheduler_t *scheduler)
{
pcmk_resource_t *rsc = NULL;
const char *rsc_id = crm_element_value(rsc_entry, PCMK_XA_ID);
crm_debug("Detected orphan resource %s on %s",
rsc_id, pcmk__node_name(node));
rsc = create_fake_resource(rsc_id, rsc_entry, scheduler);
if (rsc == NULL) {
return NULL;
}
if (!pcmk_is_set(scheduler->flags, pcmk_sched_stop_removed_resources)) {
pcmk__clear_rsc_flags(rsc, pcmk_rsc_managed);
} else {
CRM_CHECK(rsc != NULL, return NULL);
pcmk__rsc_trace(rsc, "Added orphan %s", rsc->id);
resource_location(rsc, NULL, -INFINITY, "__orphan_do_not_run__",
scheduler);
}
return rsc;
}
static void
process_rsc_state(pcmk_resource_t *rsc, pcmk_node_t *node,
enum action_fail_response on_fail)
{
pcmk_node_t *tmpnode = NULL;
char *reason = NULL;
enum action_fail_response save_on_fail = pcmk_on_fail_ignore;
CRM_ASSERT(rsc);
pcmk__rsc_trace(rsc, "Resource %s is %s on %s: on_fail=%s",
rsc->id, pcmk_role_text(rsc->role), pcmk__node_name(node),
fail2text(on_fail));
/* process current state */
if (rsc->role != pcmk_role_unknown) {
pcmk_resource_t *iter = rsc;
while (iter) {
if (g_hash_table_lookup(iter->known_on, node->details->id) == NULL) {
pcmk_node_t *n = pe__copy_node(node);
pcmk__rsc_trace(rsc, "%s%s%s known on %s",
rsc->id,
((rsc->clone_name == NULL)? "" : " also known as "),
((rsc->clone_name == NULL)? "" : rsc->clone_name),
pcmk__node_name(n));
g_hash_table_insert(iter->known_on, (gpointer) n->details->id, n);
}
if (pcmk_is_set(iter->flags, pcmk_rsc_unique)) {
break;
}
iter = iter->parent;
}
}
/* If a managed resource is believed to be running, but node is down ... */
if ((rsc->role > pcmk_role_stopped)
&& node->details->online == FALSE
&& node->details->maintenance == FALSE
&& pcmk_is_set(rsc->flags, pcmk_rsc_managed)) {
gboolean should_fence = FALSE;
/* If this is a guest node, fence it (regardless of whether fencing is
* enabled, because guest node fencing is done by recovery of the
* container resource rather than by the fencer). Mark the resource
* we're processing as failed. When the guest comes back up, its
* operation history in the CIB will be cleared, freeing the affected
* resource to run again once we are sure we know its state.
*/
if (pe__is_guest_node(node)) {
pcmk__set_rsc_flags(rsc, pcmk_rsc_failed|pcmk_rsc_stop_if_failed);
should_fence = TRUE;
} else if (pcmk_is_set(rsc->cluster->flags,
pcmk_sched_fencing_enabled)) {
if (pe__is_remote_node(node) && node->details->remote_rsc
&& !pcmk_is_set(node->details->remote_rsc->flags,
pcmk_rsc_failed)) {
/* Setting unseen means that fencing of the remote node will
* occur only if the connection resource is not going to start
* somewhere. This allows connection resources on a failed
* cluster node to move to another node without requiring the
* remote nodes to be fenced as well.
*/
node->details->unseen = TRUE;
reason = crm_strdup_printf("%s is active there (fencing will be"
" revoked if remote connection can "
"be re-established elsewhere)",
rsc->id);
}
should_fence = TRUE;
}
if (should_fence) {
if (reason == NULL) {
reason = crm_strdup_printf("%s is thought to be active there", rsc->id);
}
pe_fence_node(rsc->cluster, node, reason, FALSE);
}
free(reason);
}
/* In order to calculate priority_fencing_delay correctly, save the failure information and pass it to native_add_running(). */
save_on_fail = on_fail;
if (node->details->unclean) {
/* No extra processing needed
* Also allows resources to be started again after a node is shot
*/
on_fail = pcmk_on_fail_ignore;
}
switch (on_fail) {
case pcmk_on_fail_ignore:
/* nothing to do */
break;
case pcmk_on_fail_demote:
pcmk__set_rsc_flags(rsc, pcmk_rsc_failed);
demote_action(rsc, node, FALSE);
break;
case pcmk_on_fail_fence_node:
/* treat it as if it is still running
* but also mark the node as unclean
*/
reason = crm_strdup_printf("%s failed there", rsc->id);
pe_fence_node(rsc->cluster, node, reason, FALSE);
free(reason);
break;
case pcmk_on_fail_standby_node:
node->details->standby = TRUE;
node->details->standby_onfail = TRUE;
break;
case pcmk_on_fail_block:
/* is_managed == FALSE will prevent any
* actions being sent for the resource
*/
pcmk__clear_rsc_flags(rsc, pcmk_rsc_managed);
pcmk__set_rsc_flags(rsc, pcmk_rsc_blocked);
break;
case pcmk_on_fail_ban:
/* make sure it comes up somewhere else
* or not at all
*/
resource_location(rsc, node, -INFINITY, "__action_migration_auto__",
rsc->cluster);
break;
case pcmk_on_fail_stop:
pe__set_next_role(rsc, pcmk_role_stopped,
PCMK_META_ON_FAIL "=" PCMK_VALUE_STOP);
break;
case pcmk_on_fail_restart:
if ((rsc->role != pcmk_role_stopped)
&& (rsc->role != pcmk_role_unknown)) {
pcmk__set_rsc_flags(rsc,
pcmk_rsc_failed|pcmk_rsc_stop_if_failed);
stop_action(rsc, node, FALSE);
}
break;
case pcmk_on_fail_restart_container:
pcmk__set_rsc_flags(rsc, pcmk_rsc_failed|pcmk_rsc_stop_if_failed);
if (rsc->container && pe_rsc_is_bundled(rsc)) {
/* A bundle's remote connection can run on a different node than
* the bundle's container. We don't necessarily know where the
* container is running yet, so remember it and add a stop
* action for it later.
*/
rsc->cluster->stop_needed =
g_list_prepend(rsc->cluster->stop_needed, rsc->container);
} else if (rsc->container) {
stop_action(rsc->container, node, FALSE);
} else if ((rsc->role != pcmk_role_stopped)
&& (rsc->role != pcmk_role_unknown)) {
stop_action(rsc, node, FALSE);
}
break;
case pcmk_on_fail_reset_remote:
pcmk__set_rsc_flags(rsc, pcmk_rsc_failed|pcmk_rsc_stop_if_failed);
if (pcmk_is_set(rsc->cluster->flags, pcmk_sched_fencing_enabled)) {
tmpnode = NULL;
if (rsc->is_remote_node) {
tmpnode = pe_find_node(rsc->cluster->nodes, rsc->id);
}
if (tmpnode &&
pe__is_remote_node(tmpnode) &&
tmpnode->details->remote_was_fenced == 0) {
/* The remote connection resource failed in a way that
* should result in fencing the remote node.
*/
pe_fence_node(rsc->cluster, tmpnode,
"remote connection is unrecoverable", FALSE);
}
}
/* require the stop action regardless if fencing is occurring or not. */
if (rsc->role > pcmk_role_stopped) {
stop_action(rsc, node, FALSE);
}
/* if reconnect delay is in use, prevent the connection from exiting the
* "STOPPED" role until the failure is cleared by the delay timeout. */
if (rsc->remote_reconnect_ms) {
pe__set_next_role(rsc, pcmk_role_stopped, "remote reset");
}
break;
}
/* ensure a remote-node connection failure forces an unclean remote-node
* to be fenced. By setting unseen = FALSE, the remote-node failure will
* result in a fencing operation regardless if we're going to attempt to
* reconnect to the remote-node in this transition or not. */
if (pcmk_is_set(rsc->flags, pcmk_rsc_failed) && rsc->is_remote_node) {
tmpnode = pe_find_node(rsc->cluster->nodes, rsc->id);
if (tmpnode && tmpnode->details->unclean) {
tmpnode->details->unseen = FALSE;
}
}
if ((rsc->role != pcmk_role_stopped)
&& (rsc->role != pcmk_role_unknown)) {
if (pcmk_is_set(rsc->flags, pcmk_rsc_removed)) {
if (pcmk_is_set(rsc->flags, pcmk_rsc_managed)) {
pcmk__config_warn("Detected active orphan %s running on %s",
rsc->id, pcmk__node_name(node));
} else {
pcmk__config_warn("Resource '%s' must be stopped manually on "
"%s because cluster is configured not to "
"stop active orphans",
rsc->id, pcmk__node_name(node));
}
}
native_add_running(rsc, node, rsc->cluster,
(save_on_fail != pcmk_on_fail_ignore));
switch (on_fail) {
case pcmk_on_fail_ignore:
break;
case pcmk_on_fail_demote:
case pcmk_on_fail_block:
pcmk__set_rsc_flags(rsc, pcmk_rsc_failed);
break;
default:
pcmk__set_rsc_flags(rsc,
pcmk_rsc_failed|pcmk_rsc_stop_if_failed);
break;
}
} else if (rsc->clone_name && strchr(rsc->clone_name, ':') != NULL) {
/* Only do this for older status sections that included instance numbers
* Otherwise stopped instances will appear as orphans
*/
pcmk__rsc_trace(rsc, "Resetting clone_name %s for %s (stopped)",
rsc->clone_name, rsc->id);
free(rsc->clone_name);
rsc->clone_name = NULL;
} else {
GList *possible_matches = pe__resource_actions(rsc, node,
PCMK_ACTION_STOP, FALSE);
GList *gIter = possible_matches;
for (; gIter != NULL; gIter = gIter->next) {
pcmk_action_t *stop = (pcmk_action_t *) gIter->data;
pcmk__set_action_flags(stop, pcmk_action_optional);
}
g_list_free(possible_matches);
}
/* A successful stop after migrate_to on the migration source doesn't make
* the partially migrated resource stopped on the migration target.
*/
if ((rsc->role == pcmk_role_stopped)
&& rsc->partial_migration_source
&& rsc->partial_migration_source->details == node->details
&& rsc->partial_migration_target
&& rsc->running_on) {
rsc->role = pcmk_role_started;
}
}
/* create active recurring operations as optional */
static void
process_recurring(pcmk_node_t *node, pcmk_resource_t *rsc,
int start_index, int stop_index,
GList *sorted_op_list, pcmk_scheduler_t *scheduler)
{
int counter = -1;
const char *task = NULL;
const char *status = NULL;
GList *gIter = sorted_op_list;
CRM_ASSERT(rsc);
pcmk__rsc_trace(rsc, "%s: Start index %d, stop index = %d",
rsc->id, start_index, stop_index);
for (; gIter != NULL; gIter = gIter->next) {
xmlNode *rsc_op = (xmlNode *) gIter->data;
guint interval_ms = 0;
char *key = NULL;
const char *id = ID(rsc_op);
counter++;
if (node->details->online == FALSE) {
pcmk__rsc_trace(rsc, "Skipping %s on %s: node is offline",
rsc->id, pcmk__node_name(node));
break;
/* Need to check if there's a monitor for role="Stopped" */
} else if (start_index < stop_index && counter <= stop_index) {
pcmk__rsc_trace(rsc, "Skipping %s on %s: resource is not active",
id, pcmk__node_name(node));
continue;
} else if (counter < start_index) {
pcmk__rsc_trace(rsc, "Skipping %s on %s: old %d",
id, pcmk__node_name(node), counter);
continue;
}
crm_element_value_ms(rsc_op, PCMK_META_INTERVAL, &interval_ms);
if (interval_ms == 0) {
pcmk__rsc_trace(rsc, "Skipping %s on %s: non-recurring",
id, pcmk__node_name(node));
continue;
}
status = crm_element_value(rsc_op, PCMK__XA_OP_STATUS);
if (pcmk__str_eq(status, "-1", pcmk__str_casei)) {
pcmk__rsc_trace(rsc, "Skipping %s on %s: status",
id, pcmk__node_name(node));
continue;
}
task = crm_element_value(rsc_op, PCMK_XA_OPERATION);
/* create the action */
key = pcmk__op_key(rsc->id, task, interval_ms);
pcmk__rsc_trace(rsc, "Creating %s on %s", key, pcmk__node_name(node));
custom_action(rsc, key, task, node, TRUE, scheduler);
}
}
void
calculate_active_ops(const GList *sorted_op_list, int *start_index,
int *stop_index)
{
int counter = -1;
int implied_monitor_start = -1;
int implied_clone_start = -1;
const char *task = NULL;
const char *status = NULL;
*stop_index = -1;
*start_index = -1;
for (const GList *iter = sorted_op_list; iter != NULL; iter = iter->next) {
const xmlNode *rsc_op = (const xmlNode *) iter->data;
counter++;
task = crm_element_value(rsc_op, PCMK_XA_OPERATION);
status = crm_element_value(rsc_op, PCMK__XA_OP_STATUS);
if (pcmk__str_eq(task, PCMK_ACTION_STOP, pcmk__str_casei)
&& pcmk__str_eq(status, "0", pcmk__str_casei)) {
*stop_index = counter;
} else if (pcmk__strcase_any_of(task, PCMK_ACTION_START,
PCMK_ACTION_MIGRATE_FROM, NULL)) {
*start_index = counter;
} else if ((implied_monitor_start <= *stop_index)
&& pcmk__str_eq(task, PCMK_ACTION_MONITOR,
pcmk__str_casei)) {
const char *rc = crm_element_value(rsc_op, PCMK__XA_RC_CODE);
if (pcmk__strcase_any_of(rc, "0", "8", NULL)) {
implied_monitor_start = counter;
}
} else if (pcmk__strcase_any_of(task, PCMK_ACTION_PROMOTE,
PCMK_ACTION_DEMOTE, NULL)) {
implied_clone_start = counter;
}
}
if (*start_index == -1) {
if (implied_clone_start != -1) {
*start_index = implied_clone_start;
} else if (implied_monitor_start != -1) {
*start_index = implied_monitor_start;
}
}
}
// If resource history entry has shutdown lock, remember lock node and time
static void
unpack_shutdown_lock(const xmlNode *rsc_entry, pcmk_resource_t *rsc,
const pcmk_node_t *node, pcmk_scheduler_t *scheduler)
{
time_t lock_time = 0; // When lock started (i.e. node shutdown time)
if ((crm_element_value_epoch(rsc_entry, PCMK_OPT_SHUTDOWN_LOCK,
&lock_time) == pcmk_ok) && (lock_time != 0)) {
if ((scheduler->shutdown_lock > 0)
&& (get_effective_time(scheduler)
> (lock_time + scheduler->shutdown_lock))) {
pcmk__rsc_info(rsc, "Shutdown lock for %s on %s expired",
rsc->id, pcmk__node_name(node));
pe__clear_resource_history(rsc, node);
} else {
/* @COMPAT I don't like breaking const signatures, but
* rsc->lock_node should really be const -- we just can't change it
* until the next API compatibility break.
*/
rsc->lock_node = (pcmk_node_t *) node;
rsc->lock_time = lock_time;
}
}
}
/*!
* \internal
* \brief Unpack one \c PCMK__XE_LRM_RESOURCE entry from a node's CIB status
*
* \param[in,out] node Node whose status is being unpacked
* \param[in] rsc_entry \c PCMK__XE_LRM_RESOURCE XML being unpacked
* \param[in,out] scheduler Scheduler data
*
* \return Resource corresponding to the entry, or NULL if no operation history
*/
static pcmk_resource_t *
unpack_lrm_resource(pcmk_node_t *node, const xmlNode *lrm_resource,
pcmk_scheduler_t *scheduler)
{
GList *gIter = NULL;
int stop_index = -1;
int start_index = -1;
enum rsc_role_e req_role = pcmk_role_unknown;
const char *rsc_id = ID(lrm_resource);
pcmk_resource_t *rsc = NULL;
GList *op_list = NULL;
GList *sorted_op_list = NULL;
xmlNode *rsc_op = NULL;
xmlNode *last_failure = NULL;
enum action_fail_response on_fail = pcmk_on_fail_ignore;
enum rsc_role_e saved_role = pcmk_role_unknown;
if (rsc_id == NULL) {
pcmk__config_err("Ignoring invalid " PCMK__XE_LRM_RESOURCE
" entry: No " PCMK_XA_ID);
crm_log_xml_info(lrm_resource, "missing-id");
return NULL;
}
crm_trace("Unpacking " PCMK__XE_LRM_RESOURCE " for %s on %s",
rsc_id, pcmk__node_name(node));
/* Build a list of individual PCMK__XE_LRM_RSC_OP entries, so we can sort
* them
*/
for (rsc_op = first_named_child(lrm_resource, PCMK__XE_LRM_RSC_OP);
rsc_op != NULL; rsc_op = crm_next_same_xml(rsc_op)) {
op_list = g_list_prepend(op_list, rsc_op);
}
if (!pcmk_is_set(scheduler->flags, pcmk_sched_shutdown_lock)) {
if (op_list == NULL) {
// If there are no operations, there is nothing to do
return NULL;
}
}
/* find the resource */
rsc = unpack_find_resource(scheduler, node, rsc_id);
if (rsc == NULL) {
if (op_list == NULL) {
// If there are no operations, there is nothing to do
return NULL;
} else {
rsc = process_orphan_resource(lrm_resource, node, scheduler);
}
}
CRM_ASSERT(rsc != NULL);
// Check whether the resource is "shutdown-locked" to this node
if (pcmk_is_set(scheduler->flags, pcmk_sched_shutdown_lock)) {
unpack_shutdown_lock(lrm_resource, rsc, node, scheduler);
}
/* process operations */
saved_role = rsc->role;
rsc->role = pcmk_role_unknown;
sorted_op_list = g_list_sort(op_list, sort_op_by_callid);
for (gIter = sorted_op_list; gIter != NULL; gIter = gIter->next) {
xmlNode *rsc_op = (xmlNode *) gIter->data;
unpack_rsc_op(rsc, node, rsc_op, &last_failure, &on_fail);
}
/* create active recurring operations as optional */
calculate_active_ops(sorted_op_list, &start_index, &stop_index);
process_recurring(node, rsc, start_index, stop_index, sorted_op_list,
scheduler);
/* no need to free the contents */
g_list_free(sorted_op_list);
process_rsc_state(rsc, node, on_fail);
if (get_target_role(rsc, &req_role)) {
if ((rsc->next_role == pcmk_role_unknown)
|| (req_role < rsc->next_role)) {
pe__set_next_role(rsc, req_role, PCMK_META_TARGET_ROLE);
} else if (req_role > rsc->next_role) {
pcmk__rsc_info(rsc,
"%s: Not overwriting calculated next role %s"
" with requested next role %s",
rsc->id, pcmk_role_text(rsc->next_role),
pcmk_role_text(req_role));
}
}
if (saved_role > rsc->role) {
rsc->role = saved_role;
}
return rsc;
}
static void
handle_orphaned_container_fillers(const xmlNode *lrm_rsc_list,
pcmk_scheduler_t *scheduler)
{
for (const xmlNode *rsc_entry = pcmk__xe_first_child(lrm_rsc_list);
rsc_entry != NULL; rsc_entry = pcmk__xe_next(rsc_entry)) {
pcmk_resource_t *rsc;
pcmk_resource_t *container;
const char *rsc_id;
const char *container_id;
if (!pcmk__str_eq((const char *) rsc_entry->name, PCMK__XE_LRM_RESOURCE,
pcmk__str_casei)) {
continue;
}
container_id = crm_element_value(rsc_entry, PCMK__META_CONTAINER);
rsc_id = crm_element_value(rsc_entry, PCMK_XA_ID);
if (container_id == NULL || rsc_id == NULL) {
continue;
}
container = pe_find_resource(scheduler->resources, container_id);
if (container == NULL) {
continue;
}
rsc = pe_find_resource(scheduler->resources, rsc_id);
if ((rsc == NULL) || (rsc->container != NULL)
|| !pcmk_is_set(rsc->flags, pcmk_rsc_removed_filler)) {
continue;
}
pcmk__rsc_trace(rsc, "Mapped container of orphaned resource %s to %s",
rsc->id, container_id);
rsc->container = container;
container->fillers = g_list_append(container->fillers, rsc);
}
}
/*!
* \internal
* \brief Unpack one node's lrm status section
*
* \param[in,out] node Node whose status is being unpacked
* \param[in] xml CIB node state XML
* \param[in,out] scheduler Scheduler data
*/
static void
unpack_node_lrm(pcmk_node_t *node, const xmlNode *xml,
pcmk_scheduler_t *scheduler)
{
bool found_orphaned_container_filler = false;
// Drill down to PCMK__XE_LRM_RESOURCES section
xml = find_xml_node(xml, PCMK__XE_LRM, FALSE);
if (xml == NULL) {
return;
}
xml = find_xml_node(xml, PCMK__XE_LRM_RESOURCES, FALSE);
if (xml == NULL) {
return;
}
// Unpack each PCMK__XE_LRM_RESOURCE entry
for (const xmlNode *rsc_entry = first_named_child(xml,
PCMK__XE_LRM_RESOURCE);
rsc_entry != NULL; rsc_entry = crm_next_same_xml(rsc_entry)) {
pcmk_resource_t *rsc = unpack_lrm_resource(node, rsc_entry, scheduler);
if ((rsc != NULL)
&& pcmk_is_set(rsc->flags, pcmk_rsc_removed_filler)) {
found_orphaned_container_filler = true;
}
}
/* Now that all resource state has been unpacked for this node, map any
* orphaned container fillers to their container resource.
*/
if (found_orphaned_container_filler) {
handle_orphaned_container_fillers(xml, scheduler);
}
}
static void
set_active(pcmk_resource_t *rsc)
{
const pcmk_resource_t *top = pe__const_top_resource(rsc, false);
if (top && pcmk_is_set(top->flags, pcmk_rsc_promotable)) {
rsc->role = pcmk_role_unpromoted;
} else {
rsc->role = pcmk_role_started;
}
}
static void
set_node_score(gpointer key, gpointer value, gpointer user_data)
{
pcmk_node_t *node = value;
int *score = user_data;
node->weight = *score;
}
#define XPATH_NODE_STATE "/" PCMK_XE_CIB "/" PCMK_XE_STATUS \
"/" PCMK__XE_NODE_STATE
#define SUB_XPATH_LRM_RESOURCE "/" PCMK__XE_LRM \
"/" PCMK__XE_LRM_RESOURCES \
"/" PCMK__XE_LRM_RESOURCE
#define SUB_XPATH_LRM_RSC_OP "/" PCMK__XE_LRM_RSC_OP
static xmlNode *
find_lrm_op(const char *resource, const char *op, const char *node, const char *source,
int target_rc, pcmk_scheduler_t *scheduler)
{
GString *xpath = NULL;
xmlNode *xml = NULL;
CRM_CHECK((resource != NULL) && (op != NULL) && (node != NULL),
return NULL);
xpath = g_string_sized_new(256);
pcmk__g_strcat(xpath,
XPATH_NODE_STATE "[@" PCMK_XA_UNAME "='", node, "']"
SUB_XPATH_LRM_RESOURCE "[@" PCMK_XA_ID "='", resource, "']"
SUB_XPATH_LRM_RSC_OP "[@" PCMK_XA_OPERATION "='", op, "'",
NULL);
/* Need to check against transition_magic too? */
if ((source != NULL) && (strcmp(op, PCMK_ACTION_MIGRATE_TO) == 0)) {
pcmk__g_strcat(xpath,
" and @" PCMK__META_MIGRATE_TARGET "='", source, "']",
NULL);
} else if ((source != NULL)
&& (strcmp(op, PCMK_ACTION_MIGRATE_FROM) == 0)) {
pcmk__g_strcat(xpath,
" and @" PCMK__META_MIGRATE_SOURCE "='", source, "']",
NULL);
} else {
g_string_append_c(xpath, ']');
}
xml = get_xpath_object((const char *) xpath->str, scheduler->input,
LOG_DEBUG);
g_string_free(xpath, TRUE);
if (xml && target_rc >= 0) {
int rc = PCMK_OCF_UNKNOWN_ERROR;
int status = PCMK_EXEC_ERROR;
crm_element_value_int(xml, PCMK__XA_RC_CODE, &rc);
crm_element_value_int(xml, PCMK__XA_OP_STATUS, &status);
if ((rc != target_rc) || (status != PCMK_EXEC_DONE)) {
return NULL;
}
}
return xml;
}
static xmlNode *
find_lrm_resource(const char *rsc_id, const char *node_name,
pcmk_scheduler_t *scheduler)
{
GString *xpath = NULL;
xmlNode *xml = NULL;
CRM_CHECK((rsc_id != NULL) && (node_name != NULL), return NULL);
xpath = g_string_sized_new(256);
pcmk__g_strcat(xpath,
XPATH_NODE_STATE "[@" PCMK_XA_UNAME "='", node_name, "']"
SUB_XPATH_LRM_RESOURCE "[@" PCMK_XA_ID "='", rsc_id, "']",
NULL);
xml = get_xpath_object((const char *) xpath->str, scheduler->input,
LOG_DEBUG);
g_string_free(xpath, TRUE);
return xml;
}
/*!
* \internal
* \brief Check whether a resource has no completed action history on a node
*
* \param[in,out] rsc Resource to check
* \param[in] node_name Node to check
*
* \return true if \p rsc_id is unknown on \p node_name, otherwise false
*/
static bool
unknown_on_node(pcmk_resource_t *rsc, const char *node_name)
{
bool result = false;
xmlXPathObjectPtr search;
char *xpath = NULL;
xpath = crm_strdup_printf(XPATH_NODE_STATE "[@" PCMK_XA_UNAME "='%s']"
SUB_XPATH_LRM_RESOURCE "[@" PCMK_XA_ID "='%s']"
SUB_XPATH_LRM_RSC_OP
"[@" PCMK__XA_RC_CODE "!='%d']",
node_name, rsc->id, PCMK_OCF_UNKNOWN);
search = xpath_search(rsc->cluster->input, xpath);
result = (numXpathResults(search) == 0);
freeXpathObject(search);
free(xpath);
return result;
}
/*!
* \brief Check whether a probe/monitor indicating the resource was not running
* on a node happened after some event
*
* \param[in] rsc_id Resource being checked
* \param[in] node_name Node being checked
* \param[in] xml_op Event that monitor is being compared to
* \param[in] same_node Whether the operations are on the same node
* \param[in,out] scheduler Scheduler data
*
* \return true if such a monitor happened after event, false otherwise
*/
static bool
monitor_not_running_after(const char *rsc_id, const char *node_name,
const xmlNode *xml_op, bool same_node,
pcmk_scheduler_t *scheduler)
{
/* Any probe/monitor operation on the node indicating it was not running
* there
*/
xmlNode *monitor = find_lrm_op(rsc_id, PCMK_ACTION_MONITOR, node_name,
NULL, PCMK_OCF_NOT_RUNNING, scheduler);
return (monitor && pe__is_newer_op(monitor, xml_op, same_node) > 0);
}
/*!
* \brief Check whether any non-monitor operation on a node happened after some
* event
*
* \param[in] rsc_id Resource being checked
* \param[in] node_name Node being checked
* \param[in] xml_op Event that non-monitor is being compared to
* \param[in] same_node Whether the operations are on the same node
* \param[in,out] scheduler Scheduler data
*
* \return true if such a operation happened after event, false otherwise
*/
static bool
non_monitor_after(const char *rsc_id, const char *node_name,
const xmlNode *xml_op, bool same_node,
pcmk_scheduler_t *scheduler)
{
xmlNode *lrm_resource = NULL;
lrm_resource = find_lrm_resource(rsc_id, node_name, scheduler);
if (lrm_resource == NULL) {
return false;
}
for (xmlNode *op = first_named_child(lrm_resource, PCMK__XE_LRM_RSC_OP);
op != NULL; op = crm_next_same_xml(op)) {
const char * task = NULL;
if (op == xml_op) {
continue;
}
task = crm_element_value(op, PCMK_XA_OPERATION);
if (pcmk__str_any_of(task, PCMK_ACTION_START, PCMK_ACTION_STOP,
PCMK_ACTION_MIGRATE_TO, PCMK_ACTION_MIGRATE_FROM,
NULL)
&& pe__is_newer_op(op, xml_op, same_node) > 0) {
return true;
}
}
return false;
}
/*!
* \brief Check whether the resource has newer state on a node after a migration
* attempt
*
* \param[in] rsc_id Resource being checked
* \param[in] node_name Node being checked
* \param[in] migrate_to Any migrate_to event that is being compared to
* \param[in] migrate_from Any migrate_from event that is being compared to
* \param[in,out] scheduler Scheduler data
*
* \return true if such a operation happened after event, false otherwise
*/
static bool
newer_state_after_migrate(const char *rsc_id, const char *node_name,
const xmlNode *migrate_to,
const xmlNode *migrate_from,
pcmk_scheduler_t *scheduler)
{
const xmlNode *xml_op = migrate_to;
const char *source = NULL;
const char *target = NULL;
bool same_node = false;
if (migrate_from) {
xml_op = migrate_from;
}
source = crm_element_value(xml_op, PCMK__META_MIGRATE_SOURCE);
target = crm_element_value(xml_op, PCMK__META_MIGRATE_TARGET);
/* It's preferred to compare to the migrate event on the same node if
* existing, since call ids are more reliable.
*/
if (pcmk__str_eq(node_name, target, pcmk__str_casei)) {
if (migrate_from) {
xml_op = migrate_from;
same_node = true;
} else {
xml_op = migrate_to;
}
} else if (pcmk__str_eq(node_name, source, pcmk__str_casei)) {
if (migrate_to) {
xml_op = migrate_to;
same_node = true;
} else {
xml_op = migrate_from;
}
}
/* If there's any newer non-monitor operation on the node, or any newer
* probe/monitor operation on the node indicating it was not running there,
* the migration events potentially no longer matter for the node.
*/
return non_monitor_after(rsc_id, node_name, xml_op, same_node, scheduler)
|| monitor_not_running_after(rsc_id, node_name, xml_op, same_node,
scheduler);
}
/*!
* \internal
* \brief Parse migration source and target node names from history entry
*
* \param[in] entry Resource history entry for a migration action
* \param[in] source_node If not NULL, source must match this node
* \param[in] target_node If not NULL, target must match this node
* \param[out] source_name Where to store migration source node name
* \param[out] target_name Where to store migration target node name
*
* \return Standard Pacemaker return code
*/
static int
get_migration_node_names(const xmlNode *entry, const pcmk_node_t *source_node,
const pcmk_node_t *target_node,
const char **source_name, const char **target_name)
{
*source_name = crm_element_value(entry, PCMK__META_MIGRATE_SOURCE);
*target_name = crm_element_value(entry, PCMK__META_MIGRATE_TARGET);
if ((*source_name == NULL) || (*target_name == NULL)) {
pcmk__config_err("Ignoring resource history entry %s without "
PCMK__META_MIGRATE_SOURCE " and "
PCMK__META_MIGRATE_TARGET, ID(entry));
return pcmk_rc_unpack_error;
}
if ((source_node != NULL)
&& !pcmk__str_eq(*source_name, source_node->details->uname,
pcmk__str_casei|pcmk__str_null_matches)) {
pcmk__config_err("Ignoring resource history entry %s because "
PCMK__META_MIGRATE_SOURCE "='%s' does not match %s",
ID(entry), *source_name, pcmk__node_name(source_node));
return pcmk_rc_unpack_error;
}
if ((target_node != NULL)
&& !pcmk__str_eq(*target_name, target_node->details->uname,
pcmk__str_casei|pcmk__str_null_matches)) {
pcmk__config_err("Ignoring resource history entry %s because "
PCMK__META_MIGRATE_TARGET "='%s' does not match %s",
ID(entry), *target_name, pcmk__node_name(target_node));
return pcmk_rc_unpack_error;
}
return pcmk_rc_ok;
}
/*
* \internal
* \brief Add a migration source to a resource's list of dangling migrations
*
* If the migrate_to and migrate_from actions in a live migration both
* succeeded, but there is no stop on the source, the migration is considered
* "dangling." Add the source to the resource's dangling migration list, which
* will be used to schedule a stop on the source without affecting the target.
*
* \param[in,out] rsc Resource involved in migration
* \param[in] node Migration source
*/
static void
add_dangling_migration(pcmk_resource_t *rsc, const pcmk_node_t *node)
{
pcmk__rsc_trace(rsc, "Dangling migration of %s requires stop on %s",
rsc->id, pcmk__node_name(node));
rsc->role = pcmk_role_stopped;
rsc->dangling_migrations = g_list_prepend(rsc->dangling_migrations,
(gpointer) node);
}
/*!
* \internal
* \brief Update resource role etc. after a successful migrate_to action
*
* \param[in,out] history Parsed action result history
*/
static void
unpack_migrate_to_success(struct action_history *history)
{
/* A complete migration sequence is:
* 1. migrate_to on source node (which succeeded if we get to this function)
* 2. migrate_from on target node
* 3. stop on source node
*
* If no migrate_from has happened, the migration is considered to be
* "partial". If the migrate_from succeeded but no stop has happened, the
* migration is considered to be "dangling".
*
* If a successful migrate_to and stop have happened on the source node, we
* still need to check for a partial migration, due to scenarios (easier to
* produce with batch-limit=1) like:
*
* - A resource is migrating from node1 to node2, and a migrate_to is
* initiated for it on node1.
*
* - node2 goes into standby mode while the migrate_to is pending, which
* aborts the transition.
*
* - Upon completion of the migrate_to, a new transition schedules a stop
* on both nodes and a start on node1.
*
* - If the new transition is aborted for any reason while the resource is
* stopping on node1, the transition after that stop completes will see
* the migrate_to and stop on the source, but it's still a partial
* migration, and the resource must be stopped on node2 because it is
* potentially active there due to the migrate_to.
*
* We also need to take into account that either node's history may be
* cleared at any point in the migration process.
*/
int from_rc = PCMK_OCF_OK;
int from_status = PCMK_EXEC_PENDING;
pcmk_node_t *target_node = NULL;
xmlNode *migrate_from = NULL;
const char *source = NULL;
const char *target = NULL;
bool source_newer_op = false;
bool target_newer_state = false;
bool active_on_target = false;
// Get source and target node names from XML
if (get_migration_node_names(history->xml, history->node, NULL, &source,
&target) != pcmk_rc_ok) {
return;
}
// Check for newer state on the source
source_newer_op = non_monitor_after(history->rsc->id, source, history->xml,
true, history->rsc->cluster);
// Check for a migrate_from action from this source on the target
migrate_from = find_lrm_op(history->rsc->id, PCMK_ACTION_MIGRATE_FROM,
target, source, -1, history->rsc->cluster);
if (migrate_from != NULL) {
if (source_newer_op) {
/* There's a newer non-monitor operation on the source and a
* migrate_from on the target, so this migrate_to is irrelevant to
* the resource's state.
*/
return;
}
crm_element_value_int(migrate_from, PCMK__XA_RC_CODE, &from_rc);
crm_element_value_int(migrate_from, PCMK__XA_OP_STATUS, &from_status);
}
/* If the resource has newer state on both the source and target after the
* migration events, this migrate_to is irrelevant to the resource's state.
*/
target_newer_state = newer_state_after_migrate(history->rsc->id, target,
history->xml, migrate_from,
history->rsc->cluster);
if (source_newer_op && target_newer_state) {
return;
}
/* Check for dangling migration (migrate_from succeeded but stop not done).
* We know there's no stop because we already returned if the target has a
* migrate_from and the source has any newer non-monitor operation.
*/
if ((from_rc == PCMK_OCF_OK) && (from_status == PCMK_EXEC_DONE)) {
add_dangling_migration(history->rsc, history->node);
return;
}
/* Without newer state, this migrate_to implies the resource is active.
* (Clones are not allowed to migrate, so role can't be promoted.)
*/
history->rsc->role = pcmk_role_started;
target_node = pe_find_node(history->rsc->cluster->nodes, target);
active_on_target = !target_newer_state && (target_node != NULL)
&& target_node->details->online;
if (from_status != PCMK_EXEC_PENDING) { // migrate_from failed on target
if (active_on_target) {
native_add_running(history->rsc, target_node, history->rsc->cluster,
TRUE);
} else {
// Mark resource as failed, require recovery, and prevent migration
pcmk__set_rsc_flags(history->rsc,
pcmk_rsc_failed|pcmk_rsc_stop_if_failed);
pcmk__clear_rsc_flags(history->rsc, pcmk_rsc_migratable);
}
return;
}
// The migrate_from is pending, complete but erased, or to be scheduled
/* If there is no history at all for the resource on an online target, then
* it was likely cleaned. Just return, and we'll schedule a probe. Once we
* have the probe result, it will be reflected in target_newer_state.
*/
if ((target_node != NULL) && target_node->details->online
&& unknown_on_node(history->rsc, target)) {
return;
}
if (active_on_target) {
pcmk_node_t *source_node = pe_find_node(history->rsc->cluster->nodes,
source);
native_add_running(history->rsc, target_node, history->rsc->cluster,
FALSE);
if ((source_node != NULL) && source_node->details->online) {
/* This is a partial migration: the migrate_to completed
* successfully on the source, but the migrate_from has not
* completed. Remember the source and target; if the newly
* chosen target remains the same when we schedule actions
* later, we may continue with the migration.
*/
history->rsc->partial_migration_target = target_node;
history->rsc->partial_migration_source = source_node;
}
} else if (!source_newer_op) {
// Mark resource as failed, require recovery, and prevent migration
pcmk__set_rsc_flags(history->rsc,
pcmk_rsc_failed|pcmk_rsc_stop_if_failed);
pcmk__clear_rsc_flags(history->rsc, pcmk_rsc_migratable);
}
}
/*!
* \internal
* \brief Update resource role etc. after a failed migrate_to action
*
* \param[in,out] history Parsed action result history
*/
static void
unpack_migrate_to_failure(struct action_history *history)
{
xmlNode *target_migrate_from = NULL;
const char *source = NULL;
const char *target = NULL;
// Get source and target node names from XML
if (get_migration_node_names(history->xml, history->node, NULL, &source,
&target) != pcmk_rc_ok) {
return;
}
/* If a migration failed, we have to assume the resource is active. Clones
* are not allowed to migrate, so role can't be promoted.
*/
history->rsc->role = pcmk_role_started;
// Check for migrate_from on the target
target_migrate_from = find_lrm_op(history->rsc->id,
PCMK_ACTION_MIGRATE_FROM, target, source,
PCMK_OCF_OK, history->rsc->cluster);
if (/* If the resource state is unknown on the target, it will likely be
* probed there.
* Don't just consider it running there. We will get back here anyway in
* case the probe detects it's running there.
*/
!unknown_on_node(history->rsc, target)
/* If the resource has newer state on the target after the migration
* events, this migrate_to no longer matters for the target.
*/
&& !newer_state_after_migrate(history->rsc->id, target, history->xml,
target_migrate_from,
history->rsc->cluster)) {
/* The resource has no newer state on the target, so assume it's still
* active there.
* (if it is up).
*/
pcmk_node_t *target_node = pe_find_node(history->rsc->cluster->nodes,
target);
if (target_node && target_node->details->online) {
native_add_running(history->rsc, target_node, history->rsc->cluster,
FALSE);
}
} else if (!non_monitor_after(history->rsc->id, source, history->xml, true,
history->rsc->cluster)) {
/* We know the resource has newer state on the target, but this
* migrate_to still matters for the source as long as there's no newer
* non-monitor operation there.
*/
// Mark node as having dangling migration so we can force a stop later
history->rsc->dangling_migrations =
g_list_prepend(history->rsc->dangling_migrations,
(gpointer) history->node);
}
}
/*!
* \internal
* \brief Update resource role etc. after a failed migrate_from action
*
* \param[in,out] history Parsed action result history
*/
static void
unpack_migrate_from_failure(struct action_history *history)
{
xmlNode *source_migrate_to = NULL;
const char *source = NULL;
const char *target = NULL;
// Get source and target node names from XML
if (get_migration_node_names(history->xml, NULL, history->node, &source,
&target) != pcmk_rc_ok) {
return;
}
/* If a migration failed, we have to assume the resource is active. Clones
* are not allowed to migrate, so role can't be promoted.
*/
history->rsc->role = pcmk_role_started;
// Check for a migrate_to on the source
source_migrate_to = find_lrm_op(history->rsc->id, PCMK_ACTION_MIGRATE_TO,
source, target, PCMK_OCF_OK,
history->rsc->cluster);
if (/* If the resource state is unknown on the source, it will likely be
* probed there.
* Don't just consider it running there. We will get back here anyway in
* case the probe detects it's running there.
*/
!unknown_on_node(history->rsc, source)
/* If the resource has newer state on the source after the migration
* events, this migrate_from no longer matters for the source.
*/
&& !newer_state_after_migrate(history->rsc->id, source,
source_migrate_to, history->xml,
history->rsc->cluster)) {
/* The resource has no newer state on the source, so assume it's still
* active there (if it is up).
*/
pcmk_node_t *source_node = pe_find_node(history->rsc->cluster->nodes,
source);
if (source_node && source_node->details->online) {
native_add_running(history->rsc, source_node, history->rsc->cluster,
TRUE);
}
}
}
/*!
* \internal
* \brief Add an action to cluster's list of failed actions
*
* \param[in,out] history Parsed action result history
*/
static void
record_failed_op(struct action_history *history)
{
if (!(history->node->details->online)) {
return;
}
for (const xmlNode *xIter = history->rsc->cluster->failed->children;
xIter != NULL; xIter = xIter->next) {
const char *key = pcmk__xe_history_key(xIter);
const char *uname = crm_element_value(xIter, PCMK_XA_UNAME);
if (pcmk__str_eq(history->key, key, pcmk__str_none)
&& pcmk__str_eq(uname, history->node->details->uname,
pcmk__str_casei)) {
crm_trace("Skipping duplicate entry %s on %s",
history->key, pcmk__node_name(history->node));
return;
}
}
crm_trace("Adding entry for %s on %s to failed action list",
history->key, pcmk__node_name(history->node));
crm_xml_add(history->xml, PCMK_XA_UNAME, history->node->details->uname);
crm_xml_add(history->xml, PCMK__XA_RSC_ID, history->rsc->id);
add_node_copy(history->rsc->cluster->failed, history->xml);
}
static char *
last_change_str(const xmlNode *xml_op)
{
time_t when;
char *result = NULL;
if (crm_element_value_epoch(xml_op, PCMK_XA_LAST_RC_CHANGE,
&when) == pcmk_ok) {
char *when_s = pcmk__epoch2str(&when, 0);
const char *p = strchr(when_s, ' ');
// Skip day of week to make message shorter
if ((p != NULL) && (*(++p) != '\0')) {
result = strdup(p);
CRM_ASSERT(result != NULL);
}
free(when_s);
}
if (result == NULL) {
result = strdup("unknown time");
CRM_ASSERT(result != NULL);
}
return result;
}
/*!
* \internal
* \brief Compare two on-fail values
*
* \param[in] first One on-fail value to compare
* \param[in] second The other on-fail value to compare
*
* \return A negative number if second is more severe than first, zero if they
* are equal, or a positive number if first is more severe than second.
* \note This is only needed until the action_fail_response values can be
* renumbered at the next API compatibility break.
*/
static int
cmp_on_fail(enum action_fail_response first, enum action_fail_response second)
{
switch (first) {
case pcmk_on_fail_demote:
switch (second) {
case pcmk_on_fail_ignore:
return 1;
case pcmk_on_fail_demote:
return 0;
default:
return -1;
}
break;
case pcmk_on_fail_reset_remote:
switch (second) {
case pcmk_on_fail_ignore:
case pcmk_on_fail_demote:
case pcmk_on_fail_restart:
return 1;
case pcmk_on_fail_reset_remote:
return 0;
default:
return -1;
}
break;
case pcmk_on_fail_restart_container:
switch (second) {
case pcmk_on_fail_ignore:
case pcmk_on_fail_demote:
case pcmk_on_fail_restart:
case pcmk_on_fail_reset_remote:
return 1;
case pcmk_on_fail_restart_container:
return 0;
default:
return -1;
}
break;
default:
break;
}
switch (second) {
case pcmk_on_fail_demote:
return (first == pcmk_on_fail_ignore)? -1 : 1;
case pcmk_on_fail_reset_remote:
switch (first) {
case pcmk_on_fail_ignore:
case pcmk_on_fail_demote:
case pcmk_on_fail_restart:
return -1;
default:
return 1;
}
break;
case pcmk_on_fail_restart_container:
switch (first) {
case pcmk_on_fail_ignore:
case pcmk_on_fail_demote:
case pcmk_on_fail_restart:
case pcmk_on_fail_reset_remote:
return -1;
default:
return 1;
}
break;
default:
break;
}
return first - second;
}
/*!
* \internal
* \brief Ban a resource (or its clone if an anonymous instance) from all nodes
*
* \param[in,out] rsc Resource to ban
*/
static void
ban_from_all_nodes(pcmk_resource_t *rsc)
{
int score = -INFINITY;
pcmk_resource_t *fail_rsc = rsc;
if (fail_rsc->parent != NULL) {
pcmk_resource_t *parent = uber_parent(fail_rsc);
if (pe_rsc_is_anon_clone(parent)) {
/* For anonymous clones, if an operation with
* PCMK_META_ON_FAIL=PCMK_VALUE_STOP fails for any instance, the
* entire clone must stop.
*/
fail_rsc = parent;
}
}
// Ban the resource from all nodes
crm_notice("%s will not be started under current conditions", fail_rsc->id);
if (fail_rsc->allowed_nodes != NULL) {
g_hash_table_destroy(fail_rsc->allowed_nodes);
}
fail_rsc->allowed_nodes = pe__node_list2table(rsc->cluster->nodes);
g_hash_table_foreach(fail_rsc->allowed_nodes, set_node_score, &score);
}
/*!
* \internal
* \brief Get configured failure handling and role after failure for an action
*
* \param[in,out] history Unpacked action history entry
* \param[out] on_fail Where to set configured failure handling
* \param[out] fail_role Where to set to role after failure
*/
static void
unpack_failure_handling(struct action_history *history,
enum action_fail_response *on_fail,
enum rsc_role_e *fail_role)
{
xmlNode *config = pcmk__find_action_config(history->rsc, history->task,
history->interval_ms, true);
GHashTable *meta = pcmk__unpack_action_meta(history->rsc, history->node,
history->task,
history->interval_ms, config);
const char *on_fail_str = g_hash_table_lookup(meta, PCMK_META_ON_FAIL);
*on_fail = pcmk__parse_on_fail(history->rsc, history->task,
history->interval_ms, on_fail_str);
*fail_role = pcmk__role_after_failure(history->rsc, history->task, *on_fail,
meta);
g_hash_table_destroy(meta);
}
/*!
* \internal
* \brief Update resource role, failure handling, etc., after a failed action
*
* \param[in,out] history Parsed action result history
* \param[in] config_on_fail Action failure handling from configuration
* \param[in] fail_role Resource's role after failure of this action
* \param[out] last_failure This will be set to the history XML
* \param[in,out] on_fail Actual handling of action result
*/
static void
unpack_rsc_op_failure(struct action_history *history,
enum action_fail_response config_on_fail,
enum rsc_role_e fail_role, xmlNode **last_failure,
enum action_fail_response *on_fail)
{
bool is_probe = false;
char *last_change_s = NULL;
*last_failure = history->xml;
is_probe = pcmk_xe_is_probe(history->xml);
last_change_s = last_change_str(history->xml);
if (!pcmk_is_set(history->rsc->cluster->flags, pcmk_sched_symmetric_cluster)
&& (history->exit_status == PCMK_OCF_NOT_INSTALLED)) {
crm_trace("Unexpected result (%s%s%s) was recorded for "
"%s of %s on %s at %s " CRM_XS " exit-status=%d id=%s",
services_ocf_exitcode_str(history->exit_status),
(pcmk__str_empty(history->exit_reason)? "" : ": "),
pcmk__s(history->exit_reason, ""),
(is_probe? "probe" : history->task), history->rsc->id,
pcmk__node_name(history->node), last_change_s,
history->exit_status, history->id);
} else {
pcmk__sched_warn("Unexpected result (%s%s%s) was recorded for %s of "
"%s on %s at %s " CRM_XS " exit-status=%d id=%s",
services_ocf_exitcode_str(history->exit_status),
(pcmk__str_empty(history->exit_reason)? "" : ": "),
pcmk__s(history->exit_reason, ""),
(is_probe? "probe" : history->task), history->rsc->id,
pcmk__node_name(history->node), last_change_s,
history->exit_status, history->id);
if (is_probe && (history->exit_status != PCMK_OCF_OK)
&& (history->exit_status != PCMK_OCF_NOT_RUNNING)
&& (history->exit_status != PCMK_OCF_RUNNING_PROMOTED)) {
/* A failed (not just unexpected) probe result could mean the user
* didn't know resources will be probed even where they can't run.
*/
crm_notice("If it is not possible for %s to run on %s, see "
"the " PCMK_XA_RESOURCE_DISCOVERY " option for location "
"constraints",
history->rsc->id, pcmk__node_name(history->node));
}
record_failed_op(history);
}
free(last_change_s);
if (cmp_on_fail(*on_fail, config_on_fail) < 0) {
pcmk__rsc_trace(history->rsc, "on-fail %s -> %s for %s",
fail2text(*on_fail), fail2text(config_on_fail),
history->key);
*on_fail = config_on_fail;
}
if (strcmp(history->task, PCMK_ACTION_STOP) == 0) {
resource_location(history->rsc, history->node, -INFINITY,
"__stop_fail__", history->rsc->cluster);
} else if (strcmp(history->task, PCMK_ACTION_MIGRATE_TO) == 0) {
unpack_migrate_to_failure(history);
} else if (strcmp(history->task, PCMK_ACTION_MIGRATE_FROM) == 0) {
unpack_migrate_from_failure(history);
} else if (strcmp(history->task, PCMK_ACTION_PROMOTE) == 0) {
history->rsc->role = pcmk_role_promoted;
} else if (strcmp(history->task, PCMK_ACTION_DEMOTE) == 0) {
if (config_on_fail == pcmk_on_fail_block) {
history->rsc->role = pcmk_role_promoted;
pe__set_next_role(history->rsc, pcmk_role_stopped,
"demote with on-fail=block");
} else if (history->exit_status == PCMK_OCF_NOT_RUNNING) {
history->rsc->role = pcmk_role_stopped;
} else {
/* Staying in the promoted role would put the scheduler and
* controller into a loop. Setting the role to unpromoted is not
* dangerous because the resource will be stopped as part of
* recovery, and any promotion will be ordered after that stop.
*/
history->rsc->role = pcmk_role_unpromoted;
}
}
if (is_probe && (history->exit_status == PCMK_OCF_NOT_INSTALLED)) {
/* leave stopped */
pcmk__rsc_trace(history->rsc, "Leaving %s stopped", history->rsc->id);
history->rsc->role = pcmk_role_stopped;
} else if (history->rsc->role < pcmk_role_started) {
pcmk__rsc_trace(history->rsc, "Setting %s active", history->rsc->id);
set_active(history->rsc);
}
pcmk__rsc_trace(history->rsc,
"Resource %s: role=%s unclean=%s on_fail=%s fail_role=%s",
history->rsc->id, pcmk_role_text(history->rsc->role),
pcmk__btoa(history->node->details->unclean),
fail2text(config_on_fail), pcmk_role_text(fail_role));
if ((fail_role != pcmk_role_started)
&& (history->rsc->next_role < fail_role)) {
pe__set_next_role(history->rsc, fail_role, "failure");
}
if (fail_role == pcmk_role_stopped) {
ban_from_all_nodes(history->rsc);
}
}
/*!
* \internal
* \brief Block a resource with a failed action if it cannot be recovered
*
* If resource action is a failed stop and fencing is not possible, mark the
* resource as unmanaged and blocked, since recovery cannot be done.
*
* \param[in,out] history Parsed action history entry
*/
static void
block_if_unrecoverable(struct action_history *history)
{
char *last_change_s = NULL;
if (strcmp(history->task, PCMK_ACTION_STOP) != 0) {
return; // All actions besides stop are always recoverable
}
if (pe_can_fence(history->node->details->data_set, history->node)) {
return; // Failed stops are recoverable via fencing
}
last_change_s = last_change_str(history->xml);
pcmk__sched_err("No further recovery can be attempted for %s "
"because %s on %s failed (%s%s%s) at %s "
CRM_XS " rc=%d id=%s",
history->rsc->id, history->task,
pcmk__node_name(history->node),
services_ocf_exitcode_str(history->exit_status),
(pcmk__str_empty(history->exit_reason)? "" : ": "),
pcmk__s(history->exit_reason, ""),
last_change_s, history->exit_status, history->id);
free(last_change_s);
pcmk__clear_rsc_flags(history->rsc, pcmk_rsc_managed);
pcmk__set_rsc_flags(history->rsc, pcmk_rsc_blocked);
}
/*!
* \internal
* \brief Update action history's execution status and why
*
* \param[in,out] history Parsed action history entry
* \param[out] why Where to store reason for update
* \param[in] value New value
* \param[in] reason Description of why value was changed
*/
static inline void
remap_because(struct action_history *history, const char **why, int value,
const char *reason)
{
if (history->execution_status != value) {
history->execution_status = value;
*why = reason;
}
}
/*!
* \internal
* \brief Remap informational monitor results and operation status
*
* For the monitor results, certain OCF codes are for providing extended information
* to the user about services that aren't yet failed but not entirely healthy either.
* These must be treated as the "normal" result by Pacemaker.
*
* For operation status, the action result can be used to determine an appropriate
* status for the purposes of responding to the action. The status provided by the
* executor is not directly usable since the executor does not know what was expected.
*
* \param[in,out] history Parsed action history entry
* \param[in,out] on_fail What should be done about the result
* \param[in] expired Whether result is expired
*
* \note If the result is remapped and the node is not shutting down or failed,
* the operation will be recorded in the scheduler data's list of failed
* operations to highlight it for the user.
*
* \note This may update the resource's current and next role.
*/
static void
remap_operation(struct action_history *history,
enum action_fail_response *on_fail, bool expired)
{
bool is_probe = false;
int orig_exit_status = history->exit_status;
int orig_exec_status = history->execution_status;
const char *why = NULL;
const char *task = history->task;
// Remap degraded results to their successful counterparts
history->exit_status = pcmk__effective_rc(history->exit_status);
if (history->exit_status != orig_exit_status) {
why = "degraded result";
if (!expired && (!history->node->details->shutdown
|| history->node->details->online)) {
record_failed_op(history);
}
}
if (!pe_rsc_is_bundled(history->rsc)
&& pcmk_xe_mask_probe_failure(history->xml)
&& ((history->execution_status != PCMK_EXEC_DONE)
|| (history->exit_status != PCMK_OCF_NOT_RUNNING))) {
history->execution_status = PCMK_EXEC_DONE;
history->exit_status = PCMK_OCF_NOT_RUNNING;
why = "equivalent probe result";
}
/* If the executor reported an execution status of anything but done or
* error, consider that final. But for done or error, we know better whether
* it should be treated as a failure or not, because we know the expected
* result.
*/
switch (history->execution_status) {
case PCMK_EXEC_DONE:
case PCMK_EXEC_ERROR:
break;
// These should be treated as node-fatal
case PCMK_EXEC_NO_FENCE_DEVICE:
case PCMK_EXEC_NO_SECRETS:
remap_because(history, &why, PCMK_EXEC_ERROR_HARD,
"node-fatal error");
goto remap_done;
default:
goto remap_done;
}
is_probe = pcmk_xe_is_probe(history->xml);
if (is_probe) {
task = "probe";
}
if (history->expected_exit_status < 0) {
/* Pre-1.0 Pacemaker versions, and Pacemaker 1.1.6 or earlier with
* Heartbeat 2.0.7 or earlier as the cluster layer, did not include the
* expected exit status in the transition key, which (along with the
* similar case of a corrupted transition key in the CIB) will be
* reported to this function as -1. Pacemaker 2.0+ does not support
* rolling upgrades from those versions or processing of saved CIB files
* from those versions, so we do not need to care much about this case.
*/
remap_because(history, &why, PCMK_EXEC_ERROR,
"obsolete history format");
pcmk__config_warn("Expected result not found for %s on %s "
"(corrupt or obsolete CIB?)",
history->key, pcmk__node_name(history->node));
} else if (history->exit_status == history->expected_exit_status) {
remap_because(history, &why, PCMK_EXEC_DONE, "expected result");
} else {
remap_because(history, &why, PCMK_EXEC_ERROR, "unexpected result");
pcmk__rsc_debug(history->rsc,
"%s on %s: expected %d (%s), got %d (%s%s%s)",
history->key, pcmk__node_name(history->node),
history->expected_exit_status,
services_ocf_exitcode_str(history->expected_exit_status),
history->exit_status,
services_ocf_exitcode_str(history->exit_status),
(pcmk__str_empty(history->exit_reason)? "" : ": "),
pcmk__s(history->exit_reason, ""));
}
switch (history->exit_status) {
case PCMK_OCF_OK:
if (is_probe
&& (history->expected_exit_status == PCMK_OCF_NOT_RUNNING)) {
char *last_change_s = last_change_str(history->xml);
remap_because(history, &why, PCMK_EXEC_DONE, "probe");
pcmk__rsc_info(history->rsc,
"Probe found %s active on %s at %s",
history->rsc->id, pcmk__node_name(history->node),
last_change_s);
free(last_change_s);
}
break;
case PCMK_OCF_NOT_RUNNING:
if (is_probe
|| (history->expected_exit_status == history->exit_status)
|| !pcmk_is_set(history->rsc->flags, pcmk_rsc_managed)) {
/* For probes, recurring monitors for the Stopped role, and
* unmanaged resources, "not running" is not considered a
* failure.
*/
remap_because(history, &why, PCMK_EXEC_DONE, "exit status");
history->rsc->role = pcmk_role_stopped;
*on_fail = pcmk_on_fail_ignore;
pe__set_next_role(history->rsc, pcmk_role_unknown,
"not running");
}
break;
case PCMK_OCF_RUNNING_PROMOTED:
if (is_probe
&& (history->exit_status != history->expected_exit_status)) {
char *last_change_s = last_change_str(history->xml);
remap_because(history, &why, PCMK_EXEC_DONE, "probe");
pcmk__rsc_info(history->rsc,
"Probe found %s active and promoted on %s at %s",
history->rsc->id,
pcmk__node_name(history->node), last_change_s);
free(last_change_s);
}
if (!expired
|| (history->exit_status == history->expected_exit_status)) {
history->rsc->role = pcmk_role_promoted;
}
break;
case PCMK_OCF_FAILED_PROMOTED:
if (!expired) {
history->rsc->role = pcmk_role_promoted;
}
remap_because(history, &why, PCMK_EXEC_ERROR, "exit status");
break;
case PCMK_OCF_NOT_CONFIGURED:
remap_because(history, &why, PCMK_EXEC_ERROR_FATAL, "exit status");
break;
case PCMK_OCF_UNIMPLEMENT_FEATURE:
{
guint interval_ms = 0;
crm_element_value_ms(history->xml, PCMK_META_INTERVAL,
&interval_ms);
if (interval_ms == 0) {
if (!expired) {
block_if_unrecoverable(history);
}
remap_because(history, &why, PCMK_EXEC_ERROR_HARD,
"exit status");
} else {
remap_because(history, &why, PCMK_EXEC_NOT_SUPPORTED,
"exit status");
}
}
break;
case PCMK_OCF_NOT_INSTALLED:
case PCMK_OCF_INVALID_PARAM:
case PCMK_OCF_INSUFFICIENT_PRIV:
if (!expired) {
block_if_unrecoverable(history);
}
remap_because(history, &why, PCMK_EXEC_ERROR_HARD, "exit status");
break;
default:
if (history->execution_status == PCMK_EXEC_DONE) {
char *last_change_s = last_change_str(history->xml);
crm_info("Treating unknown exit status %d from %s of %s "
"on %s at %s as failure",
history->exit_status, task, history->rsc->id,
pcmk__node_name(history->node), last_change_s);
remap_because(history, &why, PCMK_EXEC_ERROR,
"unknown exit status");
free(last_change_s);
}
break;
}
remap_done:
if (why != NULL) {
pcmk__rsc_trace(history->rsc,
"Remapped %s result from [%s: %s] to [%s: %s] "
"because of %s",
history->key, pcmk_exec_status_str(orig_exec_status),
crm_exit_str(orig_exit_status),
pcmk_exec_status_str(history->execution_status),
crm_exit_str(history->exit_status), why);
}
}
// return TRUE if start or monitor last failure but parameters changed
static bool
should_clear_for_param_change(const xmlNode *xml_op, const char *task,
pcmk_resource_t *rsc, pcmk_node_t *node)
{
if (pcmk__str_any_of(task, PCMK_ACTION_START, PCMK_ACTION_MONITOR, NULL)) {
if (pe__bundle_needs_remote_name(rsc)) {
/* We haven't allocated resources yet, so we can't reliably
* substitute addr parameters for the REMOTE_CONTAINER_HACK.
* When that's needed, defer the check until later.
*/
pe__add_param_check(xml_op, rsc, node, pcmk__check_last_failure,
rsc->cluster);
} else {
pcmk__op_digest_t *digest_data = NULL;
digest_data = rsc_action_digest_cmp(rsc, xml_op, node,
rsc->cluster);
switch (digest_data->rc) {
case pcmk__digest_unknown:
crm_trace("Resource %s history entry %s on %s"
" has no digest to compare",
rsc->id, pcmk__xe_history_key(xml_op),
node->details->id);
break;
case pcmk__digest_match:
break;
default:
return TRUE;
}
}
}
return FALSE;
}
// Order action after fencing of remote node, given connection rsc
static void
order_after_remote_fencing(pcmk_action_t *action, pcmk_resource_t *remote_conn,
pcmk_scheduler_t *scheduler)
{
pcmk_node_t *remote_node = pe_find_node(scheduler->nodes, remote_conn->id);
if (remote_node) {
pcmk_action_t *fence = pe_fence_op(remote_node, NULL, TRUE, NULL,
FALSE, scheduler);
order_actions(fence, action, pcmk__ar_first_implies_then);
}
}
static bool
should_ignore_failure_timeout(const pcmk_resource_t *rsc, const char *task,
guint interval_ms, bool is_last_failure)
{
/* Clearing failures of recurring monitors has special concerns. The
* executor reports only changes in the monitor result, so if the
* monitor is still active and still getting the same failure result,
* that will go undetected after the failure is cleared.
*
* Also, the operation history will have the time when the recurring
* monitor result changed to the given code, not the time when the
* result last happened.
*
* @TODO We probably should clear such failures only when the failure
* timeout has passed since the last occurrence of the failed result.
* However we don't record that information. We could maybe approximate
* that by clearing only if there is a more recent successful monitor or
* stop result, but we don't even have that information at this point
* since we are still unpacking the resource's operation history.
*
* This is especially important for remote connection resources with a
* reconnect interval, so in that case, we skip clearing failures
* if the remote node hasn't been fenced.
*/
if (rsc->remote_reconnect_ms
&& pcmk_is_set(rsc->cluster->flags, pcmk_sched_fencing_enabled)
&& (interval_ms != 0)
&& pcmk__str_eq(task, PCMK_ACTION_MONITOR, pcmk__str_casei)) {
pcmk_node_t *remote_node = pe_find_node(rsc->cluster->nodes, rsc->id);
if (remote_node && !remote_node->details->remote_was_fenced) {
if (is_last_failure) {
crm_info("Waiting to clear monitor failure for remote node %s"
" until fencing has occurred", rsc->id);
}
return TRUE;
}
}
return FALSE;
}
/*!
* \internal
* \brief Check operation age and schedule failure clearing when appropriate
*
* This function has two distinct purposes. The first is to check whether an
* operation history entry is expired (i.e. the resource has a failure timeout,
* the entry is older than the timeout, and the resource either has no fail
* count or its fail count is entirely older than the timeout). The second is to
* schedule fail count clearing when appropriate (i.e. the operation is expired
* and either the resource has an expired fail count or the operation is a
* last_failure for a remote connection resource with a reconnect interval,
* or the operation is a last_failure for a start or monitor operation and the
* resource's parameters have changed since the operation).
*
* \param[in,out] history Parsed action result history
*
* \return true if operation history entry is expired, otherwise false
*/
static bool
check_operation_expiry(struct action_history *history)
{
bool expired = false;
bool is_last_failure = pcmk__ends_with(history->id, "_last_failure_0");
time_t last_run = 0;
int unexpired_fail_count = 0;
const char *clear_reason = NULL;
if (history->execution_status == PCMK_EXEC_NOT_INSTALLED) {
pcmk__rsc_trace(history->rsc,
"Resource history entry %s on %s is not expired: "
"Not Installed does not expire",
history->id, pcmk__node_name(history->node));
return false; // "Not installed" must always be cleared manually
}
if ((history->rsc->failure_timeout > 0)
&& (crm_element_value_epoch(history->xml, PCMK_XA_LAST_RC_CHANGE,
&last_run) == 0)) {
/* Resource has a PCMK_META_FAILURE_TIMEOUT and history entry has a
* timestamp
*/
time_t now = get_effective_time(history->rsc->cluster);
time_t last_failure = 0;
// Is this particular operation history older than the failure timeout?
if ((now >= (last_run + history->rsc->failure_timeout))
&& !should_ignore_failure_timeout(history->rsc, history->task,
history->interval_ms,
is_last_failure)) {
expired = true;
}
// Does the resource as a whole have an unexpired fail count?
unexpired_fail_count = pe_get_failcount(history->node, history->rsc,
&last_failure,
pcmk__fc_effective,
history->xml);
// Update scheduler recheck time according to *last* failure
crm_trace("%s@%lld is %sexpired @%lld with unexpired_failures=%d timeout=%ds"
" last-failure@%lld",
history->id, (long long) last_run, (expired? "" : "not "),
(long long) now, unexpired_fail_count,
history->rsc->failure_timeout, (long long) last_failure);
last_failure += history->rsc->failure_timeout + 1;
if (unexpired_fail_count && (now < last_failure)) {
pe__update_recheck_time(last_failure, history->rsc->cluster,
"fail count expiration");
}
}
if (expired) {
if (pe_get_failcount(history->node, history->rsc, NULL,
pcmk__fc_default, history->xml)) {
// There is a fail count ignoring timeout
if (unexpired_fail_count == 0) {
// There is no fail count considering timeout
clear_reason = "it expired";
} else {
/* This operation is old, but there is an unexpired fail count.
* In a properly functioning cluster, this should only be
* possible if this operation is not a failure (otherwise the
* fail count should be expired too), so this is really just a
* failsafe.
*/
pcmk__rsc_trace(history->rsc,
"Resource history entry %s on %s is not "
"expired: Unexpired fail count",
history->id, pcmk__node_name(history->node));
expired = false;
}
} else if (is_last_failure
&& (history->rsc->remote_reconnect_ms != 0)) {
/* Clear any expired last failure when reconnect interval is set,
* even if there is no fail count.
*/
clear_reason = "reconnect interval is set";
}
}
if (!expired && is_last_failure
&& should_clear_for_param_change(history->xml, history->task,
history->rsc, history->node)) {
clear_reason = "resource parameters have changed";
}
if (clear_reason != NULL) {
pcmk_action_t *clear_op = NULL;
// Schedule clearing of the fail count
clear_op = pe__clear_failcount(history->rsc, history->node,
clear_reason, history->rsc->cluster);
if (pcmk_is_set(history->rsc->cluster->flags,
pcmk_sched_fencing_enabled)
&& (history->rsc->remote_reconnect_ms != 0)) {
/* If we're clearing a remote connection due to a reconnect
* interval, we want to wait until any scheduled fencing
* completes.
*
* We could limit this to remote_node->details->unclean, but at
* this point, that's always true (it won't be reliable until
* after unpack_node_history() is done).
*/
crm_info("Clearing %s failure will wait until any scheduled "
"fencing of %s completes",
history->task, history->rsc->id);
order_after_remote_fencing(clear_op, history->rsc,
history->rsc->cluster);
}
}
if (expired && (history->interval_ms == 0)
&& pcmk__str_eq(history->task, PCMK_ACTION_MONITOR, pcmk__str_none)) {
switch (history->exit_status) {
case PCMK_OCF_OK:
case PCMK_OCF_NOT_RUNNING:
case PCMK_OCF_RUNNING_PROMOTED:
case PCMK_OCF_DEGRADED:
case PCMK_OCF_DEGRADED_PROMOTED:
// Don't expire probes that return these values
pcmk__rsc_trace(history->rsc,
"Resource history entry %s on %s is not "
"expired: Probe result",
history->id, pcmk__node_name(history->node));
expired = false;
break;
}
}
return expired;
}
int
pe__target_rc_from_xml(const xmlNode *xml_op)
{
int target_rc = 0;
const char *key = crm_element_value(xml_op, PCMK__XA_TRANSITION_KEY);
if (key == NULL) {
return -1;
}
decode_transition_key(key, NULL, NULL, NULL, &target_rc);
return target_rc;
}
/*!
* \internal
* \brief Update a resource's state for an action result
*
* \param[in,out] history Parsed action history entry
* \param[in] exit_status Exit status to base new state on
* \param[in] last_failure Resource's last_failure entry, if known
* \param[in,out] on_fail Resource's current failure handling
*/
static void
update_resource_state(struct action_history *history, int exit_status,
const xmlNode *last_failure,
enum action_fail_response *on_fail)
{
bool clear_past_failure = false;
if ((exit_status == PCMK_OCF_NOT_INSTALLED)
|| (!pe_rsc_is_bundled(history->rsc)
&& pcmk_xe_mask_probe_failure(history->xml))) {
history->rsc->role = pcmk_role_stopped;
} else if (exit_status == PCMK_OCF_NOT_RUNNING) {
clear_past_failure = true;
} else if (pcmk__str_eq(history->task, PCMK_ACTION_MONITOR,
pcmk__str_none)) {
if ((last_failure != NULL)
&& pcmk__str_eq(history->key, pcmk__xe_history_key(last_failure),
pcmk__str_none)) {
clear_past_failure = true;
}
if (history->rsc->role < pcmk_role_started) {
set_active(history->rsc);
}
} else if (pcmk__str_eq(history->task, PCMK_ACTION_START, pcmk__str_none)) {
history->rsc->role = pcmk_role_started;
clear_past_failure = true;
} else if (pcmk__str_eq(history->task, PCMK_ACTION_STOP, pcmk__str_none)) {
history->rsc->role = pcmk_role_stopped;
clear_past_failure = true;
} else if (pcmk__str_eq(history->task, PCMK_ACTION_PROMOTE,
pcmk__str_none)) {
history->rsc->role = pcmk_role_promoted;
clear_past_failure = true;
} else if (pcmk__str_eq(history->task, PCMK_ACTION_DEMOTE,
pcmk__str_none)) {
if (*on_fail == pcmk_on_fail_demote) {
/* Demote clears an error only if
* PCMK_META_ON_FAIL=PCMK_VALUE_DEMOTE
*/
clear_past_failure = true;
}
history->rsc->role = pcmk_role_unpromoted;
} else if (pcmk__str_eq(history->task, PCMK_ACTION_MIGRATE_FROM,
pcmk__str_none)) {
history->rsc->role = pcmk_role_started;
clear_past_failure = true;
} else if (pcmk__str_eq(history->task, PCMK_ACTION_MIGRATE_TO,
pcmk__str_none)) {
unpack_migrate_to_success(history);
} else if (history->rsc->role < pcmk_role_started) {
pcmk__rsc_trace(history->rsc, "%s active on %s",
history->rsc->id, pcmk__node_name(history->node));
set_active(history->rsc);
}
if (!clear_past_failure) {
return;
}
switch (*on_fail) {
case pcmk_on_fail_stop:
case pcmk_on_fail_ban:
case pcmk_on_fail_standby_node:
case pcmk_on_fail_fence_node:
pcmk__rsc_trace(history->rsc,
"%s (%s) is not cleared by a completed %s",
history->rsc->id, fail2text(*on_fail),
history->task);
break;
case pcmk_on_fail_block:
case pcmk_on_fail_ignore:
case pcmk_on_fail_demote:
case pcmk_on_fail_restart:
case pcmk_on_fail_restart_container:
*on_fail = pcmk_on_fail_ignore;
pe__set_next_role(history->rsc, pcmk_role_unknown,
"clear past failures");
break;
case pcmk_on_fail_reset_remote:
if (history->rsc->remote_reconnect_ms == 0) {
/* With no reconnect interval, the connection is allowed to
* start again after the remote node is fenced and
* completely stopped. (With a reconnect interval, we wait
* for the failure to be cleared entirely before attempting
* to reconnect.)
*/
*on_fail = pcmk_on_fail_ignore;
pe__set_next_role(history->rsc, pcmk_role_unknown,
"clear past failures and reset remote");
}
break;
}
}
/*!
* \internal
* \brief Check whether a given history entry matters for resource state
*
* \param[in] history Parsed action history entry
*
* \return true if action can affect resource state, otherwise false
*/
static inline bool
can_affect_state(struct action_history *history)
{
#if 0
/* @COMPAT It might be better to parse only actions we know we're interested
* in, rather than exclude a couple we don't. However that would be a
* behavioral change that should be done at a major or minor series release.
* Currently, unknown operations can affect whether a resource is considered
* active and/or failed.
*/
return pcmk__str_any_of(history->task, PCMK_ACTION_MONITOR,
PCMK_ACTION_START, PCMK_ACTION_STOP,
PCMK_ACTION_PROMOTE, PCMK_ACTION_DEMOTE,
PCMK_ACTION_MIGRATE_TO, PCMK_ACTION_MIGRATE_FROM,
"asyncmon", NULL);
#else
return !pcmk__str_any_of(history->task, PCMK_ACTION_NOTIFY,
PCMK_ACTION_META_DATA, NULL);
#endif
}
/*!
* \internal
* \brief Unpack execution/exit status and exit reason from a history entry
*
* \param[in,out] history Action history entry to unpack
*
* \return Standard Pacemaker return code
*/
static int
unpack_action_result(struct action_history *history)
{
if ((crm_element_value_int(history->xml, PCMK__XA_OP_STATUS,
&(history->execution_status)) < 0)
|| (history->execution_status < PCMK_EXEC_PENDING)
|| (history->execution_status > PCMK_EXEC_MAX)
|| (history->execution_status == PCMK_EXEC_CANCELLED)) {
pcmk__config_err("Ignoring resource history entry %s for %s on %s "
"with invalid " PCMK__XA_OP_STATUS " '%s'",
history->id, history->rsc->id,
pcmk__node_name(history->node),
pcmk__s(crm_element_value(history->xml,
PCMK__XA_OP_STATUS),
""));
return pcmk_rc_unpack_error;
}
if ((crm_element_value_int(history->xml, PCMK__XA_RC_CODE,
&(history->exit_status)) < 0)
|| (history->exit_status < 0) || (history->exit_status > CRM_EX_MAX)) {
#if 0
/* @COMPAT We should ignore malformed entries, but since that would
* change behavior, it should be done at a major or minor series
* release.
*/
pcmk__config_err("Ignoring resource history entry %s for %s on %s "
"with invalid " PCMK__XA_RC_CODE " '%s'",
history->id, history->rsc->id,
pcmk__node_name(history->node),
pcmk__s(crm_element_value(history->xml,
PCMK__XA_RC_CODE),
""));
return pcmk_rc_unpack_error;
#else
history->exit_status = CRM_EX_ERROR;
#endif
}
history->exit_reason = crm_element_value(history->xml, PCMK_XA_EXIT_REASON);
return pcmk_rc_ok;
}
/*!
* \internal
* \brief Process an action history entry whose result expired
*
* \param[in,out] history Parsed action history entry
* \param[in] orig_exit_status Action exit status before remapping
*
* \return Standard Pacemaker return code (in particular, pcmk_rc_ok means the
* entry needs no further processing)
*/
static int
process_expired_result(struct action_history *history, int orig_exit_status)
{
if (!pe_rsc_is_bundled(history->rsc)
&& pcmk_xe_mask_probe_failure(history->xml)
&& (orig_exit_status != history->expected_exit_status)) {
if (history->rsc->role <= pcmk_role_stopped) {
history->rsc->role = pcmk_role_unknown;
}
crm_trace("Ignoring resource history entry %s for probe of %s on %s: "
"Masked failure expired",
history->id, history->rsc->id,
pcmk__node_name(history->node));
return pcmk_rc_ok;
}
if (history->exit_status == history->expected_exit_status) {
return pcmk_rc_undetermined; // Only failures expire
}
if (history->interval_ms == 0) {
crm_notice("Ignoring resource history entry %s for %s of %s on %s: "
"Expired failure",
history->id, history->task, history->rsc->id,
pcmk__node_name(history->node));
return pcmk_rc_ok;
}
if (history->node->details->online && !history->node->details->unclean) {
/* Reschedule the recurring action. schedule_cancel() won't work at
* this stage, so as a hacky workaround, forcibly change the restart
* digest so pcmk__check_action_config() does what we want later.
*
* @TODO We should skip this if there is a newer successful monitor.
* Also, this causes rescheduling only if the history entry
* has a PCMK__XA_OP_DIGEST (which the expire-non-blocked-failure
* scheduler regression test doesn't, but that may not be a
* realistic scenario in production).
*/
crm_notice("Rescheduling %s-interval %s of %s on %s "
"after failure expired",
pcmk__readable_interval(history->interval_ms), history->task,
history->rsc->id, pcmk__node_name(history->node));
crm_xml_add(history->xml, PCMK__XA_OP_RESTART_DIGEST,
"calculated-failure-timeout");
return pcmk_rc_ok;
}
return pcmk_rc_undetermined;
}
/*!
* \internal
* \brief Process a masked probe failure
*
* \param[in,out] history Parsed action history entry
* \param[in] orig_exit_status Action exit status before remapping
* \param[in] last_failure Resource's last_failure entry, if known
* \param[in,out] on_fail Resource's current failure handling
*/
static void
mask_probe_failure(struct action_history *history, int orig_exit_status,
const xmlNode *last_failure,
enum action_fail_response *on_fail)
{
pcmk_resource_t *ban_rsc = history->rsc;
if (!pcmk_is_set(history->rsc->flags, pcmk_rsc_unique)) {
ban_rsc = uber_parent(history->rsc);
}
crm_notice("Treating probe result '%s' for %s on %s as 'not running'",
services_ocf_exitcode_str(orig_exit_status), history->rsc->id,
pcmk__node_name(history->node));
update_resource_state(history, history->expected_exit_status, last_failure,
on_fail);
crm_xml_add(history->xml, PCMK_XA_UNAME, history->node->details->uname);
record_failed_op(history);
resource_location(ban_rsc, history->node, -INFINITY, "masked-probe-failure",
history->rsc->cluster);
}
/*!
* \internal Check whether a given failure is for a given pending action
*
* \param[in] history Parsed history entry for pending action
* \param[in] last_failure Resource's last_failure entry, if known
*
* \return true if \p last_failure is failure of pending action in \p history,
* otherwise false
* \note Both \p history and \p last_failure must come from the same
* \c PCMK__XE_LRM_RESOURCE block, as node and resource are assumed to be
* the same.
*/
static bool
failure_is_newer(const struct action_history *history,
const xmlNode *last_failure)
{
guint failure_interval_ms = 0U;
long long failure_change = 0LL;
long long this_change = 0LL;
if (last_failure == NULL) {
return false; // Resource has no last_failure entry
}
if (!pcmk__str_eq(history->task,
crm_element_value(last_failure, PCMK_XA_OPERATION),
pcmk__str_none)) {
return false; // last_failure is for different action
}
if ((crm_element_value_ms(last_failure, PCMK_META_INTERVAL,
&failure_interval_ms) != pcmk_ok)
|| (history->interval_ms != failure_interval_ms)) {
return false; // last_failure is for action with different interval
}
if ((pcmk__scan_ll(crm_element_value(history->xml, PCMK_XA_LAST_RC_CHANGE),
&this_change, 0LL) != pcmk_rc_ok)
|| (pcmk__scan_ll(crm_element_value(last_failure,
PCMK_XA_LAST_RC_CHANGE),
&failure_change, 0LL) != pcmk_rc_ok)
|| (failure_change < this_change)) {
return false; // Failure is not known to be newer
}
return true;
}
/*!
* \internal
* \brief Update a resource's role etc. for a pending action
*
* \param[in,out] history Parsed history entry for pending action
* \param[in] last_failure Resource's last_failure entry, if known
*/
static void
process_pending_action(struct action_history *history,
const xmlNode *last_failure)
{
/* For recurring monitors, a failure is recorded only in RSC_last_failure_0,
* and there might be a RSC_monitor_INTERVAL entry with the last successful
* or pending result.
*
* If last_failure contains the failure of the pending recurring monitor
* we're processing here, and is newer, the action is no longer pending.
* (Pending results have call ID -1, which sorts last, so the last failure
* if any should be known.)
*/
if (failure_is_newer(history, last_failure)) {
return;
}
if (strcmp(history->task, PCMK_ACTION_START) == 0) {
pcmk__set_rsc_flags(history->rsc, pcmk_rsc_start_pending);
set_active(history->rsc);
} else if (strcmp(history->task, PCMK_ACTION_PROMOTE) == 0) {
history->rsc->role = pcmk_role_promoted;
} else if ((strcmp(history->task, PCMK_ACTION_MIGRATE_TO) == 0)
&& history->node->details->unclean) {
/* A migrate_to action is pending on a unclean source, so force a stop
* on the target.
*/
const char *migrate_target = NULL;
pcmk_node_t *target = NULL;
migrate_target = crm_element_value(history->xml,
PCMK__META_MIGRATE_TARGET);
target = pe_find_node(history->rsc->cluster->nodes, migrate_target);
if (target != NULL) {
stop_action(history->rsc, target, FALSE);
}
}
if (history->rsc->pending_task != NULL) {
/* There should never be multiple pending actions, but as a failsafe,
* just remember the first one processed for display purposes.
*/
return;
}
if (pcmk_is_probe(history->task, history->interval_ms)) {
/* Pending probes are currently never displayed, even if pending
* operations are requested. If we ever want to change that,
* enable the below and the corresponding part of
* native.c:native_pending_task().
*/
#if 0
history->rsc->pending_task = strdup("probe");
history->rsc->pending_node = history->node;
#endif
} else {
history->rsc->pending_task = strdup(history->task);
history->rsc->pending_node = history->node;
}
}
static void
unpack_rsc_op(pcmk_resource_t *rsc, pcmk_node_t *node, xmlNode *xml_op,
xmlNode **last_failure, enum action_fail_response *on_fail)
{
int old_rc = 0;
bool expired = false;
pcmk_resource_t *parent = rsc;
enum rsc_role_e fail_role = pcmk_role_unknown;
enum action_fail_response failure_strategy = pcmk_on_fail_restart;
struct action_history history = {
.rsc = rsc,
.node = node,
.xml = xml_op,
.execution_status = PCMK_EXEC_UNKNOWN,
};
CRM_CHECK(rsc && node && xml_op, return);
history.id = ID(xml_op);
if (history.id == NULL) {
pcmk__config_err("Ignoring resource history entry for %s on %s "
"without ID", rsc->id, pcmk__node_name(node));
return;
}
// Task and interval
history.task = crm_element_value(xml_op, PCMK_XA_OPERATION);
if (history.task == NULL) {
pcmk__config_err("Ignoring resource history entry %s for %s on %s "
"without " PCMK_XA_OPERATION,
history.id, rsc->id, pcmk__node_name(node));
return;
}
crm_element_value_ms(xml_op, PCMK_META_INTERVAL, &(history.interval_ms));
if (!can_affect_state(&history)) {
pcmk__rsc_trace(rsc,
"Ignoring resource history entry %s for %s on %s "
"with irrelevant action '%s'",
history.id, rsc->id, pcmk__node_name(node),
history.task);
return;
}
if (unpack_action_result(&history) != pcmk_rc_ok) {
return; // Error already logged
}
history.expected_exit_status = pe__target_rc_from_xml(xml_op);
history.key = pcmk__xe_history_key(xml_op);
crm_element_value_int(xml_op, PCMK__XA_CALL_ID, &(history.call_id));
pcmk__rsc_trace(rsc, "Unpacking %s (%s call %d on %s): %s (%s)",
history.id, history.task, history.call_id,
pcmk__node_name(node),
pcmk_exec_status_str(history.execution_status),
crm_exit_str(history.exit_status));
if (node->details->unclean) {
pcmk__rsc_trace(rsc,
"%s is running on %s, which is unclean (further action "
"depends on value of stop's on-fail attribute)",
rsc->id, pcmk__node_name(node));
}
expired = check_operation_expiry(&history);
old_rc = history.exit_status;
remap_operation(&history, on_fail, expired);
if (expired && (process_expired_result(&history, old_rc) == pcmk_rc_ok)) {
goto done;
}
if (!pe_rsc_is_bundled(rsc) && pcmk_xe_mask_probe_failure(xml_op)) {
mask_probe_failure(&history, old_rc, *last_failure, on_fail);
goto done;
}
if (!pcmk_is_set(rsc->flags, pcmk_rsc_unique)) {
parent = uber_parent(rsc);
}
switch (history.execution_status) {
case PCMK_EXEC_PENDING:
process_pending_action(&history, *last_failure);
goto done;
case PCMK_EXEC_DONE:
update_resource_state(&history, history.exit_status, *last_failure,
on_fail);
goto done;
case PCMK_EXEC_NOT_INSTALLED:
unpack_failure_handling(&history, &failure_strategy, &fail_role);
if (failure_strategy == pcmk_on_fail_ignore) {
crm_warn("Cannot ignore failed %s of %s on %s: "
"Resource agent doesn't exist "
CRM_XS " status=%d rc=%d id=%s",
history.task, rsc->id, pcmk__node_name(node),
history.execution_status, history.exit_status,
history.id);
/* Also for printing it as "FAILED" by marking it as
* pcmk_rsc_failed later
*/
*on_fail = pcmk_on_fail_ban;
}
resource_location(parent, node, -INFINITY, "hard-error",
rsc->cluster);
unpack_rsc_op_failure(&history, failure_strategy, fail_role,
last_failure, on_fail);
goto done;
case PCMK_EXEC_NOT_CONNECTED:
if (pe__is_guest_or_remote_node(node)
&& pcmk_is_set(node->details->remote_rsc->flags,
pcmk_rsc_managed)) {
/* We should never get into a situation where a managed remote
* connection resource is considered OK but a resource action
* behind the connection gets a "not connected" status. But as a
* fail-safe in case a bug or unusual circumstances do lead to
* that, ensure the remote connection is considered failed.
*/
pcmk__set_rsc_flags(node->details->remote_rsc,
pcmk_rsc_failed|pcmk_rsc_stop_if_failed);
}
break; // Not done, do error handling
case PCMK_EXEC_ERROR:
case PCMK_EXEC_ERROR_HARD:
case PCMK_EXEC_ERROR_FATAL:
case PCMK_EXEC_TIMEOUT:
case PCMK_EXEC_NOT_SUPPORTED:
case PCMK_EXEC_INVALID:
break; // Not done, do error handling
default: // No other value should be possible at this point
break;
}
unpack_failure_handling(&history, &failure_strategy, &fail_role);
if ((failure_strategy == pcmk_on_fail_ignore)
|| ((failure_strategy == pcmk_on_fail_restart_container)
&& (strcmp(history.task, PCMK_ACTION_STOP) == 0))) {
char *last_change_s = last_change_str(xml_op);
crm_warn("Pretending failed %s (%s%s%s) of %s on %s at %s succeeded "
CRM_XS " %s",
history.task, services_ocf_exitcode_str(history.exit_status),
(pcmk__str_empty(history.exit_reason)? "" : ": "),
pcmk__s(history.exit_reason, ""), rsc->id,
pcmk__node_name(node), last_change_s, history.id);
free(last_change_s);
update_resource_state(&history, history.expected_exit_status,
*last_failure, on_fail);
crm_xml_add(xml_op, PCMK_XA_UNAME, node->details->uname);
pcmk__set_rsc_flags(rsc, pcmk_rsc_ignore_failure);
record_failed_op(&history);
if ((failure_strategy == pcmk_on_fail_restart_container)
&& cmp_on_fail(*on_fail, pcmk_on_fail_restart) <= 0) {
*on_fail = failure_strategy;
}
} else {
unpack_rsc_op_failure(&history, failure_strategy, fail_role,
last_failure, on_fail);
if (history.execution_status == PCMK_EXEC_ERROR_HARD) {
uint8_t log_level = LOG_ERR;
if (history.exit_status == PCMK_OCF_NOT_INSTALLED) {
log_level = LOG_NOTICE;
}
do_crm_log(log_level,
"Preventing %s from restarting on %s because "
"of hard failure (%s%s%s) " CRM_XS " %s",
parent->id, pcmk__node_name(node),
services_ocf_exitcode_str(history.exit_status),
(pcmk__str_empty(history.exit_reason)? "" : ": "),
pcmk__s(history.exit_reason, ""), history.id);
resource_location(parent, node, -INFINITY, "hard-error",
rsc->cluster);
} else if (history.execution_status == PCMK_EXEC_ERROR_FATAL) {
pcmk__sched_err("Preventing %s from restarting anywhere because "
"of fatal failure (%s%s%s) " CRM_XS " %s",
parent->id,
services_ocf_exitcode_str(history.exit_status),
(pcmk__str_empty(history.exit_reason)? "" : ": "),
pcmk__s(history.exit_reason, ""), history.id);
resource_location(parent, NULL, -INFINITY, "fatal-error",
rsc->cluster);
}
}
done:
pcmk__rsc_trace(rsc, "%s role on %s after %s is %s (next %s)",
rsc->id, pcmk__node_name(node), history.id,
pcmk_role_text(rsc->role),
pcmk_role_text(rsc->next_role));
}
static void
add_node_attrs(const xmlNode *xml_obj, pcmk_node_t *node, bool overwrite,
pcmk_scheduler_t *scheduler)
{
const char *cluster_name = NULL;
pe_rule_eval_data_t rule_data = {
.node_hash = NULL,
.role = pcmk_role_unknown,
.now = scheduler->now,
.match_data = NULL,
.rsc_data = NULL,
.op_data = NULL
};
g_hash_table_insert(node->details->attrs,
strdup(CRM_ATTR_UNAME), strdup(node->details->uname));
g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_ID),
strdup(node->details->id));
if (pcmk__str_eq(node->details->id, scheduler->dc_uuid, pcmk__str_casei)) {
scheduler->dc_node = node;
node->details->is_dc = TRUE;
g_hash_table_insert(node->details->attrs,
strdup(CRM_ATTR_IS_DC), strdup(PCMK_VALUE_TRUE));
} else {
g_hash_table_insert(node->details->attrs,
strdup(CRM_ATTR_IS_DC), strdup(PCMK_VALUE_FALSE));
}
cluster_name = g_hash_table_lookup(scheduler->config_hash,
PCMK_OPT_CLUSTER_NAME);
if (cluster_name) {
g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_CLUSTER_NAME),
strdup(cluster_name));
}
pe__unpack_dataset_nvpairs(xml_obj, PCMK_XE_INSTANCE_ATTRIBUTES, &rule_data,
node->details->attrs, NULL, overwrite,
scheduler);
pe__unpack_dataset_nvpairs(xml_obj, PCMK_XE_UTILIZATION, &rule_data,
node->details->utilization, NULL,
FALSE, scheduler);
if (pe_node_attribute_raw(node, CRM_ATTR_SITE_NAME) == NULL) {
const char *site_name = pe_node_attribute_raw(node, "site-name");
if (site_name) {
g_hash_table_insert(node->details->attrs,
strdup(CRM_ATTR_SITE_NAME),
strdup(site_name));
} else if (cluster_name) {
/* Default to cluster-name if unset */
g_hash_table_insert(node->details->attrs,
strdup(CRM_ATTR_SITE_NAME),
strdup(cluster_name));
}
}
}
static GList *
extract_operations(const char *node, const char *rsc, xmlNode * rsc_entry, gboolean active_filter)
{
int counter = -1;
int stop_index = -1;
int start_index = -1;
xmlNode *rsc_op = NULL;
GList *gIter = NULL;
GList *op_list = NULL;
GList *sorted_op_list = NULL;
/* extract operations */
op_list = NULL;
sorted_op_list = NULL;
for (rsc_op = pcmk__xe_first_child(rsc_entry);
rsc_op != NULL; rsc_op = pcmk__xe_next(rsc_op)) {
if (pcmk__str_eq((const char *)rsc_op->name, PCMK__XE_LRM_RSC_OP,
pcmk__str_none)) {
crm_xml_add(rsc_op, PCMK_XA_RESOURCE, rsc);
crm_xml_add(rsc_op, PCMK_XA_UNAME, node);
op_list = g_list_prepend(op_list, rsc_op);
}
}
if (op_list == NULL) {
/* if there are no operations, there is nothing to do */
return NULL;
}
sorted_op_list = g_list_sort(op_list, sort_op_by_callid);
/* create active recurring operations as optional */
if (active_filter == FALSE) {
return sorted_op_list;
}
op_list = NULL;
calculate_active_ops(sorted_op_list, &start_index, &stop_index);
for (gIter = sorted_op_list; gIter != NULL; gIter = gIter->next) {
xmlNode *rsc_op = (xmlNode *) gIter->data;
counter++;
if (start_index < stop_index) {
crm_trace("Skipping %s: not active", ID(rsc_entry));
break;
} else if (counter < start_index) {
crm_trace("Skipping %s: old", ID(rsc_op));
continue;
}
op_list = g_list_append(op_list, rsc_op);
}
g_list_free(sorted_op_list);
return op_list;
}
GList *
find_operations(const char *rsc, const char *node, gboolean active_filter,
pcmk_scheduler_t *scheduler)
{
GList *output = NULL;
GList *intermediate = NULL;
xmlNode *tmp = NULL;
xmlNode *status = find_xml_node(scheduler->input, PCMK_XE_STATUS, TRUE);
pcmk_node_t *this_node = NULL;
xmlNode *node_state = NULL;
for (node_state = pcmk__xe_first_child(status); node_state != NULL;
node_state = pcmk__xe_next(node_state)) {
if (pcmk__str_eq((const char *) node_state->name, PCMK__XE_NODE_STATE,
pcmk__str_none)) {
const char *uname = crm_element_value(node_state, PCMK_XA_UNAME);
if (node != NULL && !pcmk__str_eq(uname, node, pcmk__str_casei)) {
continue;
}
this_node = pe_find_node(scheduler->nodes, uname);
if(this_node == NULL) {
CRM_LOG_ASSERT(this_node != NULL);
continue;
} else if (pe__is_guest_or_remote_node(this_node)) {
determine_remote_online_status(scheduler, this_node);
} else {
determine_online_status(node_state, this_node, scheduler);
}
if (this_node->details->online
|| pcmk_is_set(scheduler->flags, pcmk_sched_fencing_enabled)) {
/* offline nodes run no resources...
* unless stonith is enabled in which case we need to
* make sure rsc start events happen after the stonith
*/
xmlNode *lrm_rsc = NULL;
tmp = find_xml_node(node_state, PCMK__XE_LRM, FALSE);
tmp = find_xml_node(tmp, PCMK__XE_LRM_RESOURCES, FALSE);
for (lrm_rsc = pcmk__xe_first_child(tmp); lrm_rsc != NULL;
lrm_rsc = pcmk__xe_next(lrm_rsc)) {
if (pcmk__str_eq((const char *) lrm_rsc->name,
PCMK__XE_LRM_RESOURCE, pcmk__str_none)) {
const char *rsc_id = crm_element_value(lrm_rsc,
PCMK_XA_ID);
if (rsc != NULL && !pcmk__str_eq(rsc_id, rsc, pcmk__str_casei)) {
continue;
}
intermediate = extract_operations(uname, rsc_id, lrm_rsc, active_filter);
output = g_list_concat(output, intermediate);
}
}
}
}
}
return output;
}

File Metadata

Mime Type
text/x-diff
Expires
Thu, Oct 16, 12:00 AM (2 h, 59 m)
Storage Engine
blob
Storage Format
Raw Data
Storage Handle
2530605
Default Alt Text
(426 KB)

Event Timeline