Page MenuHomeClusterLabs Projects

No OneTemporary

This file is larger than 256 KB, so syntax highlighting was skipped.
diff --git a/daemons/execd/cts-exec-helper.c b/daemons/execd/cts-exec-helper.c
index ac214a7d8b..54756cc395 100644
--- a/daemons/execd/cts-exec-helper.c
+++ b/daemons/execd/cts-exec-helper.c
@@ -1,626 +1,626 @@
/*
* Copyright 2012-2024 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 <glib.h>
#include <unistd.h>
#include <crm/crm.h>
#include <crm/services.h>
#include <crm/common/cmdline_internal.h>
#include <crm/common/mainloop.h>
#include <crm/pengine/status.h>
#include <crm/pengine/internal.h>
#include <crm/cib.h>
#include <crm/cib/internal.h>
#include <crm/lrmd.h>
#define SUMMARY "cts-exec-helper - inject commands into the Pacemaker executor and watch for events"
static int exec_call_id = 0;
static gboolean start_test(gpointer user_data);
static void try_connect(void);
static char *key = NULL;
static char *val = NULL;
static struct {
int verbose;
int quiet;
guint interval_ms;
int timeout;
int start_delay;
int cancel_call_id;
gboolean no_wait;
gboolean is_running;
gboolean no_connect;
int exec_call_opts;
const char *api_call;
const char *rsc_id;
const char *provider;
const char *class;
const char *type;
const char *action;
const char *listen;
gboolean use_tls;
lrmd_key_value_t *params;
} options;
static gboolean
interval_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
return pcmk_parse_interval_spec(optarg,
&options.interval_ms) == pcmk_rc_ok;
}
static gboolean
notify_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
if (pcmk__str_any_of(option_name, "--notify-orig", "-n", NULL)) {
options.exec_call_opts = lrmd_opt_notify_orig_only;
} else if (pcmk__str_any_of(option_name, "--notify-changes", "-o", NULL)) {
options.exec_call_opts = lrmd_opt_notify_changes_only;
}
return TRUE;
}
static gboolean
param_key_val_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
if (pcmk__str_any_of(option_name, "--param-key", "-k", NULL)) {
pcmk__str_update(&key, optarg);
} else if (pcmk__str_any_of(option_name, "--param-val", "-v", NULL)) {
pcmk__str_update(&val, optarg);
}
if (key != NULL && val != NULL) {
options.params = lrmd_key_value_add(options.params, key, val);
pcmk__str_update(&key, NULL);
pcmk__str_update(&val, NULL);
}
return TRUE;
}
static GOptionEntry basic_entries[] = {
{ "api-call", 'c', 0, G_OPTION_ARG_STRING, &options.api_call,
"Directly relates to executor API functions",
NULL },
{ "is-running", 'R', 0, G_OPTION_ARG_NONE, &options.is_running,
"Determine if a resource is registered and running",
NULL },
{ "listen", 'l', 0, G_OPTION_ARG_STRING, &options.listen,
"Listen for a specific event string",
NULL },
{ "no-wait", 'w', 0, G_OPTION_ARG_NONE, &options.no_wait,
"Make api call and do not wait for result",
NULL },
{ "notify-changes", 'o', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, notify_cb,
"Only notify client changes to recurring operations",
NULL },
{ "notify-orig", 'n', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, notify_cb,
"Only notify this client of the results of an API action",
NULL },
{ "tls", 'S', 0, G_OPTION_ARG_NONE, &options.use_tls,
"Use TLS backend for local connection",
NULL },
{ NULL }
};
static GOptionEntry api_call_entries[] = {
{ "action", 'a', 0, G_OPTION_ARG_STRING, &options.action,
NULL, NULL },
{ "cancel-call-id", 'x', 0, G_OPTION_ARG_INT, &options.cancel_call_id,
NULL, NULL },
{ "class", 'C', 0, G_OPTION_ARG_STRING, &options.class,
NULL, NULL },
{ "interval", 'i', 0, G_OPTION_ARG_CALLBACK, interval_cb,
NULL, NULL },
{ "param-key", 'k', 0, G_OPTION_ARG_CALLBACK, param_key_val_cb,
NULL, NULL },
{ "param-val", 'v', 0, G_OPTION_ARG_CALLBACK, param_key_val_cb,
NULL, NULL },
{ "provider", 'P', 0, G_OPTION_ARG_STRING, &options.provider,
NULL, NULL },
{ "rsc-id", 'r', 0, G_OPTION_ARG_STRING, &options.rsc_id,
NULL, NULL },
{ "start-delay", 's', 0, G_OPTION_ARG_INT, &options.start_delay,
NULL, NULL },
{ "timeout", 't', 0, G_OPTION_ARG_INT, &options.timeout,
NULL, NULL },
{ "type", 'T', 0, G_OPTION_ARG_STRING, &options.type,
NULL, NULL },
{ NULL }
};
static GMainLoop *mainloop = NULL;
static lrmd_t *lrmd_conn = NULL;
static char event_buf_v0[1024];
static crm_exit_t
test_exit(crm_exit_t exit_code)
{
lrmd_api_delete(lrmd_conn);
return crm_exit(exit_code);
}
#define print_result(fmt, args...) \
if (!options.quiet) { \
printf(fmt "\n" , ##args); \
}
#define report_event(event) \
snprintf(event_buf_v0, sizeof(event_buf_v0), "NEW_EVENT event_type:%s rsc_id:%s action:%s rc:%s op_status:%s", \
lrmd_event_type2str(event->type), \
event->rsc_id, \
event->op_type ? event->op_type : "none", \
services_ocf_exitcode_str(event->rc), \
pcmk_exec_status_str(event->op_status)); \
crm_info("%s", event_buf_v0);
static void
test_shutdown(int nsig)
{
lrmd_api_delete(lrmd_conn);
lrmd_conn = NULL;
}
static void
read_events(lrmd_event_data_t * event)
{
report_event(event);
if (options.listen) {
if (pcmk__str_eq(options.listen, event_buf_v0, pcmk__str_casei)) {
print_result("LISTEN EVENT SUCCESSFUL");
test_exit(CRM_EX_OK);
}
}
if (exec_call_id && (event->call_id == exec_call_id)) {
if (event->op_status == 0 && event->rc == 0) {
print_result("API-CALL SUCCESSFUL for 'exec'");
} else {
print_result("API-CALL FAILURE for 'exec', rc:%d lrmd_op_status:%s",
event->rc, pcmk_exec_status_str(event->op_status));
test_exit(CRM_EX_ERROR);
}
if (!options.listen) {
test_exit(CRM_EX_OK);
}
}
}
static gboolean
timeout_err(gpointer data)
{
print_result("LISTEN EVENT FAILURE - timeout occurred, never found");
test_exit(CRM_EX_TIMEOUT);
return FALSE;
}
static void
connection_events(lrmd_event_data_t * event)
{
int rc = event->connection_rc;
if (event->type != lrmd_event_connect) {
/* ignore */
return;
}
if (!rc) {
crm_info("Executor client connection established");
start_test(NULL);
return;
} else {
sleep(1);
try_connect();
crm_notice("Executor client connection failed");
}
}
static void
try_connect(void)
{
int tries = 10;
static int num_tries = 0;
int rc = 0;
lrmd_conn->cmds->set_callback(lrmd_conn, connection_events);
for (; num_tries < tries; num_tries++) {
rc = lrmd_conn->cmds->connect_async(lrmd_conn, crm_system_name, 3000);
if (!rc) {
return; /* we'll hear back in async callback */
}
sleep(1);
}
print_result("API CONNECTION FAILURE");
test_exit(CRM_EX_ERROR);
}
static gboolean
start_test(gpointer user_data)
{
int rc = 0;
if (!options.no_connect) {
if (!lrmd_conn->cmds->is_connected(lrmd_conn)) {
try_connect();
/* async connect -- this function will get called back into */
return 0;
}
}
lrmd_conn->cmds->set_callback(lrmd_conn, read_events);
if (options.timeout) {
g_timeout_add(options.timeout, timeout_err, NULL);
}
if (!options.api_call) {
return 0;
}
if (pcmk__str_eq(options.api_call, "exec", pcmk__str_casei)) {
rc = lrmd_conn->cmds->exec(lrmd_conn,
options.rsc_id,
options.action,
NULL,
options.interval_ms,
options.timeout,
options.start_delay,
options.exec_call_opts,
options.params);
if (rc > 0) {
exec_call_id = rc;
print_result("API-CALL 'exec' action pending, waiting on response");
}
} else if (pcmk__str_eq(options.api_call, "register_rsc", pcmk__str_casei)) {
rc = lrmd_conn->cmds->register_rsc(lrmd_conn,
options.rsc_id,
options.class, options.provider, options.type, 0);
} else if (pcmk__str_eq(options.api_call, "get_rsc_info", pcmk__str_casei)) {
lrmd_rsc_info_t *rsc_info;
rsc_info = lrmd_conn->cmds->get_rsc_info(lrmd_conn, options.rsc_id, 0);
if (rsc_info) {
print_result("RSC_INFO: id:%s class:%s provider:%s type:%s",
rsc_info->id, rsc_info->standard,
(rsc_info->provider? rsc_info->provider : "<none>"),
rsc_info->type);
lrmd_free_rsc_info(rsc_info);
rc = pcmk_ok;
} else {
rc = -1;
}
} else if (pcmk__str_eq(options.api_call, "unregister_rsc", pcmk__str_casei)) {
rc = lrmd_conn->cmds->unregister_rsc(lrmd_conn, options.rsc_id, 0);
} else if (pcmk__str_eq(options.api_call, "cancel", pcmk__str_casei)) {
rc = lrmd_conn->cmds->cancel(lrmd_conn, options.rsc_id, options.action,
options.interval_ms);
} else if (pcmk__str_eq(options.api_call, "metadata", pcmk__str_casei)) {
char *output = NULL;
rc = lrmd_conn->cmds->get_metadata(lrmd_conn,
options.class,
options.provider, options.type, &output, 0);
if (rc == pcmk_ok) {
print_result("%s", output);
free(output);
}
} else if (pcmk__str_eq(options.api_call, "list_agents", pcmk__str_casei)) {
lrmd_list_t *list = NULL;
lrmd_list_t *iter = NULL;
rc = lrmd_conn->cmds->list_agents(lrmd_conn, &list, options.class, options.provider);
if (rc > 0) {
print_result("%d agents found", rc);
for (iter = list; iter != NULL; iter = iter->next) {
print_result("%s", iter->val);
}
lrmd_list_freeall(list);
rc = 0;
} else {
print_result("API_CALL FAILURE - no agents found");
rc = -1;
}
} else if (pcmk__str_eq(options.api_call, "list_ocf_providers", pcmk__str_casei)) {
lrmd_list_t *list = NULL;
lrmd_list_t *iter = NULL;
rc = lrmd_conn->cmds->list_ocf_providers(lrmd_conn, options.type, &list);
if (rc > 0) {
print_result("%d providers found", rc);
for (iter = list; iter != NULL; iter = iter->next) {
print_result("%s", iter->val);
}
lrmd_list_freeall(list);
rc = 0;
} else {
print_result("API_CALL FAILURE - no providers found");
rc = -1;
}
} else if (pcmk__str_eq(options.api_call, "list_standards", pcmk__str_casei)) {
lrmd_list_t *list = NULL;
lrmd_list_t *iter = NULL;
rc = lrmd_conn->cmds->list_standards(lrmd_conn, &list);
if (rc > 0) {
print_result("%d standards found", rc);
for (iter = list; iter != NULL; iter = iter->next) {
print_result("%s", iter->val);
}
lrmd_list_freeall(list);
rc = 0;
} else {
print_result("API_CALL FAILURE - no providers found");
rc = -1;
}
} else if (pcmk__str_eq(options.api_call, "get_recurring_ops", pcmk__str_casei)) {
GList *op_list = NULL;
GList *op_item = NULL;
rc = lrmd_conn->cmds->get_recurring_ops(lrmd_conn, options.rsc_id, 0, 0,
&op_list);
for (op_item = op_list; op_item != NULL; op_item = op_item->next) {
lrmd_op_info_t *op_info = op_item->data;
print_result("RECURRING_OP: %s_%s_%s timeout=%sms",
op_info->rsc_id, op_info->action,
op_info->interval_ms_s, op_info->timeout_ms_s);
lrmd_free_op_info(op_info);
}
g_list_free(op_list);
} else if (options.api_call) {
print_result("API-CALL FAILURE unknown action '%s'", options.action);
test_exit(CRM_EX_ERROR);
}
if (rc < 0) {
print_result("API-CALL FAILURE for '%s' api_rc:%d",
options.api_call, rc);
test_exit(CRM_EX_ERROR);
}
if (options.api_call && rc == pcmk_ok) {
print_result("API-CALL SUCCESSFUL for '%s'", options.api_call);
if (!options.listen) {
test_exit(CRM_EX_OK);
}
}
if (options.no_wait) {
/* just make the call and exit regardless of anything else. */
test_exit(CRM_EX_OK);
}
return 0;
}
/*!
* \internal
* \brief Generate resource parameters from CIB if none explicitly given
*
* \return Standard Pacemaker return code
*/
static int
generate_params(void)
{
int rc = pcmk_rc_ok;
pcmk_scheduler_t *scheduler = NULL;
xmlNode *cib_xml_copy = NULL;
pcmk_resource_t *rsc = NULL;
GHashTable *params = NULL;
GHashTable *meta = NULL;
GHashTableIter iter;
char *key = NULL;
char *value = NULL;
if (options.params != NULL) {
return pcmk_rc_ok; // User specified parameters explicitly
}
// Retrieve and update CIB
rc = cib__signon_query(NULL, NULL, &cib_xml_copy);
if (rc != pcmk_rc_ok) {
return rc;
}
if (!cli_config_update(&cib_xml_copy, NULL, FALSE)) {
crm_err("Could not update CIB");
return pcmk_rc_cib_corrupt;
}
// Calculate cluster status
scheduler = pe_new_working_set();
if (scheduler == NULL) {
crm_crit("Could not allocate scheduler data");
return ENOMEM;
}
- pe__set_working_set_flags(scheduler,
+ pcmk__set_scheduler_flags(scheduler,
pcmk_sched_no_counts|pcmk_sched_no_compat);
scheduler->input = cib_xml_copy;
scheduler->now = crm_time_new(NULL);
cluster_status(scheduler);
// Find resource in CIB
rsc = pe_find_resource_with_flags(scheduler->resources, options.rsc_id,
pcmk_rsc_match_history
|pcmk_rsc_match_basename);
if (rsc == NULL) {
crm_err("Resource does not exist in config");
pe_free_working_set(scheduler);
return EINVAL;
}
// Add resource instance parameters to options.params
params = pe_rsc_params(rsc, NULL, scheduler);
if (params != NULL) {
g_hash_table_iter_init(&iter, params);
while (g_hash_table_iter_next(&iter, (gpointer *) &key,
(gpointer *) &value)) {
options.params = lrmd_key_value_add(options.params, key, value);
}
}
// Add resource meta-attributes to options.params
meta = pcmk__strkey_table(free, free);
get_meta_attributes(meta, rsc, NULL, scheduler);
g_hash_table_iter_init(&iter, meta);
while (g_hash_table_iter_next(&iter, (gpointer *) &key,
(gpointer *) &value)) {
char *crm_name = crm_meta_name(key);
options.params = lrmd_key_value_add(options.params, crm_name, value);
free(crm_name);
}
g_hash_table_destroy(meta);
pe_free_working_set(scheduler);
return rc;
}
static GOptionContext *
build_arg_context(pcmk__common_args_t *args, GOptionGroup **group) {
GOptionContext *context = NULL;
context = pcmk__build_arg_context(args, NULL, group, NULL);
pcmk__add_main_args(context, basic_entries);
pcmk__add_arg_group(context, "api-call", "API Call Options:",
"Parameters for api-call option", api_call_entries);
return context;
}
int
main(int argc, char **argv)
{
GError *error = NULL;
crm_exit_t exit_code = CRM_EX_OK;
crm_trigger_t *trig = NULL;
pcmk__common_args_t *args = pcmk__new_common_args(SUMMARY);
/* Typically we'd pass all the single character options that take an argument
* as the second parameter here (and there's a bunch of those in this tool).
* However, we control how this program is called so we can just not call it
* in a way where the preprocessing ever matters.
*/
gchar **processed_args = pcmk__cmdline_preproc(argv, NULL);
GOptionContext *context = build_arg_context(args, NULL);
if (!g_option_context_parse_strv(context, &processed_args, &error)) {
exit_code = CRM_EX_USAGE;
goto done;
}
/* We have to use crm_log_init here to set up the logging because there's
* different handling for daemons vs. command line programs, and
* pcmk__cli_init_logging is set up to only handle the latter.
*/
crm_log_init(NULL, LOG_INFO, TRUE, (args->verbosity? TRUE : FALSE), argc,
argv, FALSE);
for (int i = 0; i < args->verbosity; i++) {
crm_bump_log_level(argc, argv);
}
if (!options.listen && pcmk__strcase_any_of(options.api_call, "metadata", "list_agents",
"list_standards", "list_ocf_providers", NULL)) {
options.no_connect = TRUE;
}
if (options.is_running) {
int rc = pcmk_rc_ok;
if (options.rsc_id == NULL) {
exit_code = CRM_EX_USAGE;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"--is-running requires --rsc-id");
goto done;
}
options.interval_ms = 0;
if (options.timeout == 0) {
options.timeout = 30000;
}
rc = generate_params();
if (rc != pcmk_rc_ok) {
exit_code = pcmk_rc2exitc(rc);
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Can not determine resource status: "
"unable to get parameters from CIB");
goto done;
}
options.api_call = "exec";
options.action = PCMK_ACTION_MONITOR;
options.exec_call_opts = lrmd_opt_notify_orig_only;
}
if (!options.api_call && !options.listen) {
exit_code = CRM_EX_USAGE;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Must specify at least one of --api-call, --listen, "
"or --is-running");
goto done;
}
if (options.use_tls) {
lrmd_conn = lrmd_remote_api_new(NULL, "localhost", 0);
} else {
lrmd_conn = lrmd_api_new();
}
trig = mainloop_add_trigger(G_PRIORITY_HIGH, start_test, NULL);
mainloop_set_trigger(trig);
mainloop_add_signal(SIGTERM, test_shutdown);
crm_info("Starting");
mainloop = g_main_loop_new(NULL, FALSE);
g_main_loop_run(mainloop);
done:
g_strfreev(processed_args);
pcmk__free_arg_context(context);
free(key);
free(val);
pcmk__output_and_clear_error(&error, NULL);
return test_exit(exit_code);
}
diff --git a/include/crm/common/scheduler_internal.h b/include/crm/common/scheduler_internal.h
index 57e59eb35b..fa6bec7ed9 100644
--- a/include/crm/common/scheduler_internal.h
+++ b/include/crm/common/scheduler_internal.h
@@ -1,143 +1,169 @@
/*
* Copyright 2004-2023 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU Lesser General Public License
* version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY.
*/
#ifndef PCMK__CRM_COMMON_SCHEDULER_INTERNAL__H
# define PCMK__CRM_COMMON_SCHEDULER_INTERNAL__H
#include <crm/common/action_relation_internal.h>
#include <crm/common/bundles_internal.h>
#include <crm/common/clone_internal.h>
#include <crm/common/digests_internal.h>
#include <crm/common/failcounts_internal.h>
#include <crm/common/group_internal.h>
#include <crm/common/location_internal.h>
#include <crm/common/roles_internal.h>
#ifdef __cplusplus
extern "C" {
#endif
/* Some warnings are too noisy when logged every time a give function is called
* (for example, using a deprecated feature). As an alternative, we allow
* warnings to be logged once per scheduler sequence (transition). Each of those
* warnings needs a flag defined here.
*/
enum pcmk__sched_warnings {
pcmk__wo_blind = (1 << 0),
pcmk__wo_restart_type = (1 << 1),
pcmk__wo_role_after = (1 << 2),
pcmk__wo_poweroff = (1 << 3),
pcmk__wo_require_all = (1 << 4),
pcmk__wo_order_score = (1 << 5),
pcmk__wo_neg_threshold = (1 << 6),
pcmk__wo_remove_after = (1 << 7),
pcmk__wo_ping_node = (1 << 8),
pcmk__wo_order_inst = (1 << 9),
pcmk__wo_coloc_inst = (1 << 10),
pcmk__wo_group_order = (1 << 11),
pcmk__wo_group_coloc = (1 << 12),
pcmk__wo_upstart = (1 << 13),
pcmk__wo_nagios = (1 << 14),
pcmk__wo_set_ordering = (1 << 15),
};
enum pcmk__check_parameters {
/* Clear fail count if parameters changed for un-expired start or monitor
* last_failure.
*/
pcmk__check_last_failure,
/* Clear fail count if parameters changed for start, monitor, promote, or
* migrate_from actions for active resources.
*/
pcmk__check_active,
};
// Group of enum pcmk__sched_warnings flags for warnings we want to log once
extern uint32_t pcmk__warnings;
/*!
* \internal
* \brief Log a resource-tagged message at info severity
*
* \param[in] rsc Tag message with this resource's ID
* \param[in] fmt... printf(3)-style format and arguments
*/
#define pcmk__rsc_info(rsc, fmt, args...) \
crm_log_tag(LOG_INFO, ((rsc) == NULL)? "<NULL>" : (rsc)->id, (fmt), ##args)
/*!
* \internal
* \brief Log a resource-tagged message at debug severity
*
* \param[in] rsc Tag message with this resource's ID
* \param[in] fmt... printf(3)-style format and arguments
*/
#define pcmk__rsc_debug(rsc, fmt, args...) \
crm_log_tag(LOG_DEBUG, ((rsc) == NULL)? "<NULL>" : (rsc)->id, (fmt), ##args)
/*!
* \internal
* \brief Log a resource-tagged message at trace severity
*
* \param[in] rsc Tag message with this resource's ID
* \param[in] fmt... printf(3)-style format and arguments
*/
#define pcmk__rsc_trace(rsc, fmt, args...) \
crm_log_tag(LOG_TRACE, ((rsc) == NULL)? "<NULL>" : (rsc)->id, (fmt), ##args)
/*!
* \internal
* \brief Log an error and remember that current scheduler input has errors
*
* \param[in] fmt... printf(3)-style format and arguments
*/
#define pcmk__sched_err(fmt...) do { \
was_processing_error = TRUE; \
crm_err(fmt); \
} while (0)
/*!
* \internal
* \brief Log a warning and remember that current scheduler input has warnings
*
* \param[in] fmt... printf(3)-style format and arguments
*/
#define pcmk__sched_warn(fmt...) do { \
was_processing_warning = TRUE; \
crm_warn(fmt); \
} while (0)
/*!
* \internal
* \brief Log a warning once per scheduler run
*
* \param[in] wo_flag enum pcmk__sched_warnings value for this warning
* \param[in] fmt... printf(3)-style format and arguments
*/
#define pcmk__warn_once(wo_flag, fmt...) do { \
if (!pcmk_is_set(pcmk__warnings, wo_flag)) { \
if (wo_flag == pcmk__wo_blind) { \
crm_warn(fmt); \
} else { \
pcmk__config_warn(fmt); \
} \
pcmk__warnings = pcmk__set_flags_as(__func__, __LINE__, \
LOG_TRACE, \
"Warn-once", "logging", \
pcmk__warnings, \
(wo_flag), #wo_flag); \
} \
} while (0)
+/*!
+ * \internal
+ * \brief Set scheduler flags
+ *
+ * \param[in,out] scheduler Scheduler data
+ * \param[in] flags_to_set Group of enum pcmk_scheduler_flags to set
+ */
+#define pcmk__set_scheduler_flags(scheduler, flags_to_set) do { \
+ (scheduler)->flags = pcmk__set_flags_as(__func__, __LINE__, \
+ LOG_TRACE, "Scheduler", crm_system_name, \
+ (scheduler)->flags, (flags_to_set), #flags_to_set); \
+ } while (0)
+
+/*!
+ * \internal
+ * \brief Clear scheduler flags
+ *
+ * \param[in,out] scheduler Scheduler data
+ * \param[in] flags_to_clear Group of enum pcmk_scheduler_flags to clear
+ */
+#define pcmk__clear_scheduler_flags(scheduler, flags_to_clear) do { \
+ (scheduler)->flags = pcmk__clear_flags_as(__func__, __LINE__, \
+ LOG_TRACE, "Scheduler", crm_system_name, \
+ (scheduler)->flags, (flags_to_clear), #flags_to_clear); \
+ } while (0)
+
#ifdef __cplusplus
}
#endif
#endif // PCMK__CRM_COMMON_SCHEDULER_INTERNAL__H
diff --git a/include/crm/pengine/internal.h b/include/crm/pengine/internal.h
index 1dfee80018..8e1c2d2abc 100644
--- a/include/crm/pengine/internal.h
+++ b/include/crm/pengine/internal.h
@@ -1,623 +1,611 @@
/*
* 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 PE_INTERNAL__H
# define PE_INTERNAL__H
# include <stdbool.h>
# include <stdint.h>
# include <string.h>
# include <crm/msg_xml.h>
# include <crm/pengine/status.h>
# include <crm/pengine/remote_internal.h>
# include <crm/common/internal.h>
# include <crm/common/options_internal.h>
# include <crm/common/output_internal.h>
# include <crm/common/scheduler_internal.h>
const char *pe__resource_description(const pcmk_resource_t *rsc,
uint32_t show_opts);
bool pe__clone_is_ordered(const pcmk_resource_t *clone);
int pe__set_clone_flag(pcmk_resource_t *clone, enum pcmk__clone_flags flag);
bool pe__clone_flag_is_set(const pcmk_resource_t *clone, uint32_t flags);
bool pe__group_flag_is_set(const pcmk_resource_t *group, uint32_t flags);
pcmk_resource_t *pe__last_group_member(const pcmk_resource_t *group);
-#define pe__set_working_set_flags(scheduler, flags_to_set) do { \
- (scheduler)->flags = pcmk__set_flags_as(__func__, __LINE__, \
- LOG_TRACE, "Scheduler", crm_system_name, \
- (scheduler)->flags, (flags_to_set), #flags_to_set); \
- } while (0)
-
-#define pe__clear_working_set_flags(scheduler, flags_to_clear) do { \
- (scheduler)->flags = pcmk__clear_flags_as(__func__, __LINE__, \
- LOG_TRACE, "Scheduler", crm_system_name, \
- (scheduler)->flags, (flags_to_clear), #flags_to_clear); \
- } while (0)
-
#define pe__set_resource_flags(resource, flags_to_set) do { \
(resource)->flags = pcmk__set_flags_as(__func__, __LINE__, \
LOG_TRACE, "Resource", (resource)->id, (resource)->flags, \
(flags_to_set), #flags_to_set); \
} while (0)
#define pe__clear_resource_flags(resource, flags_to_clear) do { \
(resource)->flags = pcmk__clear_flags_as(__func__, __LINE__, \
LOG_TRACE, "Resource", (resource)->id, (resource)->flags, \
(flags_to_clear), #flags_to_clear); \
} while (0)
#define pe__set_action_flags(action, flags_to_set) do { \
(action)->flags = pcmk__set_flags_as(__func__, __LINE__, \
LOG_TRACE, \
"Action", (action)->uuid, \
(action)->flags, \
(flags_to_set), \
#flags_to_set); \
} while (0)
#define pe__clear_action_flags(action, flags_to_clear) do { \
(action)->flags = pcmk__clear_flags_as(__func__, __LINE__, \
LOG_TRACE, \
"Action", (action)->uuid, \
(action)->flags, \
(flags_to_clear), \
#flags_to_clear); \
} while (0)
#define pe__set_raw_action_flags(action_flags, action_name, flags_to_set) do { \
action_flags = pcmk__set_flags_as(__func__, __LINE__, \
LOG_TRACE, "Action", action_name, \
(action_flags), \
(flags_to_set), #flags_to_set); \
} while (0)
#define pe__clear_raw_action_flags(action_flags, action_name, flags_to_clear) do { \
action_flags = pcmk__clear_flags_as(__func__, __LINE__, \
LOG_TRACE, \
"Action", action_name, \
(action_flags), \
(flags_to_clear), \
#flags_to_clear); \
} while (0)
#define pe__set_action_flags_as(function, line, action, flags_to_set) do { \
(action)->flags = pcmk__set_flags_as((function), (line), \
LOG_TRACE, \
"Action", (action)->uuid, \
(action)->flags, \
(flags_to_set), \
#flags_to_set); \
} while (0)
#define pe__clear_action_flags_as(function, line, action, flags_to_clear) do { \
(action)->flags = pcmk__clear_flags_as((function), (line), \
LOG_TRACE, \
"Action", (action)->uuid, \
(action)->flags, \
(flags_to_clear), \
#flags_to_clear); \
} while (0)
#define pe__set_order_flags(order_flags, flags_to_set) do { \
order_flags = pcmk__set_flags_as(__func__, __LINE__, LOG_TRACE, \
"Ordering", "constraint", \
order_flags, (flags_to_set), \
#flags_to_set); \
} while (0)
#define pe__clear_order_flags(order_flags, flags_to_clear) do { \
order_flags = pcmk__clear_flags_as(__func__, __LINE__, LOG_TRACE, \
"Ordering", "constraint", \
order_flags, (flags_to_clear), \
#flags_to_clear); \
} while (0)
const pcmk_resource_t *pe__const_top_resource(const pcmk_resource_t *rsc,
bool include_bundle);
int pe__clone_max(const pcmk_resource_t *clone);
int pe__clone_node_max(const pcmk_resource_t *clone);
int pe__clone_promoted_max(const pcmk_resource_t *clone);
int pe__clone_promoted_node_max(const pcmk_resource_t *clone);
void pe__create_clone_notifications(pcmk_resource_t *clone);
void pe__free_clone_notification_data(pcmk_resource_t *clone);
void pe__create_clone_notif_pseudo_ops(pcmk_resource_t *clone,
pcmk_action_t *start,
pcmk_action_t *started,
pcmk_action_t *stop,
pcmk_action_t *stopped);
pcmk_action_t *pe__new_rsc_pseudo_action(pcmk_resource_t *rsc, const char *task,
bool optional, bool runnable);
void pe__create_promotable_pseudo_ops(pcmk_resource_t *clone,
bool any_promoting, bool any_demoting);
bool pe_can_fence(const pcmk_scheduler_t *scheduler, const pcmk_node_t *node);
void add_hash_param(GHashTable * hash, const char *name, const char *value);
char *native_parameter(pcmk_resource_t *rsc, pcmk_node_t *node, gboolean create,
const char *name, pcmk_scheduler_t *scheduler);
pcmk_node_t *native_location(const pcmk_resource_t *rsc, GList **list,
int current);
void pe_metadata(pcmk__output_t *out);
void verify_pe_options(GHashTable * options);
void native_add_running(pcmk_resource_t *rsc, pcmk_node_t *node,
pcmk_scheduler_t *scheduler, gboolean failed);
gboolean native_unpack(pcmk_resource_t *rsc, pcmk_scheduler_t *scheduler);
gboolean group_unpack(pcmk_resource_t *rsc, pcmk_scheduler_t *scheduler);
gboolean clone_unpack(pcmk_resource_t *rsc, pcmk_scheduler_t *scheduler);
gboolean pe__unpack_bundle(pcmk_resource_t *rsc, pcmk_scheduler_t *scheduler);
pcmk_resource_t *native_find_rsc(pcmk_resource_t *rsc, const char *id,
const pcmk_node_t *node, int flags);
gboolean native_active(pcmk_resource_t *rsc, gboolean all);
gboolean group_active(pcmk_resource_t *rsc, gboolean all);
gboolean clone_active(pcmk_resource_t *rsc, gboolean all);
gboolean pe__bundle_active(pcmk_resource_t *rsc, gboolean all);
//! \deprecated This function will be removed in a future release
void native_print(pcmk_resource_t *rsc, const char *pre_text, long options,
void *print_data);
//! \deprecated This function will be removed in a future release
void group_print(pcmk_resource_t *rsc, const char *pre_text, long options,
void *print_data);
//! \deprecated This function will be removed in a future release
void clone_print(pcmk_resource_t *rsc, const char *pre_text, long options,
void *print_data);
//! \deprecated This function will be removed in a future release
void pe__print_bundle(pcmk_resource_t *rsc, const char *pre_text, long options,
void *print_data);
gchar *pcmk__native_output_string(const pcmk_resource_t *rsc, const char *name,
const pcmk_node_t *node, uint32_t show_opts,
const char *target_role, bool show_nodes);
int pe__name_and_nvpairs_xml(pcmk__output_t *out, bool is_list, const char *tag_name
, size_t pairs_count, ...);
char *pe__node_display_name(pcmk_node_t *node, bool print_detail);
// Clone notifications (pe_notif.c)
void pe__order_notifs_after_fencing(const pcmk_action_t *action,
pcmk_resource_t *rsc,
pcmk_action_t *stonith_op);
static inline const char *
pe__rsc_bool_str(const pcmk_resource_t *rsc, uint64_t rsc_flag)
{
return pcmk__btoa(pcmk_is_set(rsc->flags, rsc_flag));
}
int pe__clone_xml(pcmk__output_t *out, va_list args);
int pe__clone_default(pcmk__output_t *out, va_list args);
int pe__group_xml(pcmk__output_t *out, va_list args);
int pe__group_default(pcmk__output_t *out, va_list args);
int pe__bundle_xml(pcmk__output_t *out, va_list args);
int pe__bundle_html(pcmk__output_t *out, va_list args);
int pe__bundle_text(pcmk__output_t *out, va_list args);
int pe__node_html(pcmk__output_t *out, va_list args);
int pe__node_text(pcmk__output_t *out, va_list args);
int pe__node_xml(pcmk__output_t *out, va_list args);
int pe__resource_xml(pcmk__output_t *out, va_list args);
int pe__resource_html(pcmk__output_t *out, va_list args);
int pe__resource_text(pcmk__output_t *out, va_list args);
void native_free(pcmk_resource_t *rsc);
void group_free(pcmk_resource_t *rsc);
void clone_free(pcmk_resource_t *rsc);
void pe__free_bundle(pcmk_resource_t *rsc);
enum rsc_role_e native_resource_state(const pcmk_resource_t *rsc,
gboolean current);
enum rsc_role_e group_resource_state(const pcmk_resource_t *rsc,
gboolean current);
enum rsc_role_e clone_resource_state(const pcmk_resource_t *rsc,
gboolean current);
enum rsc_role_e pe__bundle_resource_state(const pcmk_resource_t *rsc,
gboolean current);
void pe__count_common(pcmk_resource_t *rsc);
void pe__count_bundle(pcmk_resource_t *rsc);
void common_free(pcmk_resource_t *rsc);
pcmk_node_t *pe__copy_node(const pcmk_node_t *this_node);
time_t get_effective_time(pcmk_scheduler_t *scheduler);
/* Failure handling utilities (from failcounts.c) */
int pe_get_failcount(const pcmk_node_t *node, pcmk_resource_t *rsc,
time_t *last_failure, uint32_t flags,
const xmlNode *xml_op);
pcmk_action_t *pe__clear_failcount(pcmk_resource_t *rsc,
const pcmk_node_t *node, const char *reason,
pcmk_scheduler_t *scheduler);
/* Functions for finding/counting a resource's active nodes */
bool pe__count_active_node(const pcmk_resource_t *rsc, pcmk_node_t *node,
pcmk_node_t **active, unsigned int *count_all,
unsigned int *count_clean);
pcmk_node_t *pe__find_active_requires(const pcmk_resource_t *rsc,
unsigned int *count);
static inline pcmk_node_t *
pe__current_node(const pcmk_resource_t *rsc)
{
return (rsc == NULL)? NULL : rsc->fns->active_node(rsc, NULL, NULL);
}
/* Binary like operators for lists of nodes */
GHashTable *pe__node_list2table(const GList *list);
pcmk_action_t *get_pseudo_op(const char *name, pcmk_scheduler_t *scheduler);
gboolean order_actions(pcmk_action_t *lh_action, pcmk_action_t *rh_action,
uint32_t flags);
void pe__show_node_scores_as(const char *file, const char *function,
int line, bool to_log, const pcmk_resource_t *rsc,
const char *comment, GHashTable *nodes,
pcmk_scheduler_t *scheduler);
#define pe__show_node_scores(level, rsc, text, nodes, scheduler) \
pe__show_node_scores_as(__FILE__, __func__, __LINE__, \
(level), (rsc), (text), (nodes), (scheduler))
GHashTable *pcmk__unpack_action_meta(pcmk_resource_t *rsc,
const pcmk_node_t *node,
const char *action_name, guint interval_ms,
const xmlNode *action_config);
GHashTable *pcmk__unpack_action_rsc_params(const xmlNode *action_xml,
GHashTable *node_attrs,
pcmk_scheduler_t *data_set);
xmlNode *pcmk__find_action_config(const pcmk_resource_t *rsc,
const char *action_name, guint interval_ms,
bool include_disabled);
enum rsc_start_requirement pcmk__action_requires(const pcmk_resource_t *rsc,
const char *action_name);
enum action_fail_response pcmk__parse_on_fail(const pcmk_resource_t *rsc,
const char *action_name,
guint interval_ms,
const char *value);
enum rsc_role_e pcmk__role_after_failure(const pcmk_resource_t *rsc,
const char *action_name,
enum action_fail_response on_fail,
GHashTable *meta);
pcmk_action_t *custom_action(pcmk_resource_t *rsc, char *key, const char *task,
const pcmk_node_t *on_node, gboolean optional,
pcmk_scheduler_t *scheduler);
# define delete_key(rsc) pcmk__op_key(rsc->id, PCMK_ACTION_DELETE, 0)
# define delete_action(rsc, node, optional) custom_action( \
rsc, delete_key(rsc), PCMK_ACTION_DELETE, node, \
optional, rsc->cluster);
# define stop_key(rsc) pcmk__op_key(rsc->id, PCMK_ACTION_STOP, 0)
# define stop_action(rsc, node, optional) custom_action( \
rsc, stop_key(rsc), PCMK_ACTION_STOP, node, \
optional, rsc->cluster);
# define reload_key(rsc) pcmk__op_key(rsc->id, PCMK_ACTION_RELOAD_AGENT, 0)
# define start_key(rsc) pcmk__op_key(rsc->id, PCMK_ACTION_START, 0)
# define start_action(rsc, node, optional) custom_action( \
rsc, start_key(rsc), PCMK_ACTION_START, node, \
optional, rsc->cluster)
# define promote_key(rsc) pcmk__op_key(rsc->id, PCMK_ACTION_PROMOTE, 0)
# define promote_action(rsc, node, optional) custom_action( \
rsc, promote_key(rsc), PCMK_ACTION_PROMOTE, node, \
optional, rsc->cluster)
# define demote_key(rsc) pcmk__op_key(rsc->id, PCMK_ACTION_DEMOTE, 0)
# define demote_action(rsc, node, optional) custom_action( \
rsc, demote_key(rsc), PCMK_ACTION_DEMOTE, node, \
optional, rsc->cluster)
extern int pe_get_configured_timeout(pcmk_resource_t *rsc, const char *action,
pcmk_scheduler_t *scheduler);
pcmk_action_t *find_first_action(const GList *input, const char *uuid,
const char *task, const pcmk_node_t *on_node);
enum action_tasks get_complex_task(const pcmk_resource_t *rsc,
const char *name);
GList *find_actions(GList *input, const char *key, const pcmk_node_t *on_node);
GList *find_actions_exact(GList *input, const char *key,
const pcmk_node_t *on_node);
GList *pe__resource_actions(const pcmk_resource_t *rsc, const pcmk_node_t *node,
const char *task, bool require_node);
extern void pe_free_action(pcmk_action_t *action);
void resource_location(pcmk_resource_t *rsc, const pcmk_node_t *node, int score,
const char *tag, pcmk_scheduler_t *scheduler);
extern int pe__is_newer_op(const xmlNode *xml_a, const xmlNode *xml_b,
bool same_node_default);
extern gint sort_op_by_callid(gconstpointer a, gconstpointer b);
gboolean get_target_role(const pcmk_resource_t *rsc, enum rsc_role_e *role);
void pe__set_next_role(pcmk_resource_t *rsc, enum rsc_role_e role,
const char *why);
pcmk_resource_t *find_clone_instance(const pcmk_resource_t *rsc,
const char *sub_id);
extern void destroy_ticket(gpointer data);
pcmk_ticket_t *ticket_new(const char *ticket_id, pcmk_scheduler_t *scheduler);
// Resources for manipulating resource names
const char *pe_base_name_end(const char *id);
char *clone_strip(const char *last_rsc_id);
char *clone_zero(const char *last_rsc_id);
static inline bool
pe_base_name_eq(const pcmk_resource_t *rsc, const char *id)
{
if (id && rsc && rsc->id) {
// Number of characters in rsc->id before any clone suffix
size_t base_len = pe_base_name_end(rsc->id) - rsc->id + 1;
return (strlen(id) == base_len) && !strncmp(id, rsc->id, base_len);
}
return false;
}
int pe__target_rc_from_xml(const xmlNode *xml_op);
gint pe__cmp_node_name(gconstpointer a, gconstpointer b);
bool is_set_recursive(const pcmk_resource_t *rsc, long long flag, bool any);
pcmk__op_digest_t *pe__calculate_digests(pcmk_resource_t *rsc, const char *task,
guint *interval_ms,
const pcmk_node_t *node,
const xmlNode *xml_op,
GHashTable *overrides,
bool calc_secure,
pcmk_scheduler_t *scheduler);
void pe__free_digests(gpointer ptr);
pcmk__op_digest_t *rsc_action_digest_cmp(pcmk_resource_t *rsc,
const xmlNode *xml_op,
pcmk_node_t *node,
pcmk_scheduler_t *scheduler);
pcmk_action_t *pe_fence_op(pcmk_node_t *node, const char *op, bool optional,
const char *reason, bool priority_delay,
pcmk_scheduler_t *scheduler);
void trigger_unfencing(pcmk_resource_t *rsc, pcmk_node_t *node,
const char *reason, pcmk_action_t *dependency,
pcmk_scheduler_t *scheduler);
char *pe__action2reason(const pcmk_action_t *action, enum pe_action_flags flag);
void pe_action_set_reason(pcmk_action_t *action, const char *reason,
bool overwrite);
void pe__add_action_expected_result(pcmk_action_t *action, int expected_result);
void pe__set_resource_flags_recursive(pcmk_resource_t *rsc, uint64_t flags);
void pe__clear_resource_flags_recursive(pcmk_resource_t *rsc, uint64_t flags);
void pe__clear_resource_flags_on_all(pcmk_scheduler_t *scheduler,
uint64_t flag);
gboolean add_tag_ref(GHashTable * tags, const char * tag_name, const char * obj_ref);
//! \deprecated This function will be removed in a future release
void print_rscs_brief(GList *rsc_list, const char * pre_text, long options,
void * print_data, gboolean print_all);
int pe__rscs_brief_output(pcmk__output_t *out, GList *rsc_list, unsigned int options);
void pe_fence_node(pcmk_scheduler_t *scheduler, pcmk_node_t *node,
const char *reason, bool priority_delay);
pcmk_node_t *pe_create_node(const char *id, const char *uname, const char *type,
const char *score, pcmk_scheduler_t *scheduler);
//! \deprecated This function will be removed in a future release
void common_print(pcmk_resource_t *rsc, const char *pre_text, const char *name,
const pcmk_node_t *node, long options, void *print_data);
int pe__common_output_text(pcmk__output_t *out, const pcmk_resource_t *rsc,
const char *name, const pcmk_node_t *node,
unsigned int options);
int pe__common_output_html(pcmk__output_t *out, const pcmk_resource_t *rsc,
const char *name, const pcmk_node_t *node,
unsigned int options);
GList *pe__bundle_containers(const pcmk_resource_t *bundle);
int pe__bundle_max(const pcmk_resource_t *rsc);
bool pe__node_is_bundle_instance(const pcmk_resource_t *bundle,
const pcmk_node_t *node);
pcmk_resource_t *pe__bundled_resource(const pcmk_resource_t *rsc);
const pcmk_resource_t *pe__get_rsc_in_container(const pcmk_resource_t *instance);
pcmk_resource_t *pe__first_container(const pcmk_resource_t *bundle);
void pe__foreach_bundle_replica(pcmk_resource_t *bundle,
bool (*fn)(pcmk__bundle_replica_t *, void *),
void *user_data);
void pe__foreach_const_bundle_replica(const pcmk_resource_t *bundle,
bool (*fn)(const pcmk__bundle_replica_t *,
void *),
void *user_data);
pcmk_resource_t *pe__find_bundle_replica(const pcmk_resource_t *bundle,
const pcmk_node_t *node);
bool pe__bundle_needs_remote_name(pcmk_resource_t *rsc);
const char *pe__add_bundle_remote_name(pcmk_resource_t *rsc,
pcmk_scheduler_t *scheduler,
xmlNode *xml, const char *field);
const char *pe__node_attribute_calculated(const pcmk_node_t *node,
const char *name,
const pcmk_resource_t *rsc,
enum pcmk__rsc_node node_type,
bool force_host);
const char *pe_node_attribute_raw(const pcmk_node_t *node, const char *name);
bool pe__is_universal_clone(const pcmk_resource_t *rsc,
const pcmk_scheduler_t *scheduler);
void pe__add_param_check(const xmlNode *rsc_op, pcmk_resource_t *rsc,
pcmk_node_t *node, enum pcmk__check_parameters,
pcmk_scheduler_t *scheduler);
void pe__foreach_param_check(pcmk_scheduler_t *scheduler,
void (*cb)(pcmk_resource_t*, pcmk_node_t*,
const xmlNode*,
enum pcmk__check_parameters));
void pe__free_param_checks(pcmk_scheduler_t *scheduler);
bool pe__shutdown_requested(const pcmk_node_t *node);
void pe__update_recheck_time(time_t recheck, pcmk_scheduler_t *scheduler,
const char *reason);
/*!
* \internal
* \brief Register xml formatting message functions.
*
* \param[in,out] out Output object to register messages with
*/
void pe__register_messages(pcmk__output_t *out);
void pe__unpack_dataset_nvpairs(const xmlNode *xml_obj, const char *set_name,
const pe_rule_eval_data_t *rule_data,
GHashTable *hash, const char *always_first,
gboolean overwrite,
pcmk_scheduler_t *scheduler);
bool pe__resource_is_disabled(const pcmk_resource_t *rsc);
void pe__clear_resource_history(pcmk_resource_t *rsc, const pcmk_node_t *node);
GList *pe__rscs_with_tag(pcmk_scheduler_t *scheduler, const char *tag_name);
GList *pe__unames_with_tag(pcmk_scheduler_t *scheduler, const char *tag_name);
bool pe__rsc_has_tag(pcmk_scheduler_t *scheduler, const char *rsc,
const char *tag);
bool pe__uname_has_tag(pcmk_scheduler_t *scheduler, const char *node,
const char *tag);
bool pe__rsc_running_on_only(const pcmk_resource_t *rsc,
const pcmk_node_t *node);
bool pe__rsc_running_on_any(pcmk_resource_t *rsc, GList *node_list);
GList *pe__filter_rsc_list(GList *rscs, GList *filter);
GList * pe__build_node_name_list(pcmk_scheduler_t *scheduler, const char *s);
GList * pe__build_rsc_list(pcmk_scheduler_t *scheduler, const char *s);
bool pcmk__rsc_filtered_by_node(pcmk_resource_t *rsc, GList *only_node);
gboolean pe__bundle_is_filtered(const pcmk_resource_t *rsc, GList *only_rsc,
gboolean check_parent);
gboolean pe__clone_is_filtered(const pcmk_resource_t *rsc, GList *only_rsc,
gboolean check_parent);
gboolean pe__group_is_filtered(const pcmk_resource_t *rsc, GList *only_rsc,
gboolean check_parent);
gboolean pe__native_is_filtered(const pcmk_resource_t *rsc, GList *only_rsc,
gboolean check_parent);
xmlNode *pe__failed_probe_for_rsc(const pcmk_resource_t *rsc, const char *name);
const char *pe__clone_child_id(const pcmk_resource_t *rsc);
int pe__sum_node_health_scores(const pcmk_node_t *node, int base_health);
int pe__node_health(pcmk_node_t *node);
static inline enum pcmk__health_strategy
pe__health_strategy(pcmk_scheduler_t *scheduler)
{
return pcmk__parse_health_strategy(pe_pref(scheduler->config_hash,
PCMK_OPT_NODE_HEALTH_STRATEGY));
}
static inline int
pe__health_score(const char *option, pcmk_scheduler_t *scheduler)
{
return char2score(pe_pref(scheduler->config_hash, option));
}
/*!
* \internal
* \brief Return a string suitable for logging as a node name
*
* \param[in] node Node to return a node name string for
*
* \return Node name if available, otherwise node ID if available,
* otherwise "unspecified node" if node is NULL or "unidentified node"
* if node has neither a name nor ID.
*/
static inline const char *
pe__node_name(const pcmk_node_t *node)
{
if (node == NULL) {
return "unspecified node";
} else if (node->details->uname != NULL) {
return node->details->uname;
} else if (node->details->id != NULL) {
return node->details->id;
} else {
return "unidentified node";
}
}
/*!
* \internal
* \brief Check whether two node objects refer to the same node
*
* \param[in] node1 First node object to compare
* \param[in] node2 Second node object to compare
*
* \return true if \p node1 and \p node2 refer to the same node
*/
static inline bool
pe__same_node(const pcmk_node_t *node1, const pcmk_node_t *node2)
{
return (node1 != NULL) && (node2 != NULL)
&& (node1->details == node2->details);
}
/*!
* \internal
* \brief Get the operation key from an action history entry
*
* \param[in] xml Action history entry
*
* \return Entry's operation key
*/
static inline const char *
pe__xe_history_key(const xmlNode *xml)
{
if (xml == NULL) {
return NULL;
} else {
/* @COMPAT Pacemaker <= 1.1.5 did not add the key, and used the ID
* instead. Checking for that allows us to process old saved CIBs,
* including some regression tests.
*/
const char *key = crm_element_value(xml, XML_LRM_ATTR_TASK_KEY);
return pcmk__str_empty(key)? ID(xml) : key;
}
}
#endif
diff --git a/lib/pacemaker/pcmk_rule.c b/lib/pacemaker/pcmk_rule.c
index 68b1b74724..0c673f8a05 100644
--- a/lib/pacemaker/pcmk_rule.c
+++ b/lib/pacemaker/pcmk_rule.c
@@ -1,296 +1,296 @@
/*
* Copyright 2022-2024 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 <crm/cib/internal.h>
#include <crm/common/cib.h>
#include <crm/common/iso8601.h>
#include <crm/msg_xml.h>
#include <crm/pengine/internal.h>
#include <crm/pengine/rules_internal.h>
#include <pacemaker-internal.h>
/*!
* \internal
* \brief Evaluate a date expression for a specific time
*
* \param[in] expr date_expression XML
* \param[in] now Time for which to evaluate expression
*
* \return Standard Pacemaker return code
*/
static int
eval_date_expression(const xmlNode *expr, crm_time_t *now)
{
pe_rule_eval_data_t rule_data = {
.node_hash = NULL,
.role = pcmk_role_unknown,
.now = now,
.match_data = NULL,
.rsc_data = NULL,
.op_data = NULL
};
return pe__eval_date_expr(expr, &rule_data, NULL);
}
/*!
* \internal
* \brief Initialize scheduler data for checking rules
*
* Make our own copies of the CIB XML and date/time object, if they're not
* \c NULL. This way we don't have to take ownership of the objects passed via
* the API.
*
* \param[in,out] out Output object
* \param[in] input The CIB XML to check (if \c NULL, use current CIB)
* \param[in] date Check whether the rule is in effect at this date
* and time (if \c NULL, use current date and time)
* \param[out] scheduler Where to store initialized scheduler data
*
* \return Standard Pacemaker return code
*/
static int
init_rule_check(pcmk__output_t *out, xmlNodePtr input, const crm_time_t *date,
pcmk_scheduler_t **scheduler)
{
// Allows for cleaner syntax than dereferencing the scheduler argument
pcmk_scheduler_t *new_scheduler = NULL;
new_scheduler = pe_new_working_set();
if (new_scheduler == NULL) {
return ENOMEM;
}
- pe__set_working_set_flags(new_scheduler,
+ pcmk__set_scheduler_flags(new_scheduler,
pcmk_sched_no_counts|pcmk_sched_no_compat);
// Populate the scheduler data
// Make our own copy of the given input or fetch the CIB and use that
if (input != NULL) {
new_scheduler->input = copy_xml(input);
if (new_scheduler->input == NULL) {
out->err(out, "Failed to copy input XML");
pe_free_working_set(new_scheduler);
return ENOMEM;
}
} else {
int rc = cib__signon_query(out, NULL, &(new_scheduler->input));
if (rc != pcmk_rc_ok) {
pe_free_working_set(new_scheduler);
return rc;
}
}
// Make our own copy of the given crm_time_t object; otherwise
// cluster_status() populates with the current time
if (date != NULL) {
// pcmk_copy_time() guarantees non-NULL
new_scheduler->now = pcmk_copy_time(date);
}
// Unpack everything
cluster_status(new_scheduler);
*scheduler = new_scheduler;
return pcmk_rc_ok;
}
#define XPATH_NODE_RULE "//" XML_TAG_RULE "[@" PCMK_XA_ID "='%s']"
/*!
* \internal
* \brief Check whether a given rule is in effect
*
* \param[in] scheduler Scheduler data
* \param[in] rule_id The ID of the rule to check
* \param[out] error Where to store a rule evaluation error message
*
* \return Standard Pacemaker return code
*/
static int
eval_rule(pcmk_scheduler_t *scheduler, const char *rule_id, const char **error)
{
xmlNodePtr cib_constraints = NULL;
xmlNodePtr match = NULL;
xmlXPathObjectPtr xpath_obj = NULL;
char *xpath = NULL;
int rc = pcmk_rc_ok;
int num_results = 0;
*error = NULL;
/* Rules are under the constraints node in the XML, so first find that. */
cib_constraints = pcmk_find_cib_element(scheduler->input,
XML_CIB_TAG_CONSTRAINTS);
/* Get all rules matching the given ID that are also simple enough for us
* to check. For the moment, these rules must only have a single
* date_expression child and:
* - Do not have a date_spec operation, or
* - Have a date_spec operation that contains years= but does not contain
* moon=.
*
* We do this in steps to provide better error messages. First, check that
* there's any rule with the given ID.
*/
xpath = crm_strdup_printf(XPATH_NODE_RULE, rule_id);
xpath_obj = xpath_search(cib_constraints, xpath);
num_results = numXpathResults(xpath_obj);
free(xpath);
freeXpathObject(xpath_obj);
if (num_results == 0) {
*error = "Rule not found";
return ENXIO;
}
if (num_results > 1) {
// Should not be possible; schema prevents this
*error = "Found more than one rule with matching ID";
return pcmk_rc_duplicate_id;
}
/* Next, make sure it has exactly one date_expression. */
xpath = crm_strdup_printf(XPATH_NODE_RULE "//date_expression", rule_id);
xpath_obj = xpath_search(cib_constraints, xpath);
num_results = numXpathResults(xpath_obj);
free(xpath);
freeXpathObject(xpath_obj);
if (num_results != 1) {
if (num_results == 0) {
*error = "Rule does not have a date expression";
} else {
*error = "Rule has more than one date expression";
}
return EOPNOTSUPP;
}
/* Then, check that it's something we actually support. */
xpath = crm_strdup_printf(XPATH_NODE_RULE "//date_expression["
"@" XML_EXPR_ATTR_OPERATION "!='date_spec']",
rule_id);
xpath_obj = xpath_search(cib_constraints, xpath);
num_results = numXpathResults(xpath_obj);
free(xpath);
if (num_results == 0) {
freeXpathObject(xpath_obj);
xpath = crm_strdup_printf(XPATH_NODE_RULE "//date_expression["
"@" XML_EXPR_ATTR_OPERATION "='date_spec' "
"and date_spec/@years "
"and not(date_spec/@moon)]", rule_id);
xpath_obj = xpath_search(cib_constraints, xpath);
num_results = numXpathResults(xpath_obj);
free(xpath);
if (num_results == 0) {
freeXpathObject(xpath_obj);
*error = "Rule must either not use date_spec, or use date_spec "
"with years= but not moon=";
return EOPNOTSUPP;
}
}
match = getXpathResult(xpath_obj, 0);
/* We should have ensured this with the xpath query above, but double-
* checking can't hurt.
*/
CRM_ASSERT(match != NULL);
CRM_ASSERT(find_expression_type(match) == pcmk__subexpr_datetime);
rc = eval_date_expression(match, scheduler->now);
if (rc == pcmk_rc_undetermined) {
/* pe__eval_date_expr() should return this only if something is
* malformed or missing
*/
*error = "Error parsing rule";
}
freeXpathObject(xpath_obj);
return rc;
}
/*!
* \internal
* \brief Check whether each rule in a list is in effect
*
* \param[in,out] out Output object
* \param[in] input The CIB XML to check (if \c NULL, use current CIB)
* \param[in] date Check whether the rule is in effect at this date and
* time (if \c NULL, use current date and time)
* \param[in] rule_ids The IDs of the rules to check, as a <tt>NULL</tt>-
* terminated list.
*
* \return Standard Pacemaker return code
*/
int
pcmk__check_rules(pcmk__output_t *out, xmlNodePtr input, const crm_time_t *date,
const char **rule_ids)
{
pcmk_scheduler_t *scheduler = NULL;
int rc = pcmk_rc_ok;
CRM_ASSERT(out != NULL);
if (rule_ids == NULL) {
// Trivial case; every rule specified is in effect
return pcmk_rc_ok;
}
rc = init_rule_check(out, input, date, &scheduler);
if (rc != pcmk_rc_ok) {
return rc;
}
for (const char **rule_id = rule_ids; *rule_id != NULL; rule_id++) {
const char *error = NULL;
int last_rc = eval_rule(scheduler, *rule_id, &error);
out->message(out, "rule-check", *rule_id, last_rc, error);
if (last_rc != pcmk_rc_ok) {
rc = last_rc;
}
}
pe_free_working_set(scheduler);
return rc;
}
// Documented in pacemaker.h
int
pcmk_check_rules(xmlNodePtr *xml, xmlNodePtr input, const crm_time_t *date,
const char **rule_ids)
{
pcmk__output_t *out = NULL;
int rc = pcmk_rc_ok;
rc = pcmk__xml_output_new(&out, xml);
if (rc != pcmk_rc_ok) {
return rc;
}
pcmk__register_lib_messages(out);
rc = pcmk__check_rules(out, input, date, rule_ids);
pcmk__xml_output_finish(out, pcmk_rc2exitc(rc), xml);
return rc;
}
diff --git a/lib/pacemaker/pcmk_scheduler.c b/lib/pacemaker/pcmk_scheduler.c
index 39171d2789..d9e0162dde 100644
--- a/lib/pacemaker/pcmk_scheduler.c
+++ b/lib/pacemaker/pcmk_scheduler.c
@@ -1,821 +1,821 @@
/*
* 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 <crm/crm.h>
#include <crm/cib.h>
#include <crm/msg_xml.h>
#include <crm/common/xml.h>
#include <crm/common/xml_internal.h>
#include <crm/common/scheduler_internal.h>
#include <glib.h>
#include <crm/pengine/status.h>
#include <pacemaker-internal.h>
#include "libpacemaker_private.h"
CRM_TRACE_INIT_DATA(pacemaker);
/*!
* \internal
* \brief Do deferred action checks after assignment
*
* When unpacking the resource history, the scheduler checks for resource
* configurations that have changed since an action was run. However, at that
* time, bundles using the REMOTE_CONTAINER_HACK don't have their final
* parameter information, so instead they add a deferred check to a list. This
* function processes one entry in that list.
*
* \param[in,out] rsc Resource that action history is for
* \param[in,out] node Node that action history is for
* \param[in] rsc_op Action history entry
* \param[in] check Type of deferred check to do
*/
static void
check_params(pcmk_resource_t *rsc, pcmk_node_t *node, const xmlNode *rsc_op,
enum pcmk__check_parameters check)
{
const char *reason = NULL;
pcmk__op_digest_t *digest_data = NULL;
switch (check) {
case pcmk__check_active:
if (pcmk__check_action_config(rsc, node, rsc_op)
&& pe_get_failcount(node, rsc, NULL, pcmk__fc_effective,
NULL)) {
reason = "action definition changed";
}
break;
case pcmk__check_last_failure:
digest_data = rsc_action_digest_cmp(rsc, rsc_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, ID(rsc_op), node->details->id);
break;
case pcmk__digest_match:
break;
default:
reason = "resource parameters have changed";
break;
}
break;
}
if (reason != NULL) {
pe__clear_failcount(rsc, node, reason, rsc->cluster);
}
}
/*!
* \internal
* \brief Check whether a resource has failcount clearing scheduled on a node
*
* \param[in] node Node to check
* \param[in] rsc Resource to check
*
* \return true if \p rsc has failcount clearing scheduled on \p node,
* otherwise false
*/
static bool
failcount_clear_action_exists(const pcmk_node_t *node,
const pcmk_resource_t *rsc)
{
GList *list = pe__resource_actions(rsc, node, PCMK_ACTION_CLEAR_FAILCOUNT,
TRUE);
if (list != NULL) {
g_list_free(list);
return true;
}
return false;
}
/*!
* \internal
* \brief Ban a resource from a node if it reached its failure threshold there
*
* \param[in,out] data Resource to check failure threshold for
* \param[in] user_data Node to check resource on
*/
static void
check_failure_threshold(gpointer data, gpointer user_data)
{
pcmk_resource_t *rsc = data;
const pcmk_node_t *node = user_data;
// If this is a collective resource, apply recursively to children instead
if (rsc->children != NULL) {
g_list_foreach(rsc->children, check_failure_threshold, user_data);
return;
}
if (!failcount_clear_action_exists(node, rsc)) {
/* Don't force the resource away from this node due to a failcount
* that's going to be cleared.
*
* @TODO Failcount clearing can be scheduled in
* pcmk__handle_rsc_config_changes() via process_rsc_history(), or in
* schedule_resource_actions() via check_params(). This runs well before
* then, so it cannot detect those, meaning we might check the migration
* threshold when we shouldn't. Worst case, we stop or move the
* resource, then move it back in the next transition.
*/
pcmk_resource_t *failed = NULL;
if (pcmk__threshold_reached(rsc, node, &failed)) {
resource_location(failed, node, -INFINITY, "__fail_limit__",
rsc->cluster);
}
}
}
/*!
* \internal
* \brief If resource has exclusive discovery, ban node if not allowed
*
* Location constraints have a resource-discovery option that allows users to
* specify where probes are done for the affected resource. If this is set to
* exclusive, probes will only be done on nodes listed in exclusive constraints.
* This function bans the resource from the node if the node is not listed.
*
* \param[in,out] data Resource to check
* \param[in] user_data Node to check resource on
*/
static void
apply_exclusive_discovery(gpointer data, gpointer user_data)
{
pcmk_resource_t *rsc = data;
const pcmk_node_t *node = user_data;
if (rsc->exclusive_discover
|| pe__const_top_resource(rsc, false)->exclusive_discover) {
pcmk_node_t *match = NULL;
// If this is a collective resource, apply recursively to children
g_list_foreach(rsc->children, apply_exclusive_discovery, user_data);
match = g_hash_table_lookup(rsc->allowed_nodes, node->details->id);
if ((match != NULL)
&& (match->rsc_discover_mode != pcmk_probe_exclusive)) {
match->weight = -INFINITY;
}
}
}
/*!
* \internal
* \brief Apply stickiness to a resource if appropriate
*
* \param[in,out] data Resource to check for stickiness
* \param[in] user_data Ignored
*/
static void
apply_stickiness(gpointer data, gpointer user_data)
{
pcmk_resource_t *rsc = data;
pcmk_node_t *node = NULL;
// If this is a collective resource, apply recursively to children instead
if (rsc->children != NULL) {
g_list_foreach(rsc->children, apply_stickiness, NULL);
return;
}
/* A resource is sticky if it is managed, has stickiness configured, and is
* active on a single node.
*/
if (!pcmk_is_set(rsc->flags, pcmk_rsc_managed)
|| (rsc->stickiness < 1) || !pcmk__list_of_1(rsc->running_on)) {
return;
}
node = rsc->running_on->data;
/* In a symmetric cluster, stickiness can always be used. In an
* asymmetric cluster, we have to check whether the resource is still
* allowed on the node, so we don't keep the resource somewhere it is no
* longer explicitly enabled.
*/
if (!pcmk_is_set(rsc->cluster->flags, pcmk_sched_symmetric_cluster)
&& (g_hash_table_lookup(rsc->allowed_nodes,
node->details->id) == NULL)) {
pcmk__rsc_debug(rsc,
"Ignoring %s stickiness because the cluster is "
"asymmetric and %s is not explicitly allowed",
rsc->id, pe__node_name(node));
return;
}
pcmk__rsc_debug(rsc, "Resource %s has %d stickiness on %s",
rsc->id, rsc->stickiness, pe__node_name(node));
resource_location(rsc, node, rsc->stickiness, "stickiness", rsc->cluster);
}
/*!
* \internal
* \brief Apply shutdown locks for all resources as appropriate
*
* \param[in,out] scheduler Scheduler data
*/
static void
apply_shutdown_locks(pcmk_scheduler_t *scheduler)
{
if (!pcmk_is_set(scheduler->flags, pcmk_sched_shutdown_lock)) {
return;
}
for (GList *iter = scheduler->resources; iter != NULL; iter = iter->next) {
pcmk_resource_t *rsc = (pcmk_resource_t *) iter->data;
rsc->cmds->shutdown_lock(rsc);
}
}
/*!
* \internal
* \brief Calculate the number of available nodes in the cluster
*
* \param[in,out] scheduler Scheduler data
*/
static void
count_available_nodes(pcmk_scheduler_t *scheduler)
{
if (pcmk_is_set(scheduler->flags, pcmk_sched_no_compat)) {
return;
}
// @COMPAT for API backward compatibility only (cluster does not use value)
for (GList *iter = scheduler->nodes; iter != NULL; iter = iter->next) {
pcmk_node_t *node = (pcmk_node_t *) iter->data;
if ((node != NULL) && (node->weight >= 0) && node->details->online
&& (node->details->type != node_ping)) {
scheduler->max_valid_nodes++;
}
}
crm_trace("Online node count: %d", scheduler->max_valid_nodes);
}
/*
* \internal
* \brief Apply node-specific scheduling criteria
*
* After the CIB has been unpacked, process node-specific scheduling criteria
* including shutdown locks, location constraints, resource stickiness,
* migration thresholds, and exclusive resource discovery.
*/
static void
apply_node_criteria(pcmk_scheduler_t *scheduler)
{
crm_trace("Applying node-specific scheduling criteria");
apply_shutdown_locks(scheduler);
count_available_nodes(scheduler);
pcmk__apply_locations(scheduler);
g_list_foreach(scheduler->resources, apply_stickiness, NULL);
for (GList *node_iter = scheduler->nodes; node_iter != NULL;
node_iter = node_iter->next) {
for (GList *rsc_iter = scheduler->resources; rsc_iter != NULL;
rsc_iter = rsc_iter->next) {
check_failure_threshold(rsc_iter->data, node_iter->data);
apply_exclusive_discovery(rsc_iter->data, node_iter->data);
}
}
}
/*!
* \internal
* \brief Assign resources to nodes
*
* \param[in,out] scheduler Scheduler data
*/
static void
assign_resources(pcmk_scheduler_t *scheduler)
{
GList *iter = NULL;
crm_trace("Assigning resources to nodes");
if (!pcmk__str_eq(scheduler->placement_strategy, "default",
pcmk__str_casei)) {
pcmk__sort_resources(scheduler);
}
pcmk__show_node_capacities("Original", scheduler);
if (pcmk_is_set(scheduler->flags, pcmk_sched_have_remote_nodes)) {
/* Assign remote connection resources first (which will also assign any
* colocation dependencies). If the connection is migrating, always
* prefer the partial migration target.
*/
for (iter = scheduler->resources; iter != NULL; iter = iter->next) {
pcmk_resource_t *rsc = (pcmk_resource_t *) iter->data;
if (rsc->is_remote_node) {
pcmk__rsc_trace(rsc, "Assigning remote connection resource '%s'",
rsc->id);
rsc->cmds->assign(rsc, rsc->partial_migration_target, true);
}
}
}
/* now do the rest of the resources */
for (iter = scheduler->resources; iter != NULL; iter = iter->next) {
pcmk_resource_t *rsc = (pcmk_resource_t *) iter->data;
if (!rsc->is_remote_node) {
pcmk__rsc_trace(rsc, "Assigning %s resource '%s'",
rsc->xml->name, rsc->id);
rsc->cmds->assign(rsc, NULL, true);
}
}
pcmk__show_node_capacities("Remaining", scheduler);
}
/*!
* \internal
* \brief Schedule fail count clearing on online nodes if resource is orphaned
*
* \param[in,out] data Resource to check
* \param[in] user_data Ignored
*/
static void
clear_failcounts_if_orphaned(gpointer data, gpointer user_data)
{
pcmk_resource_t *rsc = data;
if (!pcmk_is_set(rsc->flags, pcmk_rsc_removed)) {
return;
}
crm_trace("Clear fail counts for orphaned resource %s", rsc->id);
/* There's no need to recurse into rsc->children because those
* should just be unassigned clone instances.
*/
for (GList *iter = rsc->cluster->nodes; iter != NULL; iter = iter->next) {
pcmk_node_t *node = (pcmk_node_t *) iter->data;
pcmk_action_t *clear_op = NULL;
if (!node->details->online) {
continue;
}
if (pe_get_failcount(node, rsc, NULL, pcmk__fc_effective, NULL) == 0) {
continue;
}
clear_op = pe__clear_failcount(rsc, node, "it is orphaned",
rsc->cluster);
/* We can't use order_action_then_stop() here because its
* pcmk__ar_guest_allowed breaks things
*/
pcmk__new_ordering(clear_op->rsc, NULL, clear_op, rsc, stop_key(rsc),
NULL, pcmk__ar_ordered, rsc->cluster);
}
}
/*!
* \internal
* \brief Schedule any resource actions needed
*
* \param[in,out] scheduler Scheduler data
*/
static void
schedule_resource_actions(pcmk_scheduler_t *scheduler)
{
// Process deferred action checks
pe__foreach_param_check(scheduler, check_params);
pe__free_param_checks(scheduler);
if (pcmk_is_set(scheduler->flags, pcmk_sched_probe_resources)) {
crm_trace("Scheduling probes");
pcmk__schedule_probes(scheduler);
}
if (pcmk_is_set(scheduler->flags, pcmk_sched_stop_removed_resources)) {
g_list_foreach(scheduler->resources, clear_failcounts_if_orphaned,
NULL);
}
crm_trace("Scheduling resource actions");
for (GList *iter = scheduler->resources; iter != NULL; iter = iter->next) {
pcmk_resource_t *rsc = (pcmk_resource_t *) iter->data;
rsc->cmds->create_actions(rsc);
}
}
/*!
* \internal
* \brief Check whether a resource or any of its descendants are managed
*
* \param[in] rsc Resource to check
*
* \return true if resource or any descendant is managed, otherwise false
*/
static bool
is_managed(const pcmk_resource_t *rsc)
{
if (pcmk_is_set(rsc->flags, pcmk_rsc_managed)) {
return true;
}
for (GList *iter = rsc->children; iter != NULL; iter = iter->next) {
if (is_managed((pcmk_resource_t *) iter->data)) {
return true;
}
}
return false;
}
/*!
* \internal
* \brief Check whether any resources in the cluster are managed
*
* \param[in] scheduler Scheduler data
*
* \return true if any resource is managed, otherwise false
*/
static bool
any_managed_resources(const pcmk_scheduler_t *scheduler)
{
for (const GList *iter = scheduler->resources;
iter != NULL; iter = iter->next) {
if (is_managed((const pcmk_resource_t *) iter->data)) {
return true;
}
}
return false;
}
/*!
* \internal
* \brief Check whether a node requires fencing
*
* \param[in] node Node to check
* \param[in] have_managed Whether any resource in cluster is managed
*
* \return true if \p node should be fenced, otherwise false
*/
static bool
needs_fencing(const pcmk_node_t *node, bool have_managed)
{
return have_managed && node->details->unclean
&& pe_can_fence(node->details->data_set, node);
}
/*!
* \internal
* \brief Check whether a node requires shutdown
*
* \param[in] node Node to check
*
* \return true if \p node should be shut down, otherwise false
*/
static bool
needs_shutdown(const pcmk_node_t *node)
{
if (pe__is_guest_or_remote_node(node)) {
/* Do not send shutdown actions for Pacemaker Remote nodes.
* @TODO We might come up with a good use for this in the future.
*/
return false;
}
return node->details->online && node->details->shutdown;
}
/*!
* \internal
* \brief Track and order non-DC fencing
*
* \param[in,out] list List of existing non-DC fencing actions
* \param[in,out] action Fencing action to prepend to \p list
* \param[in] scheduler Scheduler data
*
* \return (Possibly new) head of \p list
*/
static GList *
add_nondc_fencing(GList *list, pcmk_action_t *action,
const pcmk_scheduler_t *scheduler)
{
if (!pcmk_is_set(scheduler->flags, pcmk_sched_concurrent_fencing)
&& (list != NULL)) {
/* Concurrent fencing is disabled, so order each non-DC
* fencing in a chain. If there is any DC fencing or
* shutdown, it will be ordered after the last action in the
* chain later.
*/
order_actions((pcmk_action_t *) list->data, action, pcmk__ar_ordered);
}
return g_list_prepend(list, action);
}
/*!
* \internal
* \brief Schedule a node for fencing
*
* \param[in,out] node Node that requires fencing
*/
static pcmk_action_t *
schedule_fencing(pcmk_node_t *node)
{
pcmk_action_t *fencing = pe_fence_op(node, NULL, FALSE, "node is unclean",
FALSE, node->details->data_set);
pcmk__sched_warn("Scheduling node %s for fencing", pe__node_name(node));
pcmk__order_vs_fence(fencing, node->details->data_set);
return fencing;
}
/*!
* \internal
* \brief Create and order node fencing and shutdown actions
*
* \param[in,out] scheduler Scheduler data
*/
static void
schedule_fencing_and_shutdowns(pcmk_scheduler_t *scheduler)
{
pcmk_action_t *dc_down = NULL;
bool integrity_lost = false;
bool have_managed = any_managed_resources(scheduler);
GList *fencing_ops = NULL;
GList *shutdown_ops = NULL;
crm_trace("Scheduling fencing and shutdowns as needed");
if (!have_managed) {
crm_notice("No fencing will be done until there are resources "
"to manage");
}
// Check each node for whether it needs fencing or shutdown
for (GList *iter = scheduler->nodes; iter != NULL; iter = iter->next) {
pcmk_node_t *node = (pcmk_node_t *) iter->data;
pcmk_action_t *fencing = NULL;
/* Guest nodes are "fenced" by recovering their container resource,
* so handle them separately.
*/
if (pe__is_guest_node(node)) {
if (node->details->remote_requires_reset && have_managed
&& pe_can_fence(scheduler, node)) {
pcmk__fence_guest(node);
}
continue;
}
if (needs_fencing(node, have_managed)) {
fencing = schedule_fencing(node);
// Track DC and non-DC fence actions separately
if (node->details->is_dc) {
dc_down = fencing;
} else {
fencing_ops = add_nondc_fencing(fencing_ops, fencing,
scheduler);
}
} else if (needs_shutdown(node)) {
pcmk_action_t *down_op = pcmk__new_shutdown_action(node);
// Track DC and non-DC shutdown actions separately
if (node->details->is_dc) {
dc_down = down_op;
} else {
shutdown_ops = g_list_prepend(shutdown_ops, down_op);
}
}
if ((fencing == NULL) && node->details->unclean) {
integrity_lost = true;
pcmk__config_warn("Node %s is unclean but cannot be fenced",
pe__node_name(node));
}
}
if (integrity_lost) {
if (!pcmk_is_set(scheduler->flags, pcmk_sched_fencing_enabled)) {
pcmk__config_warn("Resource functionality and data integrity "
"cannot be guaranteed (configure, enable, "
"and test fencing to correct this)");
} else if (!pcmk_is_set(scheduler->flags, pcmk_sched_quorate)) {
crm_notice("Unclean nodes will not be fenced until quorum is "
"attained or " PCMK_OPT_NO_QUORUM_POLICY " is set to "
"ignore");
}
}
if (dc_down != NULL) {
/* Order any non-DC shutdowns before any DC shutdown, to avoid repeated
* DC elections. However, we don't want to order non-DC shutdowns before
* a DC *fencing*, because even though we don't want a node that's
* shutting down to become DC, the DC fencing could be ordered before a
* clone stop that's also ordered before the shutdowns, thus leading to
* a graph loop.
*/
if (pcmk__str_eq(dc_down->task, PCMK_ACTION_DO_SHUTDOWN,
pcmk__str_none)) {
pcmk__order_after_each(dc_down, shutdown_ops);
}
// Order any non-DC fencing before any DC fencing or shutdown
if (pcmk_is_set(scheduler->flags, pcmk_sched_concurrent_fencing)) {
/* With concurrent fencing, order each non-DC fencing action
* separately before any DC fencing or shutdown.
*/
pcmk__order_after_each(dc_down, fencing_ops);
} else if (fencing_ops != NULL) {
/* Without concurrent fencing, the non-DC fencing actions are
* already ordered relative to each other, so we just need to order
* the DC fencing after the last action in the chain (which is the
* first item in the list).
*/
order_actions((pcmk_action_t *) fencing_ops->data, dc_down,
pcmk__ar_ordered);
}
}
g_list_free(fencing_ops);
g_list_free(shutdown_ops);
}
static void
log_resource_details(pcmk_scheduler_t *scheduler)
{
pcmk__output_t *out = scheduler->priv;
GList *all = NULL;
/* Due to the `crm_mon --node=` feature, out->message() for all the
* resource-related messages expects a list of nodes that we are allowed to
* output information for. Here, we create a wildcard to match all nodes.
*/
all = g_list_prepend(all, (gpointer) "*");
for (GList *item = scheduler->resources; item != NULL; item = item->next) {
pcmk_resource_t *rsc = (pcmk_resource_t *) item->data;
// Log all resources except inactive orphans
if (!pcmk_is_set(rsc->flags, pcmk_rsc_removed)
|| (rsc->role != pcmk_role_stopped)) {
out->message(out, crm_map_element_name(rsc->xml), 0, rsc, all, all);
}
}
g_list_free(all);
}
static void
log_all_actions(pcmk_scheduler_t *scheduler)
{
/* This only ever outputs to the log, so ignore whatever output object was
* previously set and just log instead.
*/
pcmk__output_t *prev_out = scheduler->priv;
pcmk__output_t *out = NULL;
if (pcmk__log_output_new(&out) != pcmk_rc_ok) {
return;
}
pe__register_messages(out);
pcmk__register_lib_messages(out);
pcmk__output_set_log_level(out, LOG_NOTICE);
scheduler->priv = out;
out->begin_list(out, NULL, NULL, "Actions");
pcmk__output_actions(scheduler);
out->end_list(out);
out->finish(out, CRM_EX_OK, true, NULL);
pcmk__output_free(out);
scheduler->priv = prev_out;
}
/*!
* \internal
* \brief Log all required but unrunnable actions at trace level
*
* \param[in] scheduler Scheduler data
*/
static void
log_unrunnable_actions(const pcmk_scheduler_t *scheduler)
{
const uint64_t flags = pcmk_action_optional
|pcmk_action_runnable
|pcmk_action_pseudo;
crm_trace("Required but unrunnable actions:");
for (const GList *iter = scheduler->actions;
iter != NULL; iter = iter->next) {
const pcmk_action_t *action = (const pcmk_action_t *) iter->data;
if (!pcmk_any_flags_set(action->flags, flags)) {
pcmk__log_action("\t", action, true);
}
}
}
/*!
* \internal
* \brief Unpack the CIB for scheduling
*
* \param[in,out] cib CIB XML to unpack (may be NULL if already unpacked)
* \param[in] flags Scheduler flags to set in addition to defaults
* \param[in,out] scheduler Scheduler data
*/
static void
unpack_cib(xmlNode *cib, unsigned long long flags, pcmk_scheduler_t *scheduler)
{
const char* localhost_save = NULL;
if (pcmk_is_set(scheduler->flags, pcmk_sched_have_status)) {
crm_trace("Reusing previously calculated cluster status");
- pe__set_working_set_flags(scheduler, flags);
+ pcmk__set_scheduler_flags(scheduler, flags);
return;
}
if (scheduler->localhost) {
localhost_save = scheduler->localhost;
}
CRM_ASSERT(cib != NULL);
crm_trace("Calculating cluster status");
/* This will zero the entire struct without freeing anything first, so
* callers should never call pcmk__schedule_actions() with a populated data
* set unless pcmk_sched_have_status is set (i.e. cluster_status() was
* previously called, whether directly or via pcmk__schedule_actions()).
*/
set_working_set_defaults(scheduler);
if (localhost_save) {
scheduler->localhost = localhost_save;
}
- pe__set_working_set_flags(scheduler, flags);
+ pcmk__set_scheduler_flags(scheduler, flags);
scheduler->input = cib;
cluster_status(scheduler); // Sets pcmk_sched_have_status
}
/*!
* \internal
* \brief Run the scheduler for a given CIB
*
* \param[in,out] cib CIB XML to use as scheduler input
* \param[in] flags Scheduler flags to set in addition to defaults
* \param[in,out] scheduler Scheduler data
*/
void
pcmk__schedule_actions(xmlNode *cib, unsigned long long flags,
pcmk_scheduler_t *scheduler)
{
unpack_cib(cib, flags, scheduler);
pcmk__set_assignment_methods(scheduler);
pcmk__apply_node_health(scheduler);
pcmk__unpack_constraints(scheduler);
if (pcmk_is_set(scheduler->flags, pcmk_sched_validate_only)) {
return;
}
if (!pcmk_is_set(scheduler->flags, pcmk_sched_location_only)
&& pcmk__is_daemon) {
log_resource_details(scheduler);
}
apply_node_criteria(scheduler);
if (pcmk_is_set(scheduler->flags, pcmk_sched_location_only)) {
return;
}
pcmk__create_internal_constraints(scheduler);
pcmk__handle_rsc_config_changes(scheduler);
assign_resources(scheduler);
schedule_resource_actions(scheduler);
/* Remote ordering constraints need to happen prior to calculating fencing
* because it is one more place we can mark nodes as needing fencing.
*/
pcmk__order_remote_connection_actions(scheduler);
schedule_fencing_and_shutdowns(scheduler);
pcmk__apply_orderings(scheduler);
log_all_actions(scheduler);
pcmk__create_graph(scheduler);
if (get_crm_log_level() == LOG_TRACE) {
log_unrunnable_actions(scheduler);
}
}
diff --git a/lib/pacemaker/pcmk_simulate.c b/lib/pacemaker/pcmk_simulate.c
index 1d3623ccf6..999f5eb167 100644
--- a/lib/pacemaker/pcmk_simulate.c
+++ b/lib/pacemaker/pcmk_simulate.c
@@ -1,1006 +1,1006 @@
/*
* Copyright 2021-2024 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 <crm/cib/internal.h>
#include <crm/common/output.h>
#include <crm/common/results.h>
#include <crm/common/scheduler.h>
#include <pacemaker-internal.h>
#include <pacemaker.h>
#include <stdint.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <unistd.h>
#include "libpacemaker_private.h"
static pcmk__output_t *out = NULL;
static cib_t *fake_cib = NULL;
static GList *fake_resource_list = NULL;
static const GList *fake_op_fail_list = NULL;
static void set_effective_date(pcmk_scheduler_t *scheduler, bool print_original,
const char *use_date);
/*!
* \internal
* \brief Create an action name for use in a dot graph
*
* \param[in] action Action to create name for
* \param[in] verbose If true, add action ID to name
*
* \return Newly allocated string with action name
* \note It is the caller's responsibility to free the result.
*/
static char *
create_action_name(const pcmk_action_t *action, bool verbose)
{
char *action_name = NULL;
const char *prefix = "";
const char *action_host = NULL;
const char *clone_name = NULL;
const char *task = action->task;
if (action->node != NULL) {
action_host = action->node->details->uname;
} else if (!pcmk_is_set(action->flags, pcmk_action_pseudo)) {
action_host = "<none>";
}
if (pcmk__str_eq(action->task, PCMK_ACTION_CANCEL, pcmk__str_none)) {
prefix = "Cancel ";
task = action->cancel_task;
}
if (action->rsc != NULL) {
clone_name = action->rsc->clone_name;
}
if (clone_name != NULL) {
char *key = NULL;
guint interval_ms = 0;
if (pcmk__guint_from_hash(action->meta,
XML_LRM_ATTR_INTERVAL_MS, 0,
&interval_ms) != pcmk_rc_ok) {
interval_ms = 0;
}
if (pcmk__strcase_any_of(action->task, PCMK_ACTION_NOTIFY,
PCMK_ACTION_NOTIFIED, NULL)) {
const char *n_type = g_hash_table_lookup(action->meta,
"notify_key_type");
const char *n_task = g_hash_table_lookup(action->meta,
"notify_key_operation");
CRM_ASSERT(n_type != NULL);
CRM_ASSERT(n_task != NULL);
key = pcmk__notify_key(clone_name, n_type, n_task);
} else {
key = pcmk__op_key(clone_name, task, interval_ms);
}
if (action_host != NULL) {
action_name = crm_strdup_printf("%s%s %s",
prefix, key, action_host);
} else {
action_name = crm_strdup_printf("%s%s", prefix, key);
}
free(key);
} else if (pcmk__str_eq(action->task, PCMK_ACTION_STONITH,
pcmk__str_none)) {
const char *op = g_hash_table_lookup(action->meta, "stonith_action");
action_name = crm_strdup_printf("%s%s '%s' %s",
prefix, action->task, op, action_host);
} else if (action->rsc && action_host) {
action_name = crm_strdup_printf("%s%s %s",
prefix, action->uuid, action_host);
} else if (action_host) {
action_name = crm_strdup_printf("%s%s %s",
prefix, action->task, action_host);
} else {
action_name = crm_strdup_printf("%s", action->uuid);
}
if (verbose) {
char *with_id = crm_strdup_printf("%s (%d)", action_name, action->id);
free(action_name);
action_name = with_id;
}
return action_name;
}
/*!
* \internal
* \brief Display the status of a cluster
*
* \param[in,out] scheduler Scheduler data
* \param[in] show_opts How to modify display (as pcmk_show_opt_e flags)
* \param[in] section_opts Sections to display (as pcmk_section_e flags)
* \param[in] title What to use as list title
* \param[in] print_spacer Whether to display a spacer first
*/
static void
print_cluster_status(pcmk_scheduler_t *scheduler, uint32_t show_opts,
uint32_t section_opts, const char *title,
bool print_spacer)
{
pcmk__output_t *out = scheduler->priv;
GList *all = NULL;
crm_exit_t stonith_rc = 0;
enum pcmk_pacemakerd_state state = pcmk_pacemakerd_state_invalid;
section_opts |= pcmk_section_nodes | pcmk_section_resources;
show_opts |= pcmk_show_inactive_rscs | pcmk_show_failed_detail;
all = g_list_prepend(all, (gpointer) "*");
PCMK__OUTPUT_SPACER_IF(out, print_spacer);
out->begin_list(out, NULL, NULL, "%s", title);
out->message(out, "cluster-status",
scheduler, state, stonith_rc, NULL,
false, section_opts, show_opts, NULL, all, all);
out->end_list(out);
g_list_free(all);
}
/*!
* \internal
* \brief Display a summary of all actions scheduled in a transition
*
* \param[in,out] scheduler Scheduler data (fully scheduled)
* \param[in] print_spacer Whether to display a spacer first
*/
static void
print_transition_summary(pcmk_scheduler_t *scheduler, bool print_spacer)
{
pcmk__output_t *out = scheduler->priv;
PCMK__OUTPUT_SPACER_IF(out, print_spacer);
out->begin_list(out, NULL, NULL, "Transition Summary");
pcmk__output_actions(scheduler);
out->end_list(out);
}
/*!
* \internal
* \brief Reset scheduler input, output, date, and flags
*
* \param[in,out] scheduler Scheduler data
* \param[in] input What to set as cluster input
* \param[in] out What to set as cluster output object
* \param[in] use_date What to set as cluster's current timestamp
* \param[in] flags Group of enum pcmk_scheduler_flags to set
*/
static void
reset(pcmk_scheduler_t *scheduler, xmlNodePtr input, pcmk__output_t *out,
const char *use_date, unsigned int flags)
{
scheduler->input = input;
scheduler->priv = out;
set_effective_date(scheduler, true, use_date);
if (pcmk_is_set(flags, pcmk_sim_sanitized)) {
- pe__set_working_set_flags(scheduler, pcmk_sched_sanitized);
+ pcmk__set_scheduler_flags(scheduler, pcmk_sched_sanitized);
}
if (pcmk_is_set(flags, pcmk_sim_show_scores)) {
- pe__set_working_set_flags(scheduler, pcmk_sched_output_scores);
+ pcmk__set_scheduler_flags(scheduler, pcmk_sched_output_scores);
}
if (pcmk_is_set(flags, pcmk_sim_show_utilization)) {
- pe__set_working_set_flags(scheduler, pcmk_sched_show_utilization);
+ pcmk__set_scheduler_flags(scheduler, pcmk_sched_show_utilization);
}
}
/*!
* \brief Write out a file in dot(1) format describing the actions that will
* be taken by the scheduler in response to an input CIB file.
*
* \param[in,out] scheduler Scheduler data
* \param[in] dot_file The filename to write
* \param[in] all_actions Write all actions, even those that are optional
* or are on unmanaged resources
* \param[in] verbose Add extra information, such as action IDs, to the
* output
*
* \return Standard Pacemaker return code
*/
static int
write_sim_dotfile(pcmk_scheduler_t *scheduler, const char *dot_file,
bool all_actions, bool verbose)
{
GList *iter = NULL;
FILE *dot_strm = fopen(dot_file, "w");
if (dot_strm == NULL) {
return errno;
}
fprintf(dot_strm, " digraph \"g\" {\n");
for (iter = scheduler->actions; iter != NULL; iter = iter->next) {
pcmk_action_t *action = (pcmk_action_t *) iter->data;
const char *style = "dashed";
const char *font = "black";
const char *color = "black";
char *action_name = create_action_name(action, verbose);
if (pcmk_is_set(action->flags, pcmk_action_pseudo)) {
font = "orange";
}
if (pcmk_is_set(action->flags, pcmk_action_added_to_graph)) {
style = "bold";
color = "green";
} else if ((action->rsc != NULL)
&& !pcmk_is_set(action->rsc->flags, pcmk_rsc_managed)) {
color = "red";
font = "purple";
if (!all_actions) {
goto do_not_write;
}
} else if (pcmk_is_set(action->flags, pcmk_action_optional)) {
color = "blue";
if (!all_actions) {
goto do_not_write;
}
} else {
color = "red";
CRM_LOG_ASSERT(!pcmk_is_set(action->flags, pcmk_action_runnable));
}
pe__set_action_flags(action, pcmk_action_added_to_graph);
fprintf(dot_strm, "\"%s\" [ style=%s color=\"%s\" fontcolor=\"%s\"]\n",
action_name, style, color, font);
do_not_write:
free(action_name);
}
for (iter = scheduler->actions; iter != NULL; iter = iter->next) {
pcmk_action_t *action = (pcmk_action_t *) iter->data;
for (GList *before_iter = action->actions_before;
before_iter != NULL; before_iter = before_iter->next) {
pcmk__related_action_t *before = before_iter->data;
char *before_name = NULL;
char *after_name = NULL;
const char *style = "dashed";
bool optional = true;
if (before->state == pe_link_dumped) {
optional = false;
style = "bold";
} else if ((uint32_t) before->type == pcmk__ar_none) {
continue;
} else if (pcmk_is_set(before->action->flags,
pcmk_action_added_to_graph)
&& pcmk_is_set(action->flags, pcmk_action_added_to_graph)
&& (uint32_t) before->type != pcmk__ar_if_on_same_node_or_target) {
optional = false;
}
if (all_actions || !optional) {
before_name = create_action_name(before->action, verbose);
after_name = create_action_name(action, verbose);
fprintf(dot_strm, "\"%s\" -> \"%s\" [ style = %s]\n",
before_name, after_name, style);
free(before_name);
free(after_name);
}
}
}
fprintf(dot_strm, "}\n");
fflush(dot_strm);
fclose(dot_strm);
return pcmk_rc_ok;
}
/*!
* \brief Profile the configuration updates and scheduler actions in a single
* CIB file, printing the profiling timings.
*
* \note \p scheduler->priv must have been set to a valid \p pcmk__output_t
* object before this function is called.
*
* \param[in] xml_file The CIB file to profile
* \param[in] repeat Number of times to run
* \param[in,out] scheduler Scheduler data
* \param[in] use_date The date to set the cluster's time to (may be NULL)
*/
static void
profile_file(const char *xml_file, long long repeat,
pcmk_scheduler_t *scheduler, const char *use_date)
{
pcmk__output_t *out = scheduler->priv;
xmlNode *cib_object = NULL;
clock_t start = 0;
clock_t end;
unsigned long long scheduler_flags = pcmk_sched_no_compat;
CRM_ASSERT(out != NULL);
cib_object = filename2xml(xml_file);
start = clock();
if (pcmk_find_cib_element(cib_object, XML_CIB_TAG_STATUS) == NULL) {
create_xml_node(cib_object, XML_CIB_TAG_STATUS);
}
if (cli_config_update(&cib_object, NULL, FALSE) == FALSE) {
free_xml(cib_object);
return;
}
if (validate_xml(cib_object, NULL, FALSE) != TRUE) {
free_xml(cib_object);
return;
}
if (pcmk_is_set(scheduler->flags, pcmk_sched_output_scores)) {
scheduler_flags |= pcmk_sched_output_scores;
}
if (pcmk_is_set(scheduler->flags, pcmk_sched_show_utilization)) {
scheduler_flags |= pcmk_sched_show_utilization;
}
for (int i = 0; i < repeat; ++i) {
xmlNode *input = (repeat == 1)? cib_object : copy_xml(cib_object);
scheduler->input = input;
set_effective_date(scheduler, false, use_date);
pcmk__schedule_actions(input, scheduler_flags, scheduler);
pe_reset_working_set(scheduler);
}
end = clock();
out->message(out, "profile", xml_file, start, end);
}
void
pcmk__profile_dir(const char *dir, long long repeat,
pcmk_scheduler_t *scheduler, const char *use_date)
{
pcmk__output_t *out = scheduler->priv;
struct dirent **namelist;
int file_num = scandir(dir, &namelist, 0, alphasort);
CRM_ASSERT(out != NULL);
if (file_num > 0) {
struct stat prop;
char buffer[FILENAME_MAX];
out->begin_list(out, NULL, NULL, "Timings");
while (file_num--) {
if ('.' == namelist[file_num]->d_name[0]) {
free(namelist[file_num]);
continue;
} else if (!pcmk__ends_with_ext(namelist[file_num]->d_name,
".xml")) {
free(namelist[file_num]);
continue;
}
snprintf(buffer, sizeof(buffer), "%s/%s",
dir, namelist[file_num]->d_name);
if (stat(buffer, &prop) == 0 && S_ISREG(prop.st_mode)) {
profile_file(buffer, repeat, scheduler, use_date);
}
free(namelist[file_num]);
}
free(namelist);
out->end_list(out);
}
}
/*!
* \brief Set the date of the cluster, either to the value given by
* \p use_date, or to the "execution-date" value in the CIB.
*
* \note \p scheduler->priv must have been set to a valid \p pcmk__output_t
* object before this function is called.
*
* \param[in,out] scheduler Scheduler data
* \param[in] print_original If \p true, the "execution-date" should
* also be printed
* \param[in] use_date The date to set the cluster's time to
* (may be NULL)
*/
static void
set_effective_date(pcmk_scheduler_t *scheduler, bool print_original,
const char *use_date)
{
pcmk__output_t *out = scheduler->priv;
time_t original_date = 0;
CRM_ASSERT(out != NULL);
crm_element_value_epoch(scheduler->input, "execution-date", &original_date);
if (use_date) {
scheduler->now = crm_time_new(use_date);
out->info(out, "Setting effective cluster time: %s", use_date);
crm_time_log(LOG_NOTICE, "Pretending 'now' is", scheduler->now,
crm_time_log_date | crm_time_log_timeofday);
} else if (original_date != 0) {
scheduler->now = pcmk__copy_timet(original_date);
if (print_original) {
char *when = crm_time_as_string(scheduler->now,
crm_time_log_date|crm_time_log_timeofday);
out->info(out, "Using the original execution date of: %s", when);
free(when);
}
}
}
/*!
* \internal
* \brief Simulate successfully executing a pseudo-action in a graph
*
* \param[in,out] graph Graph to update with pseudo-action result
* \param[in,out] action Pseudo-action to simulate executing
*
* \return Standard Pacemaker return code
*/
static int
simulate_pseudo_action(pcmk__graph_t *graph, pcmk__graph_action_t *action)
{
const char *node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET);
const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY);
pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed);
out->message(out, "inject-pseudo-action", node, task);
pcmk__update_graph(graph, action);
return pcmk_rc_ok;
}
/*!
* \internal
* \brief Simulate executing a resource action in a graph
*
* \param[in,out] graph Graph to update with resource action result
* \param[in,out] action Resource action to simulate executing
*
* \return Standard Pacemaker return code
*/
static int
simulate_resource_action(pcmk__graph_t *graph, pcmk__graph_action_t *action)
{
int rc;
lrmd_event_data_t *op = NULL;
int target_outcome = PCMK_OCF_OK;
const char *rtype = NULL;
const char *rclass = NULL;
const char *resource = NULL;
const char *rprovider = NULL;
const char *resource_config_name = NULL;
const char *operation = crm_element_value(action->xml, "operation");
const char *target_rc_s = crm_meta_value(action->params,
XML_ATTR_TE_TARGET_RC);
xmlNode *cib_node = NULL;
xmlNode *cib_resource = NULL;
xmlNode *action_rsc = first_named_child(action->xml, XML_CIB_TAG_RESOURCE);
char *node = crm_element_value_copy(action->xml, XML_LRM_ATTR_TARGET);
char *uuid = NULL;
const char *router_node = crm_element_value(action->xml,
XML_LRM_ATTR_ROUTER_NODE);
// Certain actions don't need to be displayed or history entries
if (pcmk__str_eq(operation, CRM_OP_REPROBE, pcmk__str_none)) {
crm_debug("No history injection for %s op on %s", operation, node);
goto done; // Confirm action and update graph
}
if (action_rsc == NULL) { // Shouldn't be possible
crm_log_xml_err(action->xml, "Bad");
free(node);
return EPROTO;
}
/* A resource might be known by different names in the configuration and in
* the action (for example, a clone instance). Grab the configuration name
* (which is preferred when writing history), and if necessary, the instance
* name.
*/
resource_config_name = crm_element_value(action_rsc, PCMK_XA_ID);
if (resource_config_name == NULL) { // Shouldn't be possible
crm_log_xml_err(action->xml, "No ID");
free(node);
return EPROTO;
}
resource = resource_config_name;
if (pe_find_resource(fake_resource_list, resource) == NULL) {
const char *longname = crm_element_value(action_rsc, XML_ATTR_ID_LONG);
if ((longname != NULL)
&& (pe_find_resource(fake_resource_list, longname) != NULL)) {
resource = longname;
}
}
// Certain actions need to be displayed but don't need history entries
if (pcmk__strcase_any_of(operation, PCMK_ACTION_DELETE,
PCMK_ACTION_META_DATA, NULL)) {
out->message(out, "inject-rsc-action", resource, operation, node,
(guint) 0);
goto done; // Confirm action and update graph
}
rclass = crm_element_value(action_rsc, XML_AGENT_ATTR_CLASS);
rtype = crm_element_value(action_rsc, XML_ATTR_TYPE);
rprovider = crm_element_value(action_rsc, XML_AGENT_ATTR_PROVIDER);
pcmk__scan_min_int(target_rc_s, &target_outcome, 0);
CRM_ASSERT(fake_cib->cmds->query(fake_cib, NULL, NULL,
cib_sync_call|cib_scope_local) == pcmk_ok);
// Ensure the action node is in the CIB
uuid = crm_element_value_copy(action->xml, XML_LRM_ATTR_TARGET_UUID);
cib_node = pcmk__inject_node(fake_cib, node,
((router_node == NULL)? uuid: node));
free(uuid);
CRM_ASSERT(cib_node != NULL);
// Add a history entry for the action
cib_resource = pcmk__inject_resource_history(out, cib_node, resource,
resource_config_name,
rclass, rtype, rprovider);
if (cib_resource == NULL) {
crm_err("Could not simulate action %d history for resource %s",
action->id, resource);
free(node);
free_xml(cib_node);
return EINVAL;
}
// Simulate and display an executor event for the action result
op = pcmk__event_from_graph_action(cib_resource, action, PCMK_EXEC_DONE,
target_outcome, "User-injected result");
out->message(out, "inject-rsc-action", resource, op->op_type, node,
op->interval_ms);
// Check whether action is in a list of desired simulated failures
for (const GList *iter = fake_op_fail_list;
iter != NULL; iter = iter->next) {
const char *spec = (const char *) iter->data;
char *key = NULL;
const char *match_name = NULL;
// Allow user to specify anonymous clone with or without instance number
key = crm_strdup_printf(PCMK__OP_FMT "@%s=", resource, op->op_type,
op->interval_ms, node);
if (strncasecmp(key, spec, strlen(key)) == 0) {
match_name = resource;
}
free(key);
// If not found, try the resource's name in the configuration
if ((match_name == NULL)
&& (strcmp(resource, resource_config_name) != 0)) {
key = crm_strdup_printf(PCMK__OP_FMT "@%s=", resource_config_name,
op->op_type, op->interval_ms, node);
if (strncasecmp(key, spec, strlen(key)) == 0) {
match_name = resource_config_name;
}
free(key);
}
if (match_name == NULL) {
continue; // This failed action entry doesn't match
}
// ${match_name}_${task}_${interval_in_ms}@${node}=${rc}
rc = sscanf(spec, "%*[^=]=%d", (int *) &op->rc);
if (rc != 1) {
out->err(out, "Invalid failed operation '%s' "
"(result code must be integer)", spec);
continue; // Keep checking other list entries
}
out->info(out, "Pretending action %d failed with rc=%d",
action->id, op->rc);
pcmk__set_graph_action_flags(action, pcmk__graph_action_failed);
graph->abort_priority = INFINITY;
pcmk__inject_failcount(out, cib_node, match_name, op->op_type,
op->interval_ms, op->rc);
break;
}
pcmk__inject_action_result(cib_resource, op, target_outcome);
lrmd_free_event(op);
rc = fake_cib->cmds->modify(fake_cib, XML_CIB_TAG_STATUS, cib_node,
cib_sync_call|cib_scope_local);
CRM_ASSERT(rc == pcmk_ok);
done:
free(node);
free_xml(cib_node);
pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed);
pcmk__update_graph(graph, action);
return pcmk_rc_ok;
}
/*!
* \internal
* \brief Simulate successfully executing a cluster action
*
* \param[in,out] graph Graph to update with action result
* \param[in,out] action Cluster action to simulate
*
* \return Standard Pacemaker return code
*/
static int
simulate_cluster_action(pcmk__graph_t *graph, pcmk__graph_action_t *action)
{
const char *node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET);
const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK);
xmlNode *rsc = first_named_child(action->xml, XML_CIB_TAG_RESOURCE);
pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed);
out->message(out, "inject-cluster-action", node, task, rsc);
pcmk__update_graph(graph, action);
return pcmk_rc_ok;
}
/*!
* \internal
* \brief Simulate successfully executing a fencing action
*
* \param[in,out] graph Graph to update with action result
* \param[in,out] action Fencing action to simulate
*
* \return Standard Pacemaker return code
*/
static int
simulate_fencing_action(pcmk__graph_t *graph, pcmk__graph_action_t *action)
{
const char *op = crm_meta_value(action->params, "stonith_action");
char *target = crm_element_value_copy(action->xml, XML_LRM_ATTR_TARGET);
out->message(out, "inject-fencing-action", target, op);
if (!pcmk__str_eq(op, PCMK_ACTION_ON, pcmk__str_casei)) {
int rc = pcmk_ok;
GString *xpath = g_string_sized_new(512);
// Set node state to offline
xmlNode *cib_node = pcmk__inject_node_state_change(fake_cib, target,
false);
CRM_ASSERT(cib_node != NULL);
crm_xml_add(cib_node, PCMK_XA_CRM_DEBUG_ORIGIN, __func__);
rc = fake_cib->cmds->replace(fake_cib, XML_CIB_TAG_STATUS, cib_node,
cib_sync_call|cib_scope_local);
CRM_ASSERT(rc == pcmk_ok);
// Simulate controller clearing node's resource history and attributes
pcmk__g_strcat(xpath,
"//" XML_CIB_TAG_STATE
"[@" XML_ATTR_UNAME "='", target, "']/" XML_CIB_TAG_LRM,
NULL);
fake_cib->cmds->remove(fake_cib, (const char *) xpath->str, NULL,
cib_xpath|cib_sync_call|cib_scope_local);
g_string_truncate(xpath, 0);
pcmk__g_strcat(xpath,
"//" XML_CIB_TAG_STATE
"[@" XML_ATTR_UNAME "='", target, "']"
"/" XML_TAG_TRANSIENT_NODEATTRS, NULL);
fake_cib->cmds->remove(fake_cib, (const char *) xpath->str, NULL,
cib_xpath|cib_sync_call|cib_scope_local);
free_xml(cib_node);
g_string_free(xpath, TRUE);
}
pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed);
pcmk__update_graph(graph, action);
free(target);
return pcmk_rc_ok;
}
enum pcmk__graph_status
pcmk__simulate_transition(pcmk_scheduler_t *scheduler, cib_t *cib,
const GList *op_fail_list)
{
pcmk__graph_t *transition = NULL;
enum pcmk__graph_status graph_rc;
pcmk__graph_functions_t simulation_fns = {
simulate_pseudo_action,
simulate_resource_action,
simulate_cluster_action,
simulate_fencing_action,
};
out = scheduler->priv;
fake_cib = cib;
fake_op_fail_list = op_fail_list;
if (!out->is_quiet(out)) {
out->begin_list(out, NULL, NULL, "Executing Cluster Transition");
}
pcmk__set_graph_functions(&simulation_fns);
transition = pcmk__unpack_graph(scheduler->graph, crm_system_name);
pcmk__log_graph(LOG_DEBUG, transition);
fake_resource_list = scheduler->resources;
do {
graph_rc = pcmk__execute_graph(transition);
} while (graph_rc == pcmk__graph_active);
fake_resource_list = NULL;
if (graph_rc != pcmk__graph_complete) {
out->err(out, "Transition failed: %s",
pcmk__graph_status2text(graph_rc));
pcmk__log_graph(LOG_ERR, transition);
out->err(out, "An invalid transition was produced");
}
pcmk__free_graph(transition);
if (!out->is_quiet(out)) {
// If not quiet, we'll need the resulting CIB for later display
xmlNode *cib_object = NULL;
int rc = fake_cib->cmds->query(fake_cib, NULL, &cib_object,
cib_sync_call|cib_scope_local);
CRM_ASSERT(rc == pcmk_ok);
pe_reset_working_set(scheduler);
scheduler->input = cib_object;
out->end_list(out);
}
return graph_rc;
}
int
pcmk__simulate(pcmk_scheduler_t *scheduler, pcmk__output_t *out,
const pcmk_injections_t *injections, unsigned int flags,
uint32_t section_opts, const char *use_date,
const char *input_file, const char *graph_file,
const char *dot_file)
{
int printed = pcmk_rc_no_output;
int rc = pcmk_rc_ok;
xmlNodePtr input = NULL;
cib_t *cib = NULL;
rc = cib__signon_query(out, &cib, &input);
if (rc != pcmk_rc_ok) {
goto simulate_done;
}
reset(scheduler, input, out, use_date, flags);
cluster_status(scheduler);
if ((cib->variant == cib_native)
&& pcmk_is_set(section_opts, pcmk_section_times)) {
if (pcmk__our_nodename == NULL) {
// Currently used only in the times section
pcmk__query_node_name(out, 0, &pcmk__our_nodename, 0);
}
scheduler->localhost = pcmk__our_nodename;
}
if (!out->is_quiet(out)) {
const bool show_pending = pcmk_is_set(flags, pcmk_sim_show_pending);
if (pcmk_is_set(scheduler->flags, pcmk_sched_in_maintenance)) {
printed = out->message(out, "maint-mode", scheduler->flags);
}
if (scheduler->disabled_resources || scheduler->blocked_resources) {
PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok);
printed = out->info(out,
"%d of %d resource instances DISABLED and "
"%d BLOCKED from further action due to failure",
scheduler->disabled_resources,
scheduler->ninstances,
scheduler->blocked_resources);
}
/* Most formatted output headers use caps for each word, but this one
* only has the first word capitalized for compatibility with pcs.
*/
print_cluster_status(scheduler, (show_pending? pcmk_show_pending : 0),
section_opts, "Current cluster status",
(printed == pcmk_rc_ok));
printed = pcmk_rc_ok;
}
// If the user requested any injections, handle them
if ((injections->node_down != NULL)
|| (injections->node_fail != NULL)
|| (injections->node_up != NULL)
|| (injections->op_inject != NULL)
|| (injections->ticket_activate != NULL)
|| (injections->ticket_grant != NULL)
|| (injections->ticket_revoke != NULL)
|| (injections->ticket_standby != NULL)
|| (injections->watchdog != NULL)) {
PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok);
pcmk__inject_scheduler_input(scheduler, cib, injections);
printed = pcmk_rc_ok;
rc = cib->cmds->query(cib, NULL, &input, cib_sync_call);
if (rc != pcmk_rc_ok) {
rc = pcmk_legacy2rc(rc);
goto simulate_done;
}
cleanup_calculations(scheduler);
reset(scheduler, input, out, use_date, flags);
cluster_status(scheduler);
}
if (input_file != NULL) {
rc = write_xml_file(input, input_file, FALSE);
if (rc < 0) {
rc = pcmk_legacy2rc(rc);
goto simulate_done;
}
}
if (pcmk_any_flags_set(flags, pcmk_sim_process | pcmk_sim_simulate)) {
pcmk__output_t *logger_out = NULL;
unsigned long long scheduler_flags = pcmk_sched_no_compat;
if (pcmk_is_set(scheduler->flags, pcmk_sched_output_scores)) {
scheduler_flags |= pcmk_sched_output_scores;
}
if (pcmk_is_set(scheduler->flags, pcmk_sched_show_utilization)) {
scheduler_flags |= pcmk_sched_show_utilization;
}
if (pcmk_all_flags_set(scheduler->flags,
pcmk_sched_output_scores
|pcmk_sched_show_utilization)) {
PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok);
out->begin_list(out, NULL, NULL,
"Assignment Scores and Utilization Information");
printed = pcmk_rc_ok;
} else if (pcmk_is_set(scheduler->flags, pcmk_sched_output_scores)) {
PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok);
out->begin_list(out, NULL, NULL, "Assignment Scores");
printed = pcmk_rc_ok;
} else if (pcmk_is_set(scheduler->flags, pcmk_sched_show_utilization)) {
PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok);
out->begin_list(out, NULL, NULL, "Utilization Information");
printed = pcmk_rc_ok;
} else {
rc = pcmk__log_output_new(&logger_out);
if (rc != pcmk_rc_ok) {
goto simulate_done;
}
pe__register_messages(logger_out);
pcmk__register_lib_messages(logger_out);
scheduler->priv = logger_out;
}
pcmk__schedule_actions(input, scheduler_flags, scheduler);
if (logger_out == NULL) {
out->end_list(out);
} else {
logger_out->finish(logger_out, CRM_EX_OK, true, NULL);
pcmk__output_free(logger_out);
scheduler->priv = out;
}
input = NULL; /* Don't try and free it twice */
if (graph_file != NULL) {
rc = write_xml_file(scheduler->graph, graph_file, FALSE);
if (rc < 0) {
rc = pcmk_rc_graph_error;
goto simulate_done;
}
}
if (dot_file != NULL) {
rc = write_sim_dotfile(scheduler, dot_file,
pcmk_is_set(flags, pcmk_sim_all_actions),
pcmk_is_set(flags, pcmk_sim_verbose));
if (rc != pcmk_rc_ok) {
rc = pcmk_rc_dot_error;
goto simulate_done;
}
}
if (!out->is_quiet(out)) {
print_transition_summary(scheduler, printed == pcmk_rc_ok);
}
}
rc = pcmk_rc_ok;
if (!pcmk_is_set(flags, pcmk_sim_simulate)) {
goto simulate_done;
}
PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok);
if (pcmk__simulate_transition(scheduler, cib, injections->op_fail)
!= pcmk__graph_complete) {
rc = pcmk_rc_invalid_transition;
}
if (out->is_quiet(out)) {
goto simulate_done;
}
set_effective_date(scheduler, true, use_date);
if (pcmk_is_set(flags, pcmk_sim_show_scores)) {
- pe__set_working_set_flags(scheduler, pcmk_sched_output_scores);
+ pcmk__set_scheduler_flags(scheduler, pcmk_sched_output_scores);
}
if (pcmk_is_set(flags, pcmk_sim_show_utilization)) {
- pe__set_working_set_flags(scheduler, pcmk_sched_show_utilization);
+ pcmk__set_scheduler_flags(scheduler, pcmk_sched_show_utilization);
}
cluster_status(scheduler);
print_cluster_status(scheduler, 0, section_opts, "Revised Cluster Status",
true);
simulate_done:
cib__clean_up_connection(&cib);
return rc;
}
int
pcmk_simulate(xmlNodePtr *xml, pcmk_scheduler_t *scheduler,
const pcmk_injections_t *injections, unsigned int flags,
unsigned int section_opts, const char *use_date,
const char *input_file, const char *graph_file,
const char *dot_file)
{
pcmk__output_t *out = NULL;
int rc = pcmk_rc_ok;
rc = pcmk__xml_output_new(&out, xml);
if (rc != pcmk_rc_ok) {
return rc;
}
pe__register_messages(out);
pcmk__register_lib_messages(out);
rc = pcmk__simulate(scheduler, out, injections, flags, section_opts,
use_date, input_file, graph_file, dot_file);
pcmk__xml_output_finish(out, pcmk_rc2exitc(rc), xml);
return rc;
}
diff --git a/lib/pacemaker/pcmk_status.c b/lib/pacemaker/pcmk_status.c
index 462d9c6a0e..7e97f0f66d 100644
--- a/lib/pacemaker/pcmk_status.c
+++ b/lib/pacemaker/pcmk_status.c
@@ -1,381 +1,381 @@
/*
* Copyright 2004-2023 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include <crm_internal.h>
#include <stdbool.h>
#include <stddef.h>
#include <stdint.h>
#include <crm/cib/internal.h>
#include <crm/common/output.h>
#include <crm/common/results.h>
#include <crm/fencing/internal.h>
#include <crm/pengine/internal.h>
#include <crm/stonith-ng.h>
#include <pacemaker.h>
#include <pacemaker-internal.h>
static stonith_t *
fencing_connect(void)
{
stonith_t *st = stonith_api_new();
int rc = pcmk_rc_ok;
if (st == NULL) {
return NULL;
}
rc = st->cmds->connect(st, crm_system_name, NULL);
if (rc == pcmk_rc_ok) {
return st;
} else {
stonith_api_delete(st);
return NULL;
}
}
/*!
* \internal
* \brief Output the cluster status given a fencer and CIB connection
*
* \param[in,out] out Output object
* \param[in,out] stonith Fencer connection
* \param[in,out] cib CIB connection
* \param[in] current_cib Current CIB XML
* \param[in] pcmkd_state \p pacemakerd state
* \param[in] fence_history How much of the fencing history to output
* \param[in] show Group of \p pcmk_section_e flags
* \param[in] show_opts Group of \p pcmk_show_opt_e flags
* \param[in] only_node If a node name or tag, include only the
* matching node(s) (if any) in the output.
* If \p "*" or \p NULL, include all nodes
* in the output.
* \param[in] only_rsc If a resource ID or tag, include only the
* matching resource(s) (if any) in the
* output. If \p "*" or \p NULL, include all
* resources in the output.
* \param[in] neg_location_prefix Prefix denoting a ban in a constraint ID
* \param[in] simple_output Whether to use a simple output format.
* Note: This is for use by \p crm_mon only
* and is planned to be deprecated.
*
* \return Standard Pacemaker return code
*/
int
pcmk__output_cluster_status(pcmk__output_t *out, stonith_t *stonith, cib_t *cib,
xmlNode *current_cib,
enum pcmk_pacemakerd_state pcmkd_state,
enum pcmk__fence_history fence_history,
uint32_t show, uint32_t show_opts,
const char *only_node, const char *only_rsc,
const char *neg_location_prefix, bool simple_output)
{
xmlNode *cib_copy = copy_xml(current_cib);
stonith_history_t *stonith_history = NULL;
int history_rc = 0;
pcmk_scheduler_t *scheduler = NULL;
GList *unames = NULL;
GList *resources = NULL;
int rc = pcmk_rc_ok;
if (cli_config_update(&cib_copy, NULL, FALSE) == FALSE) {
cib__clean_up_connection(&cib);
free_xml(cib_copy);
rc = pcmk_rc_schema_validation;
out->err(out, "Upgrade failed: %s", pcmk_rc_str(rc));
return rc;
}
/* get the stonith-history if there is evidence we need it */
if (fence_history != pcmk__fence_history_none) {
history_rc = pcmk__get_fencing_history(stonith, &stonith_history,
fence_history);
}
scheduler = pe_new_working_set();
CRM_ASSERT(scheduler != NULL);
- pe__set_working_set_flags(scheduler, pcmk_sched_no_compat);
+ pcmk__set_scheduler_flags(scheduler, pcmk_sched_no_compat);
scheduler->input = cib_copy;
scheduler->priv = out;
cluster_status(scheduler);
if ((cib->variant == cib_native) && pcmk_is_set(show, pcmk_section_times)) {
if (pcmk__our_nodename == NULL) {
// Currently used only in the times section
pcmk__query_node_name(out, 0, &pcmk__our_nodename, 0);
}
scheduler->localhost = pcmk__our_nodename;
}
/* Unpack constraints if any section will need them
* (tickets may be referenced in constraints but not granted yet,
* and bans need negative location constraints) */
if (pcmk_is_set(show, pcmk_section_bans)
|| pcmk_is_set(show, pcmk_section_tickets)) {
pcmk__unpack_constraints(scheduler);
}
unames = pe__build_node_name_list(scheduler, only_node);
resources = pe__build_rsc_list(scheduler, only_rsc);
/* Always print DC if NULL. */
if (scheduler->dc_node == NULL) {
show |= pcmk_section_dc;
}
if (simple_output) {
rc = pcmk__output_simple_status(out, scheduler);
} else {
out->message(out, "cluster-status",
scheduler, pcmkd_state, pcmk_rc2exitc(history_rc),
stonith_history, fence_history, show, show_opts,
neg_location_prefix, unames, resources);
}
g_list_free_full(unames, free);
g_list_free_full(resources, free);
stonith_history_free(stonith_history);
stonith_history = NULL;
pe_free_working_set(scheduler);
return rc;
}
int
pcmk_status(xmlNodePtr *xml)
{
cib_t *cib = NULL;
pcmk__output_t *out = NULL;
int rc = pcmk_rc_ok;
uint32_t show_opts = pcmk_show_pending
|pcmk_show_inactive_rscs
|pcmk_show_timing;
cib = cib_new();
if (cib == NULL) {
return pcmk_rc_cib_corrupt;
}
rc = pcmk__xml_output_new(&out, xml);
if (rc != pcmk_rc_ok) {
cib_delete(cib);
return rc;
}
pcmk__register_lib_messages(out);
pe__register_messages(out);
stonith__register_messages(out);
rc = pcmk__status(out, cib, pcmk__fence_history_full, pcmk_section_all,
show_opts, NULL, NULL, NULL, false, 0);
pcmk__xml_output_finish(out, pcmk_rc2exitc(rc), xml);
cib_delete(cib);
return rc;
}
/*!
* \internal
* \brief Query and output the cluster status
*
* The operation is considered a success if we're able to get the \p pacemakerd
* state. If possible, we'll also try to connect to the fencer and CIB and
* output their respective status information.
*
* \param[in,out] out Output object
* \param[in,out] cib CIB connection
* \param[in] fence_history How much of the fencing history to output
* \param[in] show Group of \p pcmk_section_e flags
* \param[in] show_opts Group of \p pcmk_show_opt_e flags
* \param[in] only_node If a node name or tag, include only the
* matching node(s) (if any) in the output.
* If \p "*" or \p NULL, include all nodes
* in the output.
* \param[in] only_rsc If a resource ID or tag, include only the
* matching resource(s) (if any) in the
* output. If \p "*" or \p NULL, include all
* resources in the output.
* \param[in] neg_location_prefix Prefix denoting a ban in a constraint ID
* \param[in] simple_output Whether to use a simple output format.
* Note: This is for use by \p crm_mon only
* and is planned to be deprecated.
* \param[in] timeout_ms How long to wait for a reply from the
* \p pacemakerd API. If 0,
* \p pcmk_ipc_dispatch_sync will be used.
* If positive, \p pcmk_ipc_dispatch_main
* will be used, and a new mainloop will be
* created for this purpose (freed before
* return).
*
* \return Standard Pacemaker return code
*/
int
pcmk__status(pcmk__output_t *out, cib_t *cib,
enum pcmk__fence_history fence_history, uint32_t show,
uint32_t show_opts, const char *only_node, const char *only_rsc,
const char *neg_location_prefix, bool simple_output,
unsigned int timeout_ms)
{
xmlNode *current_cib = NULL;
int rc = pcmk_rc_ok;
stonith_t *stonith = NULL;
enum pcmk_pacemakerd_state pcmkd_state = pcmk_pacemakerd_state_invalid;
time_t last_updated = 0;
if (cib == NULL) {
return ENOTCONN;
}
if (cib->variant == cib_native) {
rc = pcmk__pacemakerd_status(out, crm_system_name, timeout_ms, false,
&pcmkd_state);
if (rc != pcmk_rc_ok) {
return rc;
}
last_updated = time(NULL);
switch (pcmkd_state) {
case pcmk_pacemakerd_state_running:
case pcmk_pacemakerd_state_shutting_down:
case pcmk_pacemakerd_state_remote:
/* Fencer and CIB may still be available while shutting down or
* running on a Pacemaker Remote node
*/
break;
default:
// Fencer and CIB are definitely unavailable
out->message(out, "pacemakerd-health",
NULL, pcmkd_state, NULL, last_updated);
return rc;
}
if (fence_history != pcmk__fence_history_none) {
stonith = fencing_connect();
}
}
rc = cib__signon_query(out, &cib, &current_cib);
if (rc != pcmk_rc_ok) {
if (pcmkd_state != pcmk_pacemakerd_state_invalid) {
// Invalid at this point means we didn't query the pcmkd state
out->message(out, "pacemakerd-health",
NULL, pcmkd_state, NULL, last_updated);
}
goto done;
}
rc = pcmk__output_cluster_status(out, stonith, cib, current_cib,
pcmkd_state, fence_history, show,
show_opts, only_node, only_rsc,
neg_location_prefix, simple_output);
if (rc != pcmk_rc_ok) {
out->err(out, "Error outputting status info from the fencer or CIB");
}
done:
stonith_api_delete(stonith);
free_xml(current_cib);
return pcmk_rc_ok;
}
/*!
* \internal
* \brief Output cluster status in Nagios Plugin format
*
* \param[in,out] out Output object
* \param[in] scheduler Scheduler data
*
* \return Standard Pacemaker return code
* \note This is for a deprecated crm_mon option and should be called only for
* that.
*/
int
pcmk__output_simple_status(pcmk__output_t *out,
const pcmk_scheduler_t *scheduler)
{
int nodes_online = 0;
int nodes_standby = 0;
int nodes_maint = 0;
GString *offline_nodes = NULL;
bool no_dc = false;
bool offline = false;
bool has_warnings = false;
if (scheduler->dc_node == NULL) {
has_warnings = true;
no_dc = true;
}
for (GList *iter = scheduler->nodes; iter != NULL; iter = iter->next) {
pcmk_node_t *node = (pcmk_node_t *) iter->data;
if (node->details->standby && node->details->online) {
nodes_standby++;
} else if (node->details->maintenance && node->details->online) {
nodes_maint++;
} else if (node->details->online) {
nodes_online++;
} else {
pcmk__add_word(&offline_nodes, 1024, "offline node:");
pcmk__add_word(&offline_nodes, 0, pe__node_name(node));
has_warnings = true;
offline = true;
}
}
if (has_warnings) {
out->info(out, "CLUSTER WARN: %s%s%s",
no_dc ? "No DC" : "",
no_dc && offline ? ", " : "",
(offline? (const char *) offline_nodes->str : ""));
if (offline_nodes != NULL) {
g_string_free(offline_nodes, TRUE);
}
} else {
char *nodes_standby_s = NULL;
char *nodes_maint_s = NULL;
if (nodes_standby > 0) {
nodes_standby_s = crm_strdup_printf(", %d standby node%s",
nodes_standby,
pcmk__plural_s(nodes_standby));
}
if (nodes_maint > 0) {
nodes_maint_s = crm_strdup_printf(", %d maintenance node%s",
nodes_maint,
pcmk__plural_s(nodes_maint));
}
out->info(out, "CLUSTER OK: %d node%s online%s%s, "
"%d resource instance%s configured",
nodes_online, pcmk__plural_s(nodes_online),
nodes_standby_s != NULL ? nodes_standby_s : "",
nodes_maint_s != NULL ? nodes_maint_s : "",
scheduler->ninstances, pcmk__plural_s(scheduler->ninstances));
free(nodes_standby_s);
free(nodes_maint_s);
}
if (has_warnings) {
return pcmk_rc_error;
} else {
return pcmk_rc_ok;
}
/* coverity[leaked_storage] False positive */
}
diff --git a/lib/pengine/complex.c b/lib/pengine/complex.c
index 75bcd56527..50cfaa6de9 100644
--- a/lib/pengine/complex.c
+++ b/lib/pengine/complex.c
@@ -1,1202 +1,1202 @@
/*
* 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 <crm/pengine/rules.h>
#include <crm/pengine/internal.h>
#include <crm/msg_xml.h>
#include <crm/common/xml_internal.h>
#include <crm/common/scheduler_internal.h>
#include "pe_status_private.h"
void populate_hash(xmlNode * nvpair_list, GHashTable * hash, const char **attrs, int attrs_length);
static pcmk_node_t *active_node(const pcmk_resource_t *rsc,
unsigned int *count_all,
unsigned int *count_clean);
pcmk_rsc_methods_t resource_class_functions[] = {
{
native_unpack,
native_find_rsc,
native_parameter,
native_print,
native_active,
native_resource_state,
native_location,
native_free,
pe__count_common,
pe__native_is_filtered,
active_node,
pe__primitive_max_per_node,
},
{
group_unpack,
native_find_rsc,
native_parameter,
group_print,
group_active,
group_resource_state,
native_location,
group_free,
pe__count_common,
pe__group_is_filtered,
active_node,
pe__group_max_per_node,
},
{
clone_unpack,
native_find_rsc,
native_parameter,
clone_print,
clone_active,
clone_resource_state,
native_location,
clone_free,
pe__count_common,
pe__clone_is_filtered,
active_node,
pe__clone_max_per_node,
},
{
pe__unpack_bundle,
native_find_rsc,
native_parameter,
pe__print_bundle,
pe__bundle_active,
pe__bundle_resource_state,
native_location,
pe__free_bundle,
pe__count_bundle,
pe__bundle_is_filtered,
pe__bundle_active_node,
pe__bundle_max_per_node,
}
};
static enum pe_obj_types
get_resource_type(const char *name)
{
if (pcmk__str_eq(name, XML_CIB_TAG_RESOURCE, pcmk__str_casei)) {
return pcmk_rsc_variant_primitive;
} else if (pcmk__str_eq(name, XML_CIB_TAG_GROUP, pcmk__str_casei)) {
return pcmk_rsc_variant_group;
} else if (pcmk__str_eq(name, XML_CIB_TAG_INCARNATION, pcmk__str_casei)) {
return pcmk_rsc_variant_clone;
} else if (pcmk__str_eq(name, PCMK_XE_PROMOTABLE_LEGACY, pcmk__str_casei)) {
// @COMPAT deprecated since 2.0.0
return pcmk_rsc_variant_clone;
} else if (pcmk__str_eq(name, XML_CIB_TAG_CONTAINER, pcmk__str_casei)) {
return pcmk_rsc_variant_bundle;
}
return pcmk_rsc_variant_unknown;
}
static void
dup_attr(gpointer key, gpointer value, gpointer user_data)
{
add_hash_param(user_data, key, value);
}
static void
expand_parents_fixed_nvpairs(pcmk_resource_t *rsc,
pe_rule_eval_data_t *rule_data,
GHashTable *meta_hash, pcmk_scheduler_t *scheduler)
{
GHashTable *parent_orig_meta = pcmk__strkey_table(free, free);
pcmk_resource_t *p = rsc->parent;
if (p == NULL) {
return ;
}
/* Search all parent resources, get the fixed value of "meta_attributes" set only in the original xml, and stack it in the hash table. */
/* The fixed value of the lower parent resource takes precedence and is not overwritten. */
while(p != NULL) {
/* A hash table for comparison is generated, including the id-ref. */
pe__unpack_dataset_nvpairs(p->xml, XML_TAG_META_SETS, rule_data,
parent_orig_meta, NULL, FALSE, scheduler);
p = p->parent;
}
/* If there is a fixed value of "meta_attributes" of the parent resource, it will be processed. */
if (parent_orig_meta != NULL) {
GHashTableIter iter;
char *key = NULL;
char *value = NULL;
g_hash_table_iter_init(&iter, parent_orig_meta);
while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) {
/* Parameters set in the original xml of the parent resource will also try to overwrite the child resource. */
/* Attributes that already exist in the child lease are not updated. */
dup_attr(key, value, meta_hash);
}
}
if (parent_orig_meta != NULL) {
g_hash_table_destroy(parent_orig_meta);
}
return ;
}
void
get_meta_attributes(GHashTable * meta_hash, pcmk_resource_t * rsc,
pcmk_node_t *node, pcmk_scheduler_t *scheduler)
{
pe_rsc_eval_data_t rsc_rule_data = {
.standard = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS),
.provider = crm_element_value(rsc->xml, XML_AGENT_ATTR_PROVIDER),
.agent = crm_element_value(rsc->xml, XML_ATTR_TYPE)
};
pe_rule_eval_data_t rule_data = {
.node_hash = NULL,
.role = pcmk_role_unknown,
.now = scheduler->now,
.match_data = NULL,
.rsc_data = &rsc_rule_data,
.op_data = NULL
};
if (node) {
rule_data.node_hash = node->details->attrs;
}
for (xmlAttrPtr a = pcmk__xe_first_attr(rsc->xml); a != NULL; a = a->next) {
const char *prop_name = (const char *) a->name;
const char *prop_value = pcmk__xml_attr_value(a);
add_hash_param(meta_hash, prop_name, prop_value);
}
pe__unpack_dataset_nvpairs(rsc->xml, XML_TAG_META_SETS, &rule_data,
meta_hash, NULL, FALSE, scheduler);
/* Set the "meta_attributes" explicitly set in the parent resource to the hash table of the child resource. */
/* If it is already explicitly set as a child, it will not be overwritten. */
if (rsc->parent != NULL) {
expand_parents_fixed_nvpairs(rsc, &rule_data, meta_hash, scheduler);
}
/* check the defaults */
pe__unpack_dataset_nvpairs(scheduler->rsc_defaults, XML_TAG_META_SETS,
&rule_data, meta_hash, NULL, FALSE, scheduler);
/* If there is "meta_attributes" that the parent resource has not explicitly set, set a value that is not set from rsc_default either. */
/* The values already set up to this point will not be overwritten. */
if (rsc->parent) {
g_hash_table_foreach(rsc->parent->meta, dup_attr, meta_hash);
}
}
void
get_rsc_attributes(GHashTable *meta_hash, const pcmk_resource_t *rsc,
const pcmk_node_t *node, pcmk_scheduler_t *scheduler)
{
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
};
if (node) {
rule_data.node_hash = node->details->attrs;
}
pe__unpack_dataset_nvpairs(rsc->xml, XML_TAG_ATTR_SETS, &rule_data,
meta_hash, NULL, FALSE, scheduler);
/* set anything else based on the parent */
if (rsc->parent != NULL) {
get_rsc_attributes(meta_hash, rsc->parent, node, scheduler);
} else {
/* and finally check the defaults */
pe__unpack_dataset_nvpairs(scheduler->rsc_defaults, XML_TAG_ATTR_SETS,
&rule_data, meta_hash, NULL, FALSE,
scheduler);
}
}
static char *
template_op_key(xmlNode * op)
{
const char *name = crm_element_value(op, "name");
const char *role = crm_element_value(op, "role");
char *key = NULL;
if ((role == NULL)
|| pcmk__strcase_any_of(role, PCMK__ROLE_STARTED, PCMK__ROLE_UNPROMOTED,
PCMK__ROLE_UNPROMOTED_LEGACY, NULL)) {
role = PCMK__ROLE_UNKNOWN;
}
key = crm_strdup_printf("%s-%s", name, role);
return key;
}
static gboolean
unpack_template(xmlNode *xml_obj, xmlNode **expanded_xml,
pcmk_scheduler_t *scheduler)
{
xmlNode *cib_resources = NULL;
xmlNode *template = NULL;
xmlNode *new_xml = NULL;
xmlNode *child_xml = NULL;
xmlNode *rsc_ops = NULL;
xmlNode *template_ops = NULL;
const char *template_ref = NULL;
const char *id = NULL;
if (xml_obj == NULL) {
pcmk__config_err("No resource object for template unpacking");
return FALSE;
}
template_ref = crm_element_value(xml_obj, XML_CIB_TAG_RSC_TEMPLATE);
if (template_ref == NULL) {
return TRUE;
}
id = ID(xml_obj);
if (id == NULL) {
pcmk__config_err("'%s' object must have a id", xml_obj->name);
return FALSE;
}
if (pcmk__str_eq(template_ref, id, pcmk__str_none)) {
pcmk__config_err("The resource object '%s' should not reference itself",
id);
return FALSE;
}
cib_resources = get_xpath_object("//" XML_CIB_TAG_RESOURCES,
scheduler->input, LOG_TRACE);
if (cib_resources == NULL) {
pcmk__config_err("No resources configured");
return FALSE;
}
template = pcmk__xe_match(cib_resources, XML_CIB_TAG_RSC_TEMPLATE,
PCMK_XA_ID, template_ref);
if (template == NULL) {
pcmk__config_err("No template named '%s'", template_ref);
return FALSE;
}
new_xml = copy_xml(template);
xmlNodeSetName(new_xml, xml_obj->name);
crm_xml_add(new_xml, PCMK_XA_ID, id);
crm_xml_add(new_xml, PCMK__META_CLONE_INSTANCE_NUM,
crm_element_value(xml_obj, PCMK__META_CLONE_INSTANCE_NUM));
template_ops = find_xml_node(new_xml, "operations", FALSE);
for (child_xml = pcmk__xe_first_child(xml_obj); child_xml != NULL;
child_xml = pcmk__xe_next(child_xml)) {
xmlNode *new_child = NULL;
new_child = add_node_copy(new_xml, child_xml);
if (pcmk__str_eq((const char *)new_child->name, "operations", pcmk__str_none)) {
rsc_ops = new_child;
}
}
if (template_ops && rsc_ops) {
xmlNode *op = NULL;
GHashTable *rsc_ops_hash = pcmk__strkey_table(free, NULL);
for (op = pcmk__xe_first_child(rsc_ops); op != NULL;
op = pcmk__xe_next(op)) {
char *key = template_op_key(op);
g_hash_table_insert(rsc_ops_hash, key, op);
}
for (op = pcmk__xe_first_child(template_ops); op != NULL;
op = pcmk__xe_next(op)) {
char *key = template_op_key(op);
if (g_hash_table_lookup(rsc_ops_hash, key) == NULL) {
add_node_copy(rsc_ops, op);
}
free(key);
}
if (rsc_ops_hash) {
g_hash_table_destroy(rsc_ops_hash);
}
free_xml(template_ops);
}
/*free_xml(*expanded_xml); */
*expanded_xml = new_xml;
#if 0 /* Disable multi-level templates for now */
if (!unpack_template(new_xml, expanded_xml, scheduler)) {
free_xml(*expanded_xml);
*expanded_xml = NULL;
return FALSE;
}
#endif
return TRUE;
}
static gboolean
add_template_rsc(xmlNode *xml_obj, pcmk_scheduler_t *scheduler)
{
const char *template_ref = NULL;
const char *id = NULL;
if (xml_obj == NULL) {
pcmk__config_err("No resource object for processing resource list "
"of template");
return FALSE;
}
template_ref = crm_element_value(xml_obj, XML_CIB_TAG_RSC_TEMPLATE);
if (template_ref == NULL) {
return TRUE;
}
id = ID(xml_obj);
if (id == NULL) {
pcmk__config_err("'%s' object must have a id", xml_obj->name);
return FALSE;
}
if (pcmk__str_eq(template_ref, id, pcmk__str_none)) {
pcmk__config_err("The resource object '%s' should not reference itself",
id);
return FALSE;
}
if (add_tag_ref(scheduler->template_rsc_sets, template_ref, id) == FALSE) {
return FALSE;
}
return TRUE;
}
static bool
detect_promotable(pcmk_resource_t *rsc)
{
const char *promotable = g_hash_table_lookup(rsc->meta,
PCMK_META_PROMOTABLE);
if (crm_is_true(promotable)) {
return TRUE;
}
// @COMPAT deprecated since 2.0.0
if (pcmk__xe_is(rsc->xml, PCMK_XE_PROMOTABLE_LEGACY)) {
/* @TODO in some future version, pcmk__warn_once() here,
* then drop support in even later version
*/
g_hash_table_insert(rsc->meta, strdup(PCMK_META_PROMOTABLE),
strdup(XML_BOOLEAN_TRUE));
return TRUE;
}
return FALSE;
}
static void
free_params_table(gpointer data)
{
g_hash_table_destroy((GHashTable *) data);
}
/*!
* \brief Get a table of resource parameters
*
* \param[in,out] rsc Resource to query
* \param[in] node Node for evaluating rules (NULL for defaults)
* \param[in,out] scheduler Scheduler data
*
* \return Hash table containing resource parameter names and values
* (or NULL if \p rsc or \p scheduler is NULL)
* \note The returned table will be destroyed when the resource is freed, so
* callers should not destroy it.
*/
GHashTable *
pe_rsc_params(pcmk_resource_t *rsc, const pcmk_node_t *node,
pcmk_scheduler_t *scheduler)
{
GHashTable *params_on_node = NULL;
/* A NULL node is used to request the resource's default parameters
* (not evaluated for node), but we always want something non-NULL
* as a hash table key.
*/
const char *node_name = "";
// Sanity check
if ((rsc == NULL) || (scheduler == NULL)) {
return NULL;
}
if ((node != NULL) && (node->details->uname != NULL)) {
node_name = node->details->uname;
}
// Find the parameter table for given node
if (rsc->parameter_cache == NULL) {
rsc->parameter_cache = pcmk__strikey_table(free, free_params_table);
} else {
params_on_node = g_hash_table_lookup(rsc->parameter_cache, node_name);
}
// If none exists yet, create one with parameters evaluated for node
if (params_on_node == NULL) {
params_on_node = pcmk__strkey_table(free, free);
get_rsc_attributes(params_on_node, rsc, node, scheduler);
g_hash_table_insert(rsc->parameter_cache, strdup(node_name),
params_on_node);
}
return params_on_node;
}
/*!
* \internal
* \brief Unpack a resource's "requires" meta-attribute
*
* \param[in,out] rsc Resource being unpacked
* \param[in] value Value of "requires" meta-attribute
* \param[in] is_default Whether \p value was selected by default
*/
static void
unpack_requires(pcmk_resource_t *rsc, const char *value, bool is_default)
{
if (pcmk__str_eq(value, PCMK__VALUE_NOTHING, pcmk__str_casei)) {
} else if (pcmk__str_eq(value, PCMK__VALUE_QUORUM, pcmk__str_casei)) {
pe__set_resource_flags(rsc, pcmk_rsc_needs_quorum);
} else if (pcmk__str_eq(value, PCMK__VALUE_FENCING, pcmk__str_casei)) {
pe__set_resource_flags(rsc, pcmk_rsc_needs_fencing);
if (!pcmk_is_set(rsc->cluster->flags, pcmk_sched_fencing_enabled)) {
pcmk__config_warn("%s requires fencing but fencing is disabled",
rsc->id);
}
} else if (pcmk__str_eq(value, PCMK__VALUE_UNFENCING, pcmk__str_casei)) {
if (pcmk_is_set(rsc->flags, pcmk_rsc_fence_device)) {
pcmk__config_warn("Resetting \"" PCMK_META_REQUIRES "\" for %s "
"to \"" PCMK__VALUE_QUORUM "\" because fencing "
"devices cannot require unfencing", rsc->id);
unpack_requires(rsc, PCMK__VALUE_QUORUM, true);
return;
} else if (!pcmk_is_set(rsc->cluster->flags,
pcmk_sched_fencing_enabled)) {
pcmk__config_warn("Resetting \"" PCMK_META_REQUIRES "\" for %s "
"to \"" PCMK__VALUE_QUORUM "\" because fencing "
"is disabled", rsc->id);
unpack_requires(rsc, PCMK__VALUE_QUORUM, true);
return;
} else {
pe__set_resource_flags(rsc, pcmk_rsc_needs_fencing
|pcmk_rsc_needs_unfencing);
}
} else {
const char *orig_value = value;
if (pcmk_is_set(rsc->flags, pcmk_rsc_fence_device)) {
value = PCMK__VALUE_QUORUM;
} else if ((rsc->variant == pcmk_rsc_variant_primitive)
&& xml_contains_remote_node(rsc->xml)) {
value = PCMK__VALUE_QUORUM;
} else if (pcmk_is_set(rsc->cluster->flags,
pcmk_sched_enable_unfencing)) {
value = PCMK__VALUE_UNFENCING;
} else if (pcmk_is_set(rsc->cluster->flags,
pcmk_sched_fencing_enabled)) {
value = PCMK__VALUE_FENCING;
} else if (rsc->cluster->no_quorum_policy == pcmk_no_quorum_ignore) {
value = PCMK__VALUE_NOTHING;
} else {
value = PCMK__VALUE_QUORUM;
}
if (orig_value != NULL) {
pcmk__config_err("Resetting '" PCMK_META_REQUIRES "' for %s "
"to '%s' because '%s' is not valid",
rsc->id, value, orig_value);
}
unpack_requires(rsc, value, true);
return;
}
pcmk__rsc_trace(rsc, "\tRequired to start: %s%s", value,
(is_default? " (default)" : ""));
}
#ifndef PCMK__COMPAT_2_0
static void
warn_about_deprecated_classes(pcmk_resource_t *rsc)
{
const char *std = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS);
if (pcmk__str_eq(std, PCMK_RESOURCE_CLASS_UPSTART, pcmk__str_none)) {
pcmk__warn_once(pcmk__wo_upstart,
"Support for Upstart resources (such as %s) is "
"deprecated and will be removed in a future release",
rsc->id);
} else if (pcmk__str_eq(std, PCMK_RESOURCE_CLASS_NAGIOS, pcmk__str_none)) {
pcmk__warn_once(pcmk__wo_nagios,
"Support for Nagios resources (such as %s) is "
"deprecated and will be removed in a future release",
rsc->id);
}
}
#endif
/*!
* \internal
* \brief Unpack configuration XML for a given resource
*
* Unpack the XML object containing a resource's configuration into a new
* \c pcmk_resource_t object.
*
* \param[in] xml_obj XML node containing the resource's configuration
* \param[out] rsc Where to store the unpacked resource information
* \param[in] parent Resource's parent, if any
* \param[in,out] scheduler Scheduler data
*
* \return Standard Pacemaker return code
* \note If pcmk_rc_ok is returned, \p *rsc is guaranteed to be non-NULL, and
* the caller is responsible for freeing it using its variant-specific
* free() method. Otherwise, \p *rsc is guaranteed to be NULL.
*/
int
pe__unpack_resource(xmlNode *xml_obj, pcmk_resource_t **rsc,
pcmk_resource_t *parent, pcmk_scheduler_t *scheduler)
{
xmlNode *expanded_xml = NULL;
xmlNode *ops = NULL;
const char *value = NULL;
const char *id = NULL;
bool guest_node = false;
bool remote_node = false;
pe_rule_eval_data_t rule_data = {
.node_hash = NULL,
.role = pcmk_role_unknown,
.now = NULL,
.match_data = NULL,
.rsc_data = NULL,
.op_data = NULL
};
CRM_CHECK(rsc != NULL, return EINVAL);
CRM_CHECK((xml_obj != NULL) && (scheduler != NULL),
*rsc = NULL;
return EINVAL);
rule_data.now = scheduler->now;
crm_log_xml_trace(xml_obj, "[raw XML]");
id = crm_element_value(xml_obj, PCMK_XA_ID);
if (id == NULL) {
pcmk__config_err("Ignoring <%s> configuration without " PCMK_XA_ID,
xml_obj->name);
return pcmk_rc_unpack_error;
}
if (unpack_template(xml_obj, &expanded_xml, scheduler) == FALSE) {
return pcmk_rc_unpack_error;
}
*rsc = calloc(1, sizeof(pcmk_resource_t));
if (*rsc == NULL) {
pcmk__sched_err("Unable to allocate memory for resource '%s'", id);
return ENOMEM;
}
(*rsc)->cluster = scheduler;
if (expanded_xml) {
crm_log_xml_trace(expanded_xml, "[expanded XML]");
(*rsc)->xml = expanded_xml;
(*rsc)->orig_xml = xml_obj;
} else {
(*rsc)->xml = xml_obj;
(*rsc)->orig_xml = NULL;
}
/* Do not use xml_obj from here on, use (*rsc)->xml in case templates are involved */
(*rsc)->parent = parent;
ops = find_xml_node((*rsc)->xml, "operations", FALSE);
(*rsc)->ops_xml = expand_idref(ops, scheduler->input);
(*rsc)->variant = get_resource_type((const char *) (*rsc)->xml->name);
if ((*rsc)->variant == pcmk_rsc_variant_unknown) {
pcmk__config_err("Ignoring resource '%s' of unknown type '%s'",
id, (*rsc)->xml->name);
common_free(*rsc);
*rsc = NULL;
return pcmk_rc_unpack_error;
}
#ifndef PCMK__COMPAT_2_0
warn_about_deprecated_classes(*rsc);
#endif
(*rsc)->meta = pcmk__strkey_table(free, free);
(*rsc)->allowed_nodes = pcmk__strkey_table(NULL, free);
(*rsc)->known_on = pcmk__strkey_table(NULL, free);
value = crm_element_value((*rsc)->xml, PCMK__META_CLONE_INSTANCE_NUM);
if (value) {
(*rsc)->id = crm_strdup_printf("%s:%s", id, value);
add_hash_param((*rsc)->meta, PCMK__META_CLONE_INSTANCE_NUM, value);
} else {
(*rsc)->id = strdup(id);
}
(*rsc)->fns = &resource_class_functions[(*rsc)->variant];
get_meta_attributes((*rsc)->meta, *rsc, NULL, scheduler);
(*rsc)->parameters = pe_rsc_params(*rsc, NULL, scheduler); // \deprecated
(*rsc)->flags = 0;
pe__set_resource_flags(*rsc, pcmk_rsc_runnable|pcmk_rsc_unassigned);
if (!pcmk_is_set(scheduler->flags, pcmk_sched_in_maintenance)) {
pe__set_resource_flags(*rsc, pcmk_rsc_managed);
}
(*rsc)->rsc_cons = NULL;
(*rsc)->rsc_tickets = NULL;
(*rsc)->actions = NULL;
(*rsc)->role = pcmk_role_stopped;
(*rsc)->next_role = pcmk_role_unknown;
(*rsc)->recovery_type = pcmk_multiply_active_restart;
(*rsc)->stickiness = 0;
(*rsc)->migration_threshold = INFINITY;
(*rsc)->failure_timeout = 0;
value = g_hash_table_lookup((*rsc)->meta, PCMK_META_PRIORITY);
(*rsc)->priority = char2score(value);
value = g_hash_table_lookup((*rsc)->meta, PCMK_META_CRITICAL);
if ((value == NULL) || crm_is_true(value)) {
pe__set_resource_flags(*rsc, pcmk_rsc_critical);
}
value = g_hash_table_lookup((*rsc)->meta, PCMK_META_NOTIFY);
if (crm_is_true(value)) {
pe__set_resource_flags(*rsc, pcmk_rsc_notify);
}
if (xml_contains_remote_node((*rsc)->xml)) {
(*rsc)->is_remote_node = TRUE;
if (g_hash_table_lookup((*rsc)->meta, PCMK__META_CONTAINER)) {
guest_node = true;
} else {
remote_node = true;
}
}
value = g_hash_table_lookup((*rsc)->meta, PCMK_META_ALLOW_MIGRATE);
if (crm_is_true(value)) {
pe__set_resource_flags(*rsc, pcmk_rsc_migratable);
} else if ((value == NULL) && remote_node) {
/* By default, we want remote nodes to be able
* to float around the cluster without having to stop all the
* resources within the remote-node before moving. Allowing
* migration support enables this feature. If this ever causes
* problems, migration support can be explicitly turned off with
* PCMK_META_ALLOW_MIGRATE=false.
*/
pe__set_resource_flags(*rsc, pcmk_rsc_migratable);
}
value = g_hash_table_lookup((*rsc)->meta, PCMK_META_IS_MANAGED);
if (value != NULL && !pcmk__str_eq("default", value, pcmk__str_casei)) {
if (crm_is_true(value)) {
pe__set_resource_flags(*rsc, pcmk_rsc_managed);
} else {
pe__clear_resource_flags(*rsc, pcmk_rsc_managed);
}
}
value = g_hash_table_lookup((*rsc)->meta, PCMK_META_MAINTENANCE);
if (crm_is_true(value)) {
pe__clear_resource_flags(*rsc, pcmk_rsc_managed);
pe__set_resource_flags(*rsc, pcmk_rsc_maintenance);
}
if (pcmk_is_set(scheduler->flags, pcmk_sched_in_maintenance)) {
pe__clear_resource_flags(*rsc, pcmk_rsc_managed);
pe__set_resource_flags(*rsc, pcmk_rsc_maintenance);
}
if (pe_rsc_is_clone(pe__const_top_resource(*rsc, false))) {
value = g_hash_table_lookup((*rsc)->meta, PCMK_META_GLOBALLY_UNIQUE);
if (crm_is_true(value)) {
pe__set_resource_flags(*rsc, pcmk_rsc_unique);
}
if (detect_promotable(*rsc)) {
pe__set_resource_flags(*rsc, pcmk_rsc_promotable);
}
} else {
pe__set_resource_flags(*rsc, pcmk_rsc_unique);
}
// @COMPAT Deprecated meta-attribute
value = g_hash_table_lookup((*rsc)->meta, PCMK__META_RESTART_TYPE);
if (pcmk__str_eq(value, "restart", pcmk__str_casei)) {
(*rsc)->restart_type = pe_restart_restart;
pcmk__rsc_trace(*rsc, "%s dependency restart handling: restart",
(*rsc)->id);
pcmk__warn_once(pcmk__wo_restart_type,
"Support for " PCMK__META_RESTART_TYPE " is deprecated "
"and will be removed in a future release");
} else {
(*rsc)->restart_type = pe_restart_ignore;
pcmk__rsc_trace(*rsc, "%s dependency restart handling: ignore",
(*rsc)->id);
}
value = g_hash_table_lookup((*rsc)->meta, PCMK_META_MULTIPLE_ACTIVE);
if (pcmk__str_eq(value, "stop_only", pcmk__str_casei)) {
(*rsc)->recovery_type = pcmk_multiply_active_stop;
pcmk__rsc_trace(*rsc, "%s multiple running resource recovery: stop only",
(*rsc)->id);
} else if (pcmk__str_eq(value, "block", pcmk__str_casei)) {
(*rsc)->recovery_type = pcmk_multiply_active_block;
pcmk__rsc_trace(*rsc, "%s multiple running resource recovery: block",
(*rsc)->id);
} else if (pcmk__str_eq(value, "stop_unexpected", pcmk__str_casei)) {
(*rsc)->recovery_type = pcmk_multiply_active_unexpected;
pcmk__rsc_trace(*rsc,
"%s multiple running resource recovery: "
"stop unexpected instances",
(*rsc)->id);
} else { // "stop_start"
if (!pcmk__str_eq(value, "stop_start",
pcmk__str_casei|pcmk__str_null_matches)) {
pcmk__config_warn("%s is not a valid value for "
PCMK_META_MULTIPLE_ACTIVE
", using default of \"stop_start\"", value);
}
(*rsc)->recovery_type = pcmk_multiply_active_restart;
pcmk__rsc_trace(*rsc,
"%s multiple running resource recovery: stop/start",
(*rsc)->id);
}
value = g_hash_table_lookup((*rsc)->meta, PCMK_META_RESOURCE_STICKINESS);
if (value != NULL && !pcmk__str_eq("default", value, pcmk__str_casei)) {
(*rsc)->stickiness = char2score(value);
}
value = g_hash_table_lookup((*rsc)->meta, PCMK_META_MIGRATION_THRESHOLD);
if (value != NULL && !pcmk__str_eq("default", value, pcmk__str_casei)) {
(*rsc)->migration_threshold = char2score(value);
if ((*rsc)->migration_threshold < 0) {
/* @TODO We use 1 here to preserve previous behavior, but this
* should probably use the default (INFINITY) or 0 (to disable)
* instead.
*/
pcmk__warn_once(pcmk__wo_neg_threshold,
PCMK_META_MIGRATION_THRESHOLD
" must be non-negative, using 1 instead");
(*rsc)->migration_threshold = 1;
}
}
if (pcmk__str_eq(crm_element_value((*rsc)->xml, XML_AGENT_ATTR_CLASS),
PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) {
- pe__set_working_set_flags(scheduler, pcmk_sched_have_fencing);
+ pcmk__set_scheduler_flags(scheduler, pcmk_sched_have_fencing);
pe__set_resource_flags(*rsc, pcmk_rsc_fence_device);
}
value = g_hash_table_lookup((*rsc)->meta, PCMK_META_REQUIRES);
unpack_requires(*rsc, value, false);
value = g_hash_table_lookup((*rsc)->meta, PCMK_META_FAILURE_TIMEOUT);
if (value != NULL) {
guint interval_ms = 0U;
// Stored as seconds
pcmk_parse_interval_spec(value, &interval_ms);
(*rsc)->failure_timeout = (int) (interval_ms / 1000);
}
if (remote_node) {
GHashTable *params = pe_rsc_params(*rsc, NULL, scheduler);
/* Grabbing the value now means that any rules based on node attributes
* will evaluate to false, so such rules should not be used with
* PCMK_REMOTE_RA_RECONNECT_INTERVAL.
*
* @TODO Evaluate per node before using
*/
value = g_hash_table_lookup(params, PCMK_REMOTE_RA_RECONNECT_INTERVAL);
if (value) {
/* reconnect delay works by setting failure_timeout and preventing the
* connection from starting until the failure is cleared. */
pcmk_parse_interval_spec(value, &((*rsc)->remote_reconnect_ms));
/* We want to override any default failure_timeout in use when remote
* PCMK_REMOTE_RA_RECONNECT_INTERVAL is in use.
*/
(*rsc)->failure_timeout = (*rsc)->remote_reconnect_ms / 1000;
}
}
get_target_role(*rsc, &((*rsc)->next_role));
pcmk__rsc_trace(*rsc, "%s desired next state: %s", (*rsc)->id,
(*rsc)->next_role != pcmk_role_unknown? role2text((*rsc)->next_role) : "default");
if ((*rsc)->fns->unpack(*rsc, scheduler) == FALSE) {
(*rsc)->fns->free(*rsc);
*rsc = NULL;
return pcmk_rc_unpack_error;
}
if (pcmk_is_set(scheduler->flags, pcmk_sched_symmetric_cluster)) {
// This tag must stay exactly the same because it is tested elsewhere
resource_location(*rsc, NULL, 0, "symmetric_default", scheduler);
} else if (guest_node) {
/* remote resources tied to a container resource must always be allowed
* to opt-in to the cluster. Whether the connection resource is actually
* allowed to be placed on a node is dependent on the container resource */
resource_location(*rsc, NULL, 0, "remote_connection_default",
scheduler);
}
pcmk__rsc_trace(*rsc, "%s action notification: %s", (*rsc)->id,
pcmk_is_set((*rsc)->flags, pcmk_rsc_notify)? "required" : "not required");
(*rsc)->utilization = pcmk__strkey_table(free, free);
pe__unpack_dataset_nvpairs((*rsc)->xml, XML_TAG_UTILIZATION, &rule_data,
(*rsc)->utilization, NULL, FALSE, scheduler);
if (expanded_xml) {
if (add_template_rsc(xml_obj, scheduler) == FALSE) {
(*rsc)->fns->free(*rsc);
*rsc = NULL;
return pcmk_rc_unpack_error;
}
}
return pcmk_rc_ok;
}
gboolean
is_parent(pcmk_resource_t *child, pcmk_resource_t *rsc)
{
pcmk_resource_t *parent = child;
if (parent == NULL || rsc == NULL) {
return FALSE;
}
while (parent->parent != NULL) {
if (parent->parent == rsc) {
return TRUE;
}
parent = parent->parent;
}
return FALSE;
}
pcmk_resource_t *
uber_parent(pcmk_resource_t *rsc)
{
pcmk_resource_t *parent = rsc;
if (parent == NULL) {
return NULL;
}
while ((parent->parent != NULL)
&& (parent->parent->variant != pcmk_rsc_variant_bundle)) {
parent = parent->parent;
}
return parent;
}
/*!
* \internal
* \brief Get the topmost parent of a resource as a const pointer
*
* \param[in] rsc Resource to check
* \param[in] include_bundle If true, go all the way to bundle
*
* \return \p NULL if \p rsc is NULL, \p rsc if \p rsc has no parent,
* the bundle if \p rsc is bundled and \p include_bundle is true,
* otherwise the topmost parent of \p rsc up to a clone
*/
const pcmk_resource_t *
pe__const_top_resource(const pcmk_resource_t *rsc, bool include_bundle)
{
const pcmk_resource_t *parent = rsc;
if (parent == NULL) {
return NULL;
}
while (parent->parent != NULL) {
if (!include_bundle
&& (parent->parent->variant == pcmk_rsc_variant_bundle)) {
break;
}
parent = parent->parent;
}
return parent;
}
void
common_free(pcmk_resource_t * rsc)
{
if (rsc == NULL) {
return;
}
pcmk__rsc_trace(rsc, "Freeing %s %d", rsc->id, rsc->variant);
g_list_free(rsc->rsc_cons);
g_list_free(rsc->rsc_cons_lhs);
g_list_free(rsc->rsc_tickets);
g_list_free(rsc->dangling_migrations);
if (rsc->parameter_cache != NULL) {
g_hash_table_destroy(rsc->parameter_cache);
}
if (rsc->meta != NULL) {
g_hash_table_destroy(rsc->meta);
}
if (rsc->utilization != NULL) {
g_hash_table_destroy(rsc->utilization);
}
if ((rsc->parent == NULL)
&& pcmk_is_set(rsc->flags, pcmk_rsc_removed)) {
free_xml(rsc->xml);
rsc->xml = NULL;
free_xml(rsc->orig_xml);
rsc->orig_xml = NULL;
/* if rsc->orig_xml, then rsc->xml is an expanded xml from a template */
} else if (rsc->orig_xml) {
free_xml(rsc->xml);
rsc->xml = NULL;
}
if (rsc->running_on) {
g_list_free(rsc->running_on);
rsc->running_on = NULL;
}
if (rsc->known_on) {
g_hash_table_destroy(rsc->known_on);
rsc->known_on = NULL;
}
if (rsc->actions) {
g_list_free(rsc->actions);
rsc->actions = NULL;
}
if (rsc->allowed_nodes) {
g_hash_table_destroy(rsc->allowed_nodes);
rsc->allowed_nodes = NULL;
}
g_list_free(rsc->fillers);
g_list_free(rsc->rsc_location);
pcmk__rsc_trace(rsc, "Resource freed");
free(rsc->id);
free(rsc->clone_name);
free(rsc->allocated_to);
free(rsc->variant_opaque);
free(rsc->pending_task);
free(rsc);
}
/*!
* \internal
* \brief Count a node and update most preferred to it as appropriate
*
* \param[in] rsc An active resource
* \param[in] node A node that \p rsc is active on
* \param[in,out] active This will be set to \p node if \p node is more
* preferred than the current value
* \param[in,out] count_all If not NULL, this will be incremented
* \param[in,out] count_clean If not NULL, this will be incremented if \p node
* is online and clean
*
* \return true if the count should continue, or false if sufficiently known
*/
bool
pe__count_active_node(const pcmk_resource_t *rsc, pcmk_node_t *node,
pcmk_node_t **active, unsigned int *count_all,
unsigned int *count_clean)
{
bool keep_looking = false;
bool is_happy = false;
CRM_CHECK((rsc != NULL) && (node != NULL) && (active != NULL),
return false);
is_happy = node->details->online && !node->details->unclean;
if (count_all != NULL) {
++*count_all;
}
if ((count_clean != NULL) && is_happy) {
++*count_clean;
}
if ((count_all != NULL) || (count_clean != NULL)) {
keep_looking = true; // We're counting, so go through entire list
}
if (rsc->partial_migration_source != NULL) {
if (node->details == rsc->partial_migration_source->details) {
*active = node; // This is the migration source
} else {
keep_looking = true;
}
} else if (!pcmk_is_set(rsc->flags, pcmk_rsc_needs_fencing)) {
if (is_happy && ((*active == NULL) || !(*active)->details->online
|| (*active)->details->unclean)) {
*active = node; // This is the first clean node
} else {
keep_looking = true;
}
}
if (*active == NULL) {
*active = node; // This is the first node checked
}
return keep_looking;
}
// Shared implementation of pcmk_rsc_methods_t:active_node()
static pcmk_node_t *
active_node(const pcmk_resource_t *rsc, unsigned int *count_all,
unsigned int *count_clean)
{
pcmk_node_t *active = NULL;
if (count_all != NULL) {
*count_all = 0;
}
if (count_clean != NULL) {
*count_clean = 0;
}
if (rsc == NULL) {
return NULL;
}
for (GList *iter = rsc->running_on; iter != NULL; iter = iter->next) {
if (!pe__count_active_node(rsc, (pcmk_node_t *) iter->data, &active,
count_all, count_clean)) {
break; // Don't waste time iterating if we don't have to
}
}
return active;
}
/*!
* \brief
* \internal Find and count active nodes according to "requires"
*
* \param[in] rsc Resource to check
* \param[out] count If not NULL, will be set to count of active nodes
*
* \return An active node (or NULL if resource is not active anywhere)
*
* \note This is a convenience wrapper for active_node() where the count of all
* active nodes or only clean active nodes is desired according to the
* "requires" meta-attribute.
*/
pcmk_node_t *
pe__find_active_requires(const pcmk_resource_t *rsc, unsigned int *count)
{
if (rsc == NULL) {
if (count != NULL) {
*count = 0;
}
return NULL;
} else if (pcmk_is_set(rsc->flags, pcmk_rsc_needs_fencing)) {
return rsc->fns->active_node(rsc, count, NULL);
} else {
return rsc->fns->active_node(rsc, NULL, count);
}
}
void
pe__count_common(pcmk_resource_t *rsc)
{
if (rsc->children != NULL) {
for (GList *item = rsc->children; item != NULL; item = item->next) {
((pcmk_resource_t *) item->data)->fns->count(item->data);
}
} else if (!pcmk_is_set(rsc->flags, pcmk_rsc_removed)
|| (rsc->role > pcmk_role_stopped)) {
rsc->cluster->ninstances++;
if (pe__resource_is_disabled(rsc)) {
rsc->cluster->disabled_resources++;
}
if (pcmk_is_set(rsc->flags, pcmk_rsc_blocked)) {
rsc->cluster->blocked_resources++;
}
}
}
/*!
* \internal
* \brief Update a resource's next role
*
* \param[in,out] rsc Resource to be updated
* \param[in] role Resource's new next role
* \param[in] why Human-friendly reason why role is changing (for logs)
*/
void
pe__set_next_role(pcmk_resource_t *rsc, enum rsc_role_e role, const char *why)
{
CRM_ASSERT((rsc != NULL) && (why != NULL));
if (rsc->next_role != role) {
pcmk__rsc_trace(rsc, "Resetting next role for %s from %s to %s (%s)",
rsc->id, role2text(rsc->next_role), role2text(role),
why);
rsc->next_role = role;
}
}
diff --git a/lib/pengine/status.c b/lib/pengine/status.c
index fa2fa3f9ef..ec88d1cbe6 100644
--- a/lib/pengine/status.c
+++ b/lib/pengine/status.c
@@ -1,484 +1,484 @@
/*
* 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 <sys/param.h>
#include <crm/crm.h>
#include <crm/msg_xml.h>
#include <crm/common/xml.h>
#include <glib.h>
#include <crm/pengine/internal.h>
#include <pe_status_private.h>
/*!
* \brief Create a new object to hold scheduler data
*
* \return New, initialized scheduler data on success, else NULL (and set errno)
* \note Only pcmk_scheduler_t objects created with this function (as opposed
* to statically declared or directly allocated) should be used with the
* functions in this library, to allow for future extensions to the
* data type. The caller is responsible for freeing the memory with
* pe_free_working_set() when the instance is no longer needed.
*/
pcmk_scheduler_t *
pe_new_working_set(void)
{
pcmk_scheduler_t *scheduler = calloc(1, sizeof(pcmk_scheduler_t));
if (scheduler != NULL) {
set_working_set_defaults(scheduler);
}
return scheduler;
}
/*!
* \brief Free scheduler data
*
* \param[in,out] scheduler Scheduler data to free
*/
void
pe_free_working_set(pcmk_scheduler_t *scheduler)
{
if (scheduler != NULL) {
pe_reset_working_set(scheduler);
scheduler->priv = NULL;
free(scheduler);
}
}
/*
* Unpack everything
* At the end you'll have:
* - A list of nodes
* - A list of resources (each with any dependencies on other resources)
* - A list of constraints between resources and nodes
* - A list of constraints between start/stop actions
* - A list of nodes that need to be stonith'd
* - A list of nodes that need to be shutdown
* - A list of the possible stop/start actions (without dependencies)
*/
gboolean
cluster_status(pcmk_scheduler_t * scheduler)
{
xmlNode *section = NULL;
if ((scheduler == NULL) || (scheduler->input == NULL)) {
return FALSE;
}
crm_trace("Beginning unpack");
if (scheduler->failed != NULL) {
free_xml(scheduler->failed);
}
scheduler->failed = create_xml_node(NULL, "failed-ops");
if (scheduler->now == NULL) {
scheduler->now = crm_time_new(NULL);
}
if (scheduler->dc_uuid == NULL) {
scheduler->dc_uuid = crm_element_value_copy(scheduler->input,
XML_ATTR_DC_UUID);
}
if (pcmk__xe_attr_is_true(scheduler->input, PCMK_XA_HAVE_QUORUM)) {
- pe__set_working_set_flags(scheduler, pcmk_sched_quorate);
+ pcmk__set_scheduler_flags(scheduler, pcmk_sched_quorate);
} else {
- pe__clear_working_set_flags(scheduler, pcmk_sched_quorate);
+ pcmk__clear_scheduler_flags(scheduler, pcmk_sched_quorate);
}
scheduler->op_defaults = get_xpath_object("//" XML_CIB_TAG_OPCONFIG,
scheduler->input, LOG_NEVER);
scheduler->rsc_defaults = get_xpath_object("//" XML_CIB_TAG_RSCCONFIG,
scheduler->input, LOG_NEVER);
section = get_xpath_object("//" XML_CIB_TAG_CRMCONFIG, scheduler->input,
LOG_TRACE);
unpack_config(section, scheduler);
if (!pcmk_any_flags_set(scheduler->flags,
pcmk_sched_location_only|pcmk_sched_quorate)
&& (scheduler->no_quorum_policy != pcmk_no_quorum_ignore)) {
pcmk__sched_warn("Fencing and resource management disabled "
"due to lack of quorum");
}
section = get_xpath_object("//" XML_CIB_TAG_NODES, scheduler->input,
LOG_TRACE);
unpack_nodes(section, scheduler);
section = get_xpath_object("//" XML_CIB_TAG_RESOURCES, scheduler->input,
LOG_TRACE);
if (!pcmk_is_set(scheduler->flags, pcmk_sched_location_only)) {
unpack_remote_nodes(section, scheduler);
}
unpack_resources(section, scheduler);
section = get_xpath_object("//" XML_CIB_TAG_TAGS, scheduler->input,
LOG_NEVER);
unpack_tags(section, scheduler);
if (!pcmk_is_set(scheduler->flags, pcmk_sched_location_only)) {
section = get_xpath_object("//"XML_CIB_TAG_STATUS, scheduler->input,
LOG_TRACE);
unpack_status(section, scheduler);
}
if (!pcmk_is_set(scheduler->flags, pcmk_sched_no_counts)) {
for (GList *item = scheduler->resources; item != NULL;
item = item->next) {
((pcmk_resource_t *) (item->data))->fns->count(item->data);
}
crm_trace("Cluster resource count: %d (%d disabled, %d blocked)",
scheduler->ninstances, scheduler->disabled_resources,
scheduler->blocked_resources);
}
- pe__set_working_set_flags(scheduler, pcmk_sched_have_status);
+ pcmk__set_scheduler_flags(scheduler, pcmk_sched_have_status);
return TRUE;
}
/*!
* \internal
* \brief Free a list of pcmk_resource_t
*
* \param[in,out] resources List to free
*
* \note When the scheduler's resource list is freed, that includes the original
* storage for the uname and id of any Pacemaker Remote nodes in the
* scheduler's node list, so take care not to use those afterward.
* \todo Refactor pcmk_node_t to strdup() the node name.
*/
static void
pe_free_resources(GList *resources)
{
pcmk_resource_t *rsc = NULL;
GList *iterator = resources;
while (iterator != NULL) {
rsc = (pcmk_resource_t *) iterator->data;
iterator = iterator->next;
rsc->fns->free(rsc);
}
if (resources != NULL) {
g_list_free(resources);
}
}
static void
pe_free_actions(GList *actions)
{
GList *iterator = actions;
while (iterator != NULL) {
pe_free_action(iterator->data);
iterator = iterator->next;
}
if (actions != NULL) {
g_list_free(actions);
}
}
static void
pe_free_nodes(GList *nodes)
{
for (GList *iterator = nodes; iterator != NULL; iterator = iterator->next) {
pcmk_node_t *node = (pcmk_node_t *) iterator->data;
// Shouldn't be possible, but to be safe ...
if (node == NULL) {
continue;
}
if (node->details == NULL) {
free(node);
continue;
}
/* This is called after pe_free_resources(), which means that we can't
* use node->details->uname for Pacemaker Remote nodes.
*/
crm_trace("Freeing node %s", (pe__is_guest_or_remote_node(node)?
"(guest or remote)" : pe__node_name(node)));
if (node->details->attrs != NULL) {
g_hash_table_destroy(node->details->attrs);
}
if (node->details->utilization != NULL) {
g_hash_table_destroy(node->details->utilization);
}
if (node->details->digest_cache != NULL) {
g_hash_table_destroy(node->details->digest_cache);
}
g_list_free(node->details->running_rsc);
g_list_free(node->details->allocated_rsc);
free(node->details);
free(node);
}
if (nodes != NULL) {
g_list_free(nodes);
}
}
static void
pe__free_ordering(GList *constraints)
{
GList *iterator = constraints;
while (iterator != NULL) {
pcmk__action_relation_t *order = iterator->data;
iterator = iterator->next;
free(order->task1);
free(order->task2);
free(order);
}
if (constraints != NULL) {
g_list_free(constraints);
}
}
static void
pe__free_location(GList *constraints)
{
GList *iterator = constraints;
while (iterator != NULL) {
pcmk__location_t *cons = iterator->data;
iterator = iterator->next;
g_list_free_full(cons->nodes, free);
free(cons->id);
free(cons);
}
if (constraints != NULL) {
g_list_free(constraints);
}
}
/*!
* \brief Reset scheduler data to defaults without freeing it or constraints
*
* \param[in,out] scheduler Scheduler data to reset
*
* \deprecated This function is deprecated as part of the API;
* pe_reset_working_set() should be used instead.
*/
void
cleanup_calculations(pcmk_scheduler_t *scheduler)
{
if (scheduler == NULL) {
return;
}
- pe__clear_working_set_flags(scheduler, pcmk_sched_have_status);
+ pcmk__clear_scheduler_flags(scheduler, pcmk_sched_have_status);
if (scheduler->config_hash != NULL) {
g_hash_table_destroy(scheduler->config_hash);
}
if (scheduler->singletons != NULL) {
g_hash_table_destroy(scheduler->singletons);
}
if (scheduler->tickets) {
g_hash_table_destroy(scheduler->tickets);
}
if (scheduler->template_rsc_sets) {
g_hash_table_destroy(scheduler->template_rsc_sets);
}
if (scheduler->tags) {
g_hash_table_destroy(scheduler->tags);
}
free(scheduler->dc_uuid);
crm_trace("deleting resources");
pe_free_resources(scheduler->resources);
crm_trace("deleting actions");
pe_free_actions(scheduler->actions);
crm_trace("deleting nodes");
pe_free_nodes(scheduler->nodes);
pe__free_param_checks(scheduler);
g_list_free(scheduler->stop_needed);
free_xml(scheduler->graph);
crm_time_free(scheduler->now);
free_xml(scheduler->input);
free_xml(scheduler->failed);
set_working_set_defaults(scheduler);
CRM_CHECK(scheduler->ordering_constraints == NULL,;
);
CRM_CHECK(scheduler->placement_constraints == NULL,;
);
}
/*!
* \brief Reset scheduler data to default state without freeing it
*
* \param[in,out] scheduler Scheduler data to reset
*/
void
pe_reset_working_set(pcmk_scheduler_t *scheduler)
{
if (scheduler == NULL) {
return;
}
crm_trace("Deleting %d ordering constraints",
g_list_length(scheduler->ordering_constraints));
pe__free_ordering(scheduler->ordering_constraints);
scheduler->ordering_constraints = NULL;
crm_trace("Deleting %d location constraints",
g_list_length(scheduler->placement_constraints));
pe__free_location(scheduler->placement_constraints);
scheduler->placement_constraints = NULL;
crm_trace("Deleting %d colocation constraints",
g_list_length(scheduler->colocation_constraints));
g_list_free_full(scheduler->colocation_constraints, free);
scheduler->colocation_constraints = NULL;
crm_trace("Deleting %d ticket constraints",
g_list_length(scheduler->ticket_constraints));
g_list_free_full(scheduler->ticket_constraints, free);
scheduler->ticket_constraints = NULL;
cleanup_calculations(scheduler);
}
void
set_working_set_defaults(pcmk_scheduler_t *scheduler)
{
void *priv = scheduler->priv;
memset(scheduler, 0, sizeof(pcmk_scheduler_t));
scheduler->priv = priv;
scheduler->order_id = 1;
scheduler->action_id = 1;
scheduler->no_quorum_policy = pcmk_no_quorum_stop;
scheduler->flags = 0x0ULL;
- pe__set_working_set_flags(scheduler,
+ pcmk__set_scheduler_flags(scheduler,
pcmk_sched_symmetric_cluster
|pcmk_sched_stop_removed_resources
|pcmk_sched_cancel_removed_actions);
if (!strcmp(PCMK__CONCURRENT_FENCING_DEFAULT, "true")) {
- pe__set_working_set_flags(scheduler, pcmk_sched_concurrent_fencing);
+ pcmk__set_scheduler_flags(scheduler, pcmk_sched_concurrent_fencing);
}
}
pcmk_resource_t *
pe_find_resource(GList *rsc_list, const char *id)
{
return pe_find_resource_with_flags(rsc_list, id, pcmk_rsc_match_history);
}
pcmk_resource_t *
pe_find_resource_with_flags(GList *rsc_list, const char *id, enum pe_find flags)
{
GList *rIter = NULL;
for (rIter = rsc_list; id && rIter; rIter = rIter->next) {
pcmk_resource_t *parent = rIter->data;
pcmk_resource_t *match =
parent->fns->find_rsc(parent, id, NULL, flags);
if (match != NULL) {
return match;
}
}
crm_trace("No match for %s", id);
return NULL;
}
/*!
* \brief Find a node by name or ID in a list of nodes
*
* \param[in] nodes List of nodes (as pcmk_node_t*)
* \param[in] id If not NULL, ID of node to find
* \param[in] node_name If not NULL, name of node to find
*
* \return Node from \p nodes that matches \p id if any,
* otherwise node from \p nodes that matches \p uname if any,
* otherwise NULL
*/
pcmk_node_t *
pe_find_node_any(const GList *nodes, const char *id, const char *uname)
{
pcmk_node_t *match = NULL;
if (id != NULL) {
match = pe_find_node_id(nodes, id);
}
if ((match == NULL) && (uname != NULL)) {
match = pe_find_node(nodes, uname);
}
return match;
}
/*!
* \brief Find a node by ID in a list of nodes
*
* \param[in] nodes List of nodes (as pcmk_node_t*)
* \param[in] id ID of node to find
*
* \return Node from \p nodes that matches \p id if any, otherwise NULL
*/
pcmk_node_t *
pe_find_node_id(const GList *nodes, const char *id)
{
for (const GList *iter = nodes; iter != NULL; iter = iter->next) {
pcmk_node_t *node = (pcmk_node_t *) iter->data;
/* @TODO Whether node IDs should be considered case-sensitive should
* probably depend on the node type, so functionizing the comparison
* would be worthwhile
*/
if (pcmk__str_eq(node->details->id, id, pcmk__str_casei)) {
return node;
}
}
return NULL;
}
/*!
* \brief Find a node by name in a list of nodes
*
* \param[in] nodes List of nodes (as pcmk_node_t*)
* \param[in] node_name Name of node to find
*
* \return Node from \p nodes that matches \p node_name if any, otherwise NULL
*/
pcmk_node_t *
pe_find_node(const GList *nodes, const char *node_name)
{
for (const GList *iter = nodes; iter != NULL; iter = iter->next) {
pcmk_node_t *node = (pcmk_node_t *) iter->data;
if (pcmk__str_eq(node->details->uname, node_name, pcmk__str_casei)) {
return node;
}
}
return NULL;
}
diff --git a/lib/pengine/tests/native/native_find_rsc_test.c b/lib/pengine/tests/native/native_find_rsc_test.c
index 996666777d..1de10bf971 100644
--- a/lib/pengine/tests/native/native_find_rsc_test.c
+++ b/lib/pengine/tests/native/native_find_rsc_test.c
@@ -1,917 +1,917 @@
/*
* Copyright 2022-2023 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include <crm_internal.h>
#include <crm/common/unittest_internal.h>
#include <crm/common/scheduler.h>
#include <crm/common/xml.h>
#include <crm/pengine/internal.h>
#include <crm/pengine/status.h>
xmlNode *input = NULL;
pcmk_scheduler_t *scheduler = NULL;
pcmk_node_t *cluster01, *cluster02, *httpd_bundle_0;
pcmk_resource_t *exim_group, *inactive_group;
pcmk_resource_t *promotable_clone, *inactive_clone;
pcmk_resource_t *httpd_bundle, *mysql_clone_group;
static int
setup(void **state) {
char *path = NULL;
crm_xml_init();
path = crm_strdup_printf("%s/crm_mon.xml", getenv("PCMK_CTS_CLI_DIR"));
input = filename2xml(path);
free(path);
if (input == NULL) {
return 1;
}
scheduler = pe_new_working_set();
if (scheduler == NULL) {
return 1;
}
- pe__set_working_set_flags(scheduler,
+ pcmk__set_scheduler_flags(scheduler,
pcmk_sched_no_counts|pcmk_sched_no_compat);
scheduler->input = input;
cluster_status(scheduler);
/* Get references to the cluster nodes so we don't have to find them repeatedly. */
cluster01 = pe_find_node(scheduler->nodes, "cluster01");
cluster02 = pe_find_node(scheduler->nodes, "cluster02");
httpd_bundle_0 = pe_find_node(scheduler->nodes, "httpd-bundle-0");
/* Get references to several resources we use frequently. */
for (GList *iter = scheduler->resources; iter != NULL; iter = iter->next) {
pcmk_resource_t *rsc = (pcmk_resource_t *) iter->data;
if (strcmp(rsc->id, "exim-group") == 0) {
exim_group = rsc;
} else if (strcmp(rsc->id, "httpd-bundle") == 0) {
httpd_bundle = rsc;
} else if (strcmp(rsc->id, "inactive-clone") == 0) {
inactive_clone = rsc;
} else if (strcmp(rsc->id, "inactive-group") == 0) {
inactive_group = rsc;
} else if (strcmp(rsc->id, "mysql-clone-group") == 0) {
mysql_clone_group = rsc;
} else if (strcmp(rsc->id, "promotable-clone") == 0) {
promotable_clone = rsc;
}
}
return 0;
}
static int
teardown(void **state) {
pe_free_working_set(scheduler);
return 0;
}
static void
bad_args(void **state) {
pcmk_resource_t *rsc = g_list_first(scheduler->resources)->data;
char *id = rsc->id;
char *name = NULL;
assert_non_null(rsc);
assert_null(native_find_rsc(NULL, "dummy", NULL, 0));
assert_null(native_find_rsc(rsc, NULL, NULL, 0));
/* No resources exist with these names. */
name = crm_strdup_printf("%sX", rsc->id);
assert_null(native_find_rsc(rsc, name, NULL, 0));
free(name);
name = crm_strdup_printf("x%s", rsc->id);
assert_null(native_find_rsc(rsc, name, NULL, 0));
free(name);
name = g_ascii_strup(rsc->id, -1);
assert_null(native_find_rsc(rsc, name, NULL, 0));
g_free(name);
/* Fails because resource ID is NULL. */
rsc->id = NULL;
assert_null(native_find_rsc(rsc, id, NULL, 0));
rsc->id = id;
}
static void
primitive_rsc(void **state) {
pcmk_resource_t *dummy = NULL;
/* Find the "dummy" resource, which is the only one with that ID in the set. */
for (GList *iter = scheduler->resources; iter != NULL; iter = iter->next) {
pcmk_resource_t *rsc = (pcmk_resource_t *) iter->data;
if (strcmp(rsc->id, "dummy") == 0) {
dummy = rsc;
break;
}
}
assert_non_null(dummy);
/* Passes because NULL was passed for node, regardless of flags. */
assert_ptr_equal(dummy, native_find_rsc(dummy, "dummy", NULL, 0));
assert_ptr_equal(dummy,
native_find_rsc(dummy, "dummy", NULL,
pcmk_rsc_match_current_node));
/* Fails because resource is not a clone (nor cloned). */
assert_null(native_find_rsc(dummy, "dummy", NULL,
pcmk_rsc_match_clone_only));
assert_null(native_find_rsc(dummy, "dummy", cluster02,
pcmk_rsc_match_clone_only));
/* Fails because dummy is not running on cluster01, even with the right flags. */
assert_null(native_find_rsc(dummy, "dummy", cluster01,
pcmk_rsc_match_current_node));
// Fails because pcmk_rsc_match_current_node is required if a node is given
assert_null(native_find_rsc(dummy, "dummy", cluster02, 0));
/* Passes because dummy is running on cluster02. */
assert_ptr_equal(dummy,
native_find_rsc(dummy, "dummy", cluster02,
pcmk_rsc_match_current_node));
}
static void
group_rsc(void **state) {
assert_non_null(exim_group);
/* Passes because NULL was passed for node, regardless of flags. */
assert_ptr_equal(exim_group, native_find_rsc(exim_group, "exim-group", NULL, 0));
assert_ptr_equal(exim_group,
native_find_rsc(exim_group, "exim-group", NULL,
pcmk_rsc_match_current_node));
/* Fails because resource is not a clone (nor cloned). */
assert_null(native_find_rsc(exim_group, "exim-group", NULL,
pcmk_rsc_match_clone_only));
assert_null(native_find_rsc(exim_group, "exim-group", cluster01,
pcmk_rsc_match_clone_only));
/* Fails because none of exim-group's children are running on cluster01, even with the right flags. */
assert_null(native_find_rsc(exim_group, "exim-group", cluster01,
pcmk_rsc_match_current_node));
// Fails because pcmk_rsc_match_current_node is required if a node is given
assert_null(native_find_rsc(exim_group, "exim-group", cluster01, 0));
/* Passes because one of exim-group's children is running on cluster02. */
assert_ptr_equal(exim_group,
native_find_rsc(exim_group, "exim-group", cluster02,
pcmk_rsc_match_current_node));
}
static void
inactive_group_rsc(void **state) {
assert_non_null(inactive_group);
/* Passes because NULL was passed for node, regardless of flags. */
assert_ptr_equal(inactive_group, native_find_rsc(inactive_group, "inactive-group", NULL, 0));
assert_ptr_equal(inactive_group,
native_find_rsc(inactive_group, "inactive-group", NULL,
pcmk_rsc_match_current_node));
/* Fails because resource is not a clone (nor cloned). */
assert_null(native_find_rsc(inactive_group, "inactive-group", NULL,
pcmk_rsc_match_clone_only));
assert_null(native_find_rsc(inactive_group, "inactive-group", cluster01,
pcmk_rsc_match_clone_only));
/* Fails because none of inactive-group's children are running. */
assert_null(native_find_rsc(inactive_group, "inactive-group", cluster01,
pcmk_rsc_match_current_node));
assert_null(native_find_rsc(inactive_group, "inactive-group", cluster02,
pcmk_rsc_match_current_node));
}
static void
group_member_rsc(void **state) {
pcmk_resource_t *public_ip = NULL;
/* Find the "Public-IP" resource, a member of "exim-group". */
for (GList *iter = exim_group->children; iter != NULL; iter = iter->next) {
pcmk_resource_t *rsc = (pcmk_resource_t *) iter->data;
if (strcmp(rsc->id, "Public-IP") == 0) {
public_ip = rsc;
break;
}
}
assert_non_null(public_ip);
/* Passes because NULL was passed for node, regardless of flags. */
assert_ptr_equal(public_ip, native_find_rsc(public_ip, "Public-IP", NULL, 0));
assert_ptr_equal(public_ip,
native_find_rsc(public_ip, "Public-IP", NULL,
pcmk_rsc_match_current_node));
/* Fails because resource is not a clone (nor cloned). */
assert_null(native_find_rsc(public_ip, "Public-IP", NULL,
pcmk_rsc_match_clone_only));
assert_null(native_find_rsc(public_ip, "Public-IP", cluster02,
pcmk_rsc_match_clone_only));
/* Fails because Public-IP is not running on cluster01, even with the right flags. */
assert_null(native_find_rsc(public_ip, "Public-IP", cluster01,
pcmk_rsc_match_current_node));
// Fails because pcmk_rsc_match_current_node is required if a node is given
assert_null(native_find_rsc(public_ip, "Public-IP", cluster02, 0));
/* Passes because Public-IP is running on cluster02. */
assert_ptr_equal(public_ip,
native_find_rsc(public_ip, "Public-IP", cluster02,
pcmk_rsc_match_current_node));
}
static void
inactive_group_member_rsc(void **state) {
pcmk_resource_t *inactive_dummy_1 = NULL;
/* Find the "inactive-dummy-1" resource, a member of "inactive-group". */
for (GList *iter = inactive_group->children; iter != NULL; iter = iter->next) {
pcmk_resource_t *rsc = (pcmk_resource_t *) iter->data;
if (strcmp(rsc->id, "inactive-dummy-1") == 0) {
inactive_dummy_1 = rsc;
break;
}
}
assert_non_null(inactive_dummy_1);
/* Passes because NULL was passed for node, regardless of flags. */
assert_ptr_equal(inactive_dummy_1, native_find_rsc(inactive_dummy_1, "inactive-dummy-1", NULL, 0));
assert_ptr_equal(inactive_dummy_1,
native_find_rsc(inactive_dummy_1, "inactive-dummy-1", NULL,
pcmk_rsc_match_current_node));
/* Fails because resource is not a clone (nor cloned). */
assert_null(native_find_rsc(inactive_dummy_1, "inactive-dummy-1", NULL,
pcmk_rsc_match_clone_only));
assert_null(native_find_rsc(inactive_dummy_1, "inactive-dummy-1", cluster01,
pcmk_rsc_match_clone_only));
/* Fails because inactive-dummy-1 is not running. */
assert_null(native_find_rsc(inactive_dummy_1, "inactive-dummy-1", cluster01,
pcmk_rsc_match_current_node));
assert_null(native_find_rsc(inactive_dummy_1, "inactive-dummy-1", cluster02,
pcmk_rsc_match_current_node));
}
static void
clone_rsc(void **state) {
assert_non_null(promotable_clone);
/* Passes because NULL was passed for node, regardless of flags. */
assert_ptr_equal(promotable_clone, native_find_rsc(promotable_clone, "promotable-clone", NULL, 0));
assert_ptr_equal(promotable_clone,
native_find_rsc(promotable_clone, "promotable-clone", NULL,
pcmk_rsc_match_current_node));
assert_ptr_equal(promotable_clone,
native_find_rsc(promotable_clone, "promotable-clone", NULL,
pcmk_rsc_match_clone_only));
// Fails because pcmk_rsc_match_current_node is required if a node is given
assert_null(native_find_rsc(promotable_clone, "promotable-clone", cluster01, 0));
/* Passes because one of ping-clone's children is running on cluster01. */
assert_ptr_equal(promotable_clone,
native_find_rsc(promotable_clone, "promotable-clone",
cluster01, pcmk_rsc_match_current_node));
// Fails because pcmk_rsc_match_current_node is required if a node is given
assert_null(native_find_rsc(promotable_clone, "promotable-clone", cluster02, 0));
/* Passes because one of ping_clone's children is running on cluster02. */
assert_ptr_equal(promotable_clone,
native_find_rsc(promotable_clone, "promotable-clone",
cluster02, pcmk_rsc_match_current_node));
// Passes for previous reasons, plus includes pcmk_rsc_match_clone_only
assert_ptr_equal(promotable_clone,
native_find_rsc(promotable_clone, "promotable-clone",
cluster01,
pcmk_rsc_match_clone_only
|pcmk_rsc_match_current_node));
assert_ptr_equal(promotable_clone,
native_find_rsc(promotable_clone, "promotable-clone",
cluster02,
pcmk_rsc_match_clone_only
|pcmk_rsc_match_current_node));
}
static void
inactive_clone_rsc(void **state) {
assert_non_null(inactive_clone);
/* Passes because NULL was passed for node, regardless of flags. */
assert_ptr_equal(inactive_clone, native_find_rsc(inactive_clone, "inactive-clone", NULL, 0));
assert_ptr_equal(inactive_clone,
native_find_rsc(inactive_clone, "inactive-clone", NULL,
pcmk_rsc_match_current_node));
assert_ptr_equal(inactive_clone,
native_find_rsc(inactive_clone, "inactive-clone", NULL,
pcmk_rsc_match_clone_only));
/* Fails because none of inactive-clone's children are running. */
assert_null(native_find_rsc(inactive_clone, "inactive-clone", cluster01,
pcmk_rsc_match_current_node
|pcmk_rsc_match_clone_only));
assert_null(native_find_rsc(inactive_clone, "inactive-clone", cluster02,
pcmk_rsc_match_current_node
|pcmk_rsc_match_clone_only));
}
static void
clone_instance_rsc(void **state) {
pcmk_resource_t *promotable_0 = NULL;
pcmk_resource_t *promotable_1 = NULL;
/* Find the "promotable-rsc:0" and "promotable-rsc:1" resources, members of "promotable-clone". */
for (GList *iter = promotable_clone->children; iter != NULL; iter = iter->next) {
pcmk_resource_t *rsc = (pcmk_resource_t *) iter->data;
if (strcmp(rsc->id, "promotable-rsc:0") == 0) {
promotable_0 = rsc;
} else if (strcmp(rsc->id, "promotable-rsc:1") == 0) {
promotable_1 = rsc;
}
}
assert_non_null(promotable_0);
assert_non_null(promotable_1);
/* Passes because NULL was passed for node, regardless of flags. */
assert_ptr_equal(promotable_0, native_find_rsc(promotable_0, "promotable-rsc:0", NULL, 0));
assert_ptr_equal(promotable_0,
native_find_rsc(promotable_0, "promotable-rsc:0", NULL,
pcmk_rsc_match_current_node));
assert_ptr_equal(promotable_1, native_find_rsc(promotable_1, "promotable-rsc:1", NULL, 0));
assert_ptr_equal(promotable_1,
native_find_rsc(promotable_1, "promotable-rsc:1", NULL,
pcmk_rsc_match_current_node));
// Fails because pcmk_rsc_match_current_node is required if a node is given
assert_null(native_find_rsc(promotable_0, "promotable-rsc:0", cluster02, 0));
assert_null(native_find_rsc(promotable_1, "promotable-rsc:1", cluster01, 0));
/* Check that the resource is running on the node we expect. */
assert_ptr_equal(promotable_0,
native_find_rsc(promotable_0, "promotable-rsc:0",
cluster02, pcmk_rsc_match_current_node));
assert_null(native_find_rsc(promotable_0, "promotable-rsc:0", cluster01,
pcmk_rsc_match_current_node));
assert_ptr_equal(promotable_1,
native_find_rsc(promotable_1, "promotable-rsc:1",
cluster01, pcmk_rsc_match_current_node));
assert_null(native_find_rsc(promotable_1, "promotable-rsc:1", cluster02,
pcmk_rsc_match_current_node));
/* Passes because NULL was passed for node and primitive name was given, with correct flags. */
assert_ptr_equal(promotable_0,
native_find_rsc(promotable_0, "promotable-rsc", NULL,
pcmk_rsc_match_clone_only));
// Passes because pcmk_rsc_match_basename matches any instance's base name
assert_ptr_equal(promotable_0,
native_find_rsc(promotable_0, "promotable-rsc", NULL,
pcmk_rsc_match_basename));
assert_ptr_equal(promotable_1,
native_find_rsc(promotable_1, "promotable-rsc", NULL,
pcmk_rsc_match_basename));
// Passes because pcmk_rsc_match_anon_basename matches
assert_ptr_equal(promotable_0,
native_find_rsc(promotable_0, "promotable-rsc", NULL,
pcmk_rsc_match_anon_basename));
assert_ptr_equal(promotable_1,
native_find_rsc(promotable_1, "promotable-rsc", NULL,
pcmk_rsc_match_anon_basename));
/* Check that the resource is running on the node we expect. */
assert_ptr_equal(promotable_0,
native_find_rsc(promotable_0, "promotable-rsc", cluster02,
pcmk_rsc_match_basename
|pcmk_rsc_match_current_node));
assert_ptr_equal(promotable_0,
native_find_rsc(promotable_0, "promotable-rsc", cluster02,
pcmk_rsc_match_anon_basename
|pcmk_rsc_match_current_node));
assert_null(native_find_rsc(promotable_0, "promotable-rsc", cluster01,
pcmk_rsc_match_basename
|pcmk_rsc_match_current_node));
assert_null(native_find_rsc(promotable_0, "promotable-rsc", cluster01,
pcmk_rsc_match_anon_basename
|pcmk_rsc_match_current_node));
assert_ptr_equal(promotable_1,
native_find_rsc(promotable_1, "promotable-rsc", cluster01,
pcmk_rsc_match_basename
|pcmk_rsc_match_current_node));
assert_ptr_equal(promotable_1,
native_find_rsc(promotable_1, "promotable-rsc", cluster01,
pcmk_rsc_match_anon_basename
|pcmk_rsc_match_current_node));
assert_null(native_find_rsc(promotable_1, "promotable-rsc", cluster02,
pcmk_rsc_match_basename
|pcmk_rsc_match_current_node));
assert_null(native_find_rsc(promotable_1, "promotable-rsc", cluster02,
pcmk_rsc_match_anon_basename
|pcmk_rsc_match_current_node));
/* Fails because incorrect flags were given along with primitive name. */
assert_null(native_find_rsc(promotable_0, "promotable-rsc", NULL,
pcmk_rsc_match_current_node));
assert_null(native_find_rsc(promotable_1, "promotable-rsc", NULL,
pcmk_rsc_match_current_node));
/* And then we check failure possibilities again, except passing promotable_clone
* instead of promotable_X as the first argument to native_find_rsc.
*/
// Fails because pcmk_rsc_match_current_node is required if a node is given
assert_null(native_find_rsc(promotable_clone, "promotable-rsc:0", cluster02, 0));
assert_null(native_find_rsc(promotable_clone, "promotable-rsc:1", cluster01, 0));
/* Check that the resource is running on the node we expect. */
assert_ptr_equal(promotable_0,
native_find_rsc(promotable_clone, "promotable-rsc:0",
cluster02, pcmk_rsc_match_current_node));
assert_ptr_equal(promotable_0,
native_find_rsc(promotable_clone, "promotable-rsc",
cluster02,
pcmk_rsc_match_basename
|pcmk_rsc_match_current_node));
assert_ptr_equal(promotable_0,
native_find_rsc(promotable_clone, "promotable-rsc",
cluster02,
pcmk_rsc_match_anon_basename
|pcmk_rsc_match_current_node));
assert_ptr_equal(promotable_1,
native_find_rsc(promotable_clone, "promotable-rsc:1",
cluster01, pcmk_rsc_match_current_node));
assert_ptr_equal(promotable_1,
native_find_rsc(promotable_clone, "promotable-rsc",
cluster01,
pcmk_rsc_match_basename
|pcmk_rsc_match_current_node));
assert_ptr_equal(promotable_1,
native_find_rsc(promotable_clone, "promotable-rsc",
cluster01,
pcmk_rsc_match_anon_basename
|pcmk_rsc_match_current_node));
}
static void
renamed_rsc(void **state) {
pcmk_resource_t *promotable_0 = NULL;
pcmk_resource_t *promotable_1 = NULL;
/* Find the "promotable-rsc:0" and "promotable-rsc:1" resources, members of "promotable-clone". */
for (GList *iter = promotable_clone->children; iter != NULL; iter = iter->next) {
pcmk_resource_t *rsc = (pcmk_resource_t *) iter->data;
if (strcmp(rsc->id, "promotable-rsc:0") == 0) {
promotable_0 = rsc;
} else if (strcmp(rsc->id, "promotable-rsc:1") == 0) {
promotable_1 = rsc;
}
}
assert_non_null(promotable_0);
assert_non_null(promotable_1);
// Passes because pcmk_rsc_match_history means base name matches clone_name
assert_ptr_equal(promotable_0,
native_find_rsc(promotable_0, "promotable-rsc", NULL,
pcmk_rsc_match_history));
assert_ptr_equal(promotable_1,
native_find_rsc(promotable_1, "promotable-rsc", NULL,
pcmk_rsc_match_history));
}
static void
bundle_rsc(void **state) {
assert_non_null(httpd_bundle);
/* Passes because NULL was passed for node, regardless of flags. */
assert_ptr_equal(httpd_bundle, native_find_rsc(httpd_bundle, "httpd-bundle", NULL, 0));
assert_ptr_equal(httpd_bundle,
native_find_rsc(httpd_bundle, "httpd-bundle", NULL,
pcmk_rsc_match_current_node));
/* Fails because resource is not a clone (nor cloned). */
assert_null(native_find_rsc(httpd_bundle, "httpd-bundle", NULL,
pcmk_rsc_match_clone_only));
assert_null(native_find_rsc(httpd_bundle, "httpd-bundle", cluster01,
pcmk_rsc_match_clone_only));
// Fails because pcmk_rsc_match_current_node is required if a node is given
assert_null(native_find_rsc(httpd_bundle, "httpd-bundle", cluster01, 0));
/* Passes because one of httpd_bundle's children is running on cluster01. */
assert_ptr_equal(httpd_bundle,
native_find_rsc(httpd_bundle, "httpd-bundle", cluster01,
pcmk_rsc_match_current_node));
}
static bool
bundle_first_replica(pcmk__bundle_replica_t *replica, void *user_data)
{
pcmk_resource_t *ip_0 = replica->ip;
pcmk_resource_t *child_0 = replica->child;
pcmk_resource_t *container_0 = replica->container;
pcmk_resource_t *remote_0 = replica->remote;
assert_non_null(ip_0);
assert_non_null(child_0);
assert_non_null(container_0);
assert_non_null(remote_0);
/* Passes because NULL was passed for node, regardless of flags. */
assert_ptr_equal(ip_0, native_find_rsc(ip_0, "httpd-bundle-ip-192.168.122.131", NULL, 0));
assert_ptr_equal(child_0, native_find_rsc(child_0, "httpd:0", NULL, 0));
assert_ptr_equal(container_0, native_find_rsc(container_0, "httpd-bundle-docker-0", NULL, 0));
assert_ptr_equal(remote_0, native_find_rsc(remote_0, "httpd-bundle-0", NULL, 0));
// Fails because pcmk_rsc_match_current_node is required if a node is given
assert_null(native_find_rsc(ip_0, "httpd-bundle-ip-192.168.122.131", cluster01, 0));
assert_null(native_find_rsc(child_0, "httpd:0", httpd_bundle_0, 0));
assert_null(native_find_rsc(container_0, "httpd-bundle-docker-0", cluster01, 0));
assert_null(native_find_rsc(remote_0, "httpd-bundle-0", cluster01, 0));
/* Check that the resource is running on the node we expect. */
assert_ptr_equal(ip_0,
native_find_rsc(ip_0, "httpd-bundle-ip-192.168.122.131",
cluster01, pcmk_rsc_match_current_node));
assert_null(native_find_rsc(ip_0, "httpd-bundle-ip-192.168.122.131",
cluster02, pcmk_rsc_match_current_node));
assert_null(native_find_rsc(ip_0, "httpd-bundle-ip-192.168.122.131",
httpd_bundle_0, pcmk_rsc_match_current_node));
assert_ptr_equal(child_0,
native_find_rsc(child_0, "httpd:0", httpd_bundle_0,
pcmk_rsc_match_current_node));
assert_null(native_find_rsc(child_0, "httpd:0", cluster01,
pcmk_rsc_match_current_node));
assert_null(native_find_rsc(child_0, "httpd:0", cluster02,
pcmk_rsc_match_current_node));
assert_ptr_equal(container_0,
native_find_rsc(container_0, "httpd-bundle-docker-0",
cluster01, pcmk_rsc_match_current_node));
assert_null(native_find_rsc(container_0, "httpd-bundle-docker-0", cluster02,
pcmk_rsc_match_current_node));
assert_null(native_find_rsc(container_0, "httpd-bundle-docker-0",
httpd_bundle_0, pcmk_rsc_match_current_node));
assert_ptr_equal(remote_0,
native_find_rsc(remote_0, "httpd-bundle-0", cluster01,
pcmk_rsc_match_current_node));
assert_null(native_find_rsc(remote_0, "httpd-bundle-0", cluster02,
pcmk_rsc_match_current_node));
assert_null(native_find_rsc(remote_0, "httpd-bundle-0", httpd_bundle_0,
pcmk_rsc_match_current_node));
// Passes because pcmk_rsc_match_basename matches any replica's base name
assert_ptr_equal(child_0,
native_find_rsc(child_0, "httpd", NULL,
pcmk_rsc_match_basename));
// Passes because pcmk_rsc_match_anon_basename matches
assert_ptr_equal(child_0,
native_find_rsc(child_0, "httpd", NULL,
pcmk_rsc_match_anon_basename));
/* Check that the resource is running on the node we expect. */
assert_ptr_equal(child_0,
native_find_rsc(child_0, "httpd", httpd_bundle_0,
pcmk_rsc_match_basename
|pcmk_rsc_match_current_node));
assert_ptr_equal(child_0,
native_find_rsc(child_0, "httpd", httpd_bundle_0,
pcmk_rsc_match_anon_basename
|pcmk_rsc_match_current_node));
assert_null(native_find_rsc(child_0, "httpd", cluster01,
pcmk_rsc_match_basename
|pcmk_rsc_match_current_node));
assert_null(native_find_rsc(child_0, "httpd", cluster01,
pcmk_rsc_match_anon_basename
|pcmk_rsc_match_current_node));
assert_null(native_find_rsc(child_0, "httpd", cluster02,
pcmk_rsc_match_basename
|pcmk_rsc_match_current_node));
assert_null(native_find_rsc(child_0, "httpd", cluster02,
pcmk_rsc_match_anon_basename
|pcmk_rsc_match_current_node));
/* Fails because incorrect flags were given along with base name. */
assert_null(native_find_rsc(child_0, "httpd", NULL,
pcmk_rsc_match_current_node));
/* And then we check failure possibilities again, except passing httpd-bundle
* instead of X_0 as the first argument to native_find_rsc.
*/
// Fails because pcmk_rsc_match_current_node is required if a node is given
assert_null(native_find_rsc(httpd_bundle, "httpd-bundle-ip-192.168.122.131", cluster01, 0));
assert_null(native_find_rsc(httpd_bundle, "httpd:0", httpd_bundle_0, 0));
assert_null(native_find_rsc(httpd_bundle, "httpd-bundle-docker-0", cluster01, 0));
assert_null(native_find_rsc(httpd_bundle, "httpd-bundle-0", cluster01, 0));
/* Check that the resource is running on the node we expect. */
assert_ptr_equal(ip_0,
native_find_rsc(httpd_bundle,
"httpd-bundle-ip-192.168.122.131",
cluster01, pcmk_rsc_match_current_node));
assert_ptr_equal(child_0,
native_find_rsc(httpd_bundle, "httpd:0", httpd_bundle_0,
pcmk_rsc_match_current_node));
assert_ptr_equal(container_0,
native_find_rsc(httpd_bundle, "httpd-bundle-docker-0",
cluster01, pcmk_rsc_match_current_node));
assert_ptr_equal(remote_0,
native_find_rsc(httpd_bundle, "httpd-bundle-0", cluster01,
pcmk_rsc_match_current_node));
return false; // Do not iterate through any further replicas
}
static void
bundle_replica_rsc(void **state)
{
pe__foreach_bundle_replica(httpd_bundle, bundle_first_replica, NULL);
}
static void
clone_group_rsc(void **rsc) {
assert_non_null(mysql_clone_group);
/* Passes because NULL was passed for node, regardless of flags. */
assert_ptr_equal(mysql_clone_group, native_find_rsc(mysql_clone_group, "mysql-clone-group", NULL, 0));
assert_ptr_equal(mysql_clone_group,
native_find_rsc(mysql_clone_group, "mysql-clone-group",
NULL, pcmk_rsc_match_current_node));
assert_ptr_equal(mysql_clone_group,
native_find_rsc(mysql_clone_group, "mysql-clone-group",
NULL, pcmk_rsc_match_clone_only));
// Fails because pcmk_rsc_match_current_node is required if a node is given
assert_null(native_find_rsc(mysql_clone_group, "mysql-clone-group", cluster01, 0));
/* Passes because one of mysql-clone-group's children is running on cluster01. */
assert_ptr_equal(mysql_clone_group,
native_find_rsc(mysql_clone_group, "mysql-clone-group",
cluster01, pcmk_rsc_match_current_node));
// Fails because pcmk_rsc_match_current_node is required if a node is given
assert_null(native_find_rsc(mysql_clone_group, "mysql-clone-group", cluster02, 0));
/* Passes because one of mysql-clone-group's children is running on cluster02. */
assert_ptr_equal(mysql_clone_group,
native_find_rsc(mysql_clone_group, "mysql-clone-group",
cluster02, pcmk_rsc_match_current_node));
// Passes for previous reasons, plus includes pcmk_rsc_match_clone_only
assert_ptr_equal(mysql_clone_group,
native_find_rsc(mysql_clone_group, "mysql-clone-group",
cluster01,
pcmk_rsc_match_clone_only
|pcmk_rsc_match_current_node));
assert_ptr_equal(mysql_clone_group,
native_find_rsc(mysql_clone_group, "mysql-clone-group",
cluster02,
pcmk_rsc_match_clone_only
|pcmk_rsc_match_current_node));
}
static void
clone_group_instance_rsc(void **rsc) {
pcmk_resource_t *mysql_group_0 = NULL;
pcmk_resource_t *mysql_group_1 = NULL;
/* Find the "mysql-group:0" and "mysql-group:1" resources, members of "mysql-clone-group". */
for (GList *iter = mysql_clone_group->children; iter != NULL; iter = iter->next) {
pcmk_resource_t *rsc = (pcmk_resource_t *) iter->data;
if (strcmp(rsc->id, "mysql-group:0") == 0) {
mysql_group_0 = rsc;
} else if (strcmp(rsc->id, "mysql-group:1") == 0) {
mysql_group_1 = rsc;
}
}
assert_non_null(mysql_group_0);
assert_non_null(mysql_group_1);
/* Passes because NULL was passed for node, regardless of flags. */
assert_ptr_equal(mysql_group_0, native_find_rsc(mysql_group_0, "mysql-group:0", NULL, 0));
assert_ptr_equal(mysql_group_0,
native_find_rsc(mysql_group_0, "mysql-group:0", NULL,
pcmk_rsc_match_current_node));
assert_ptr_equal(mysql_group_1, native_find_rsc(mysql_group_1, "mysql-group:1", NULL, 0));
assert_ptr_equal(mysql_group_1,
native_find_rsc(mysql_group_1, "mysql-group:1", NULL,
pcmk_rsc_match_current_node));
// Fails because pcmk_rsc_match_current_node is required if a node is given
assert_null(native_find_rsc(mysql_group_0, "mysql-group:0", cluster02, 0));
assert_null(native_find_rsc(mysql_group_1, "mysql-group:1", cluster01, 0));
/* Check that the resource is running on the node we expect. */
assert_ptr_equal(mysql_group_0,
native_find_rsc(mysql_group_0, "mysql-group:0", cluster02,
pcmk_rsc_match_current_node));
assert_null(native_find_rsc(mysql_group_0, "mysql-group:0", cluster01,
pcmk_rsc_match_current_node));
assert_ptr_equal(mysql_group_1,
native_find_rsc(mysql_group_1, "mysql-group:1", cluster01,
pcmk_rsc_match_current_node));
assert_null(native_find_rsc(mysql_group_1, "mysql-group:1", cluster02,
pcmk_rsc_match_current_node));
/* Passes because NULL was passed for node and base name was given, with correct flags. */
assert_ptr_equal(mysql_group_0,
native_find_rsc(mysql_group_0, "mysql-group" , NULL,
pcmk_rsc_match_clone_only));
// Passes because pcmk_rsc_match_basename matches any base name
assert_ptr_equal(mysql_group_0,
native_find_rsc(mysql_group_0, "mysql-group" , NULL,
pcmk_rsc_match_basename));
assert_ptr_equal(mysql_group_1,
native_find_rsc(mysql_group_1, "mysql-group" , NULL,
pcmk_rsc_match_basename));
// Passes because pcmk_rsc_match_anon_basename matches
assert_ptr_equal(mysql_group_0,
native_find_rsc(mysql_group_0, "mysql-group" , NULL,
pcmk_rsc_match_anon_basename));
assert_ptr_equal(mysql_group_1,
native_find_rsc(mysql_group_1, "mysql-group" , NULL,
pcmk_rsc_match_anon_basename));
/* Check that the resource is running on the node we expect. */
assert_ptr_equal(mysql_group_0,
native_find_rsc(mysql_group_0, "mysql-group", cluster02,
pcmk_rsc_match_basename
|pcmk_rsc_match_current_node));
assert_ptr_equal(mysql_group_0,
native_find_rsc(mysql_group_0, "mysql-group", cluster02,
pcmk_rsc_match_anon_basename
|pcmk_rsc_match_current_node));
assert_null(native_find_rsc(mysql_group_0, "mysql-group", cluster01,
pcmk_rsc_match_basename
|pcmk_rsc_match_current_node));
assert_null(native_find_rsc(mysql_group_0, "mysql-group", cluster01,
pcmk_rsc_match_anon_basename
|pcmk_rsc_match_current_node));
assert_ptr_equal(mysql_group_1,
native_find_rsc(mysql_group_1, "mysql-group", cluster01,
pcmk_rsc_match_basename
|pcmk_rsc_match_current_node));
assert_ptr_equal(mysql_group_1,
native_find_rsc(mysql_group_1, "mysql-group", cluster01,
pcmk_rsc_match_anon_basename
|pcmk_rsc_match_current_node));
assert_null(native_find_rsc(mysql_group_1, "mysql-group", cluster02,
pcmk_rsc_match_basename
|pcmk_rsc_match_current_node));
assert_null(native_find_rsc(mysql_group_1, "mysql-group", cluster02,
pcmk_rsc_match_anon_basename
|pcmk_rsc_match_current_node));
/* Fails because incorrect flags were given along with base name. */
assert_null(native_find_rsc(mysql_group_0, "mysql-group", NULL,
pcmk_rsc_match_current_node));
assert_null(native_find_rsc(mysql_group_1, "mysql-group", NULL,
pcmk_rsc_match_current_node));
/* And then we check failure possibilities again, except passing mysql_clone_group
* instead of mysql_group_X as the first argument to native_find_rsc.
*/
// Fails because pcmk_rsc_match_current_node is required if a node is given
assert_null(native_find_rsc(mysql_clone_group, "mysql-group:0", cluster02, 0));
assert_null(native_find_rsc(mysql_clone_group, "mysql-group:1", cluster01, 0));
/* Check that the resource is running on the node we expect. */
assert_ptr_equal(mysql_group_0,
native_find_rsc(mysql_clone_group, "mysql-group:0",
cluster02, pcmk_rsc_match_current_node));
assert_ptr_equal(mysql_group_0,
native_find_rsc(mysql_clone_group, "mysql-group",
cluster02,
pcmk_rsc_match_basename
|pcmk_rsc_match_current_node));
assert_ptr_equal(mysql_group_0,
native_find_rsc(mysql_clone_group, "mysql-group",
cluster02,
pcmk_rsc_match_anon_basename
|pcmk_rsc_match_current_node));
assert_ptr_equal(mysql_group_1,
native_find_rsc(mysql_clone_group, "mysql-group:1",
cluster01, pcmk_rsc_match_current_node));
assert_ptr_equal(mysql_group_1,
native_find_rsc(mysql_clone_group, "mysql-group",
cluster01,
pcmk_rsc_match_basename
|pcmk_rsc_match_current_node));
assert_ptr_equal(mysql_group_1,
native_find_rsc(mysql_clone_group, "mysql-group",
cluster01,
pcmk_rsc_match_anon_basename
|pcmk_rsc_match_current_node));
}
static void
clone_group_member_rsc(void **state) {
pcmk_resource_t *mysql_proxy = NULL;
/* Find the "mysql-proxy" resource, a member of "mysql-group". */
for (GList *iter = mysql_clone_group->children; iter != NULL; iter = iter->next) {
pcmk_resource_t *rsc = (pcmk_resource_t *) iter->data;
if (strcmp(rsc->id, "mysql-group:0") == 0) {
for (GList *iter2 = rsc->children; iter2 != NULL; iter2 = iter2->next) {
pcmk_resource_t *child = (pcmk_resource_t *) iter2->data;
if (strcmp(child->id, "mysql-proxy:0") == 0) {
mysql_proxy = child;
break;
}
}
break;
}
}
assert_non_null(mysql_proxy);
/* Passes because NULL was passed for node, regardless of flags. */
assert_ptr_equal(mysql_proxy, native_find_rsc(mysql_proxy, "mysql-proxy:0", NULL, 0));
assert_ptr_equal(mysql_proxy,
native_find_rsc(mysql_proxy, "mysql-proxy:0", NULL,
pcmk_rsc_match_current_node));
/* Passes because resource's parent is a clone. */
assert_ptr_equal(mysql_proxy,
native_find_rsc(mysql_proxy, "mysql-proxy:0", NULL,
pcmk_rsc_match_clone_only));
assert_ptr_equal(mysql_proxy,
native_find_rsc(mysql_proxy, "mysql-proxy:0", cluster02,
pcmk_rsc_match_clone_only
|pcmk_rsc_match_current_node));
/* Fails because mysql-proxy:0 is not running on cluster01, even with the right flags. */
assert_null(native_find_rsc(mysql_proxy, "mysql-proxy:0", cluster01,
pcmk_rsc_match_current_node));
// Fails because pcmk_rsc_match_current_node is required if a node is given
assert_null(native_find_rsc(mysql_proxy, "mysql-proxy:0", cluster02, 0));
/* Passes because mysql-proxy:0 is running on cluster02. */
assert_ptr_equal(mysql_proxy,
native_find_rsc(mysql_proxy, "mysql-proxy:0", cluster02,
pcmk_rsc_match_current_node));
}
/* TODO: Add tests for finding on assigned node (passing a node without
* pcmk_rsc_match_current_node, after scheduling, for a resource that is
* starting/stopping/moving.
*/
PCMK__UNIT_TEST(setup, teardown,
cmocka_unit_test(bad_args),
cmocka_unit_test(primitive_rsc),
cmocka_unit_test(group_rsc),
cmocka_unit_test(inactive_group_rsc),
cmocka_unit_test(group_member_rsc),
cmocka_unit_test(inactive_group_member_rsc),
cmocka_unit_test(clone_rsc),
cmocka_unit_test(inactive_clone_rsc),
cmocka_unit_test(clone_instance_rsc),
cmocka_unit_test(renamed_rsc),
cmocka_unit_test(bundle_rsc),
cmocka_unit_test(bundle_replica_rsc),
cmocka_unit_test(clone_group_rsc),
cmocka_unit_test(clone_group_instance_rsc),
cmocka_unit_test(clone_group_member_rsc))
diff --git a/lib/pengine/tests/native/pe_base_name_eq_test.c b/lib/pengine/tests/native/pe_base_name_eq_test.c
index cb3c908ed9..f6081ba445 100644
--- a/lib/pengine/tests/native/pe_base_name_eq_test.c
+++ b/lib/pengine/tests/native/pe_base_name_eq_test.c
@@ -1,150 +1,150 @@
/*
* Copyright 2022-2023 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include <crm_internal.h>
#include <crm/common/unittest_internal.h>
#include <crm/common/xml.h>
#include <crm/common/scheduler.h>
#include <crm/pengine/internal.h>
#include <crm/pengine/status.h>
xmlNode *input = NULL;
pcmk_scheduler_t *scheduler = NULL;
pcmk_resource_t *exim_group, *promotable_0, *promotable_1, *dummy;
pcmk_resource_t *httpd_bundle, *mysql_group_0, *mysql_group_1;
static int
setup(void **state) {
char *path = NULL;
crm_xml_init();
path = crm_strdup_printf("%s/crm_mon.xml", getenv("PCMK_CTS_CLI_DIR"));
input = filename2xml(path);
free(path);
if (input == NULL) {
return 1;
}
scheduler = pe_new_working_set();
if (scheduler == NULL) {
return 1;
}
- pe__set_working_set_flags(scheduler,
+ pcmk__set_scheduler_flags(scheduler,
pcmk_sched_no_counts|pcmk_sched_no_compat);
scheduler->input = input;
cluster_status(scheduler);
/* Get references to several resources we use frequently. */
for (GList *iter = scheduler->resources; iter != NULL; iter = iter->next) {
pcmk_resource_t *rsc = (pcmk_resource_t *) iter->data;
if (strcmp(rsc->id, "dummy") == 0) {
dummy = rsc;
} else if (strcmp(rsc->id, "exim-group") == 0) {
exim_group = rsc;
} else if (strcmp(rsc->id, "httpd-bundle") == 0) {
httpd_bundle = rsc;
} else if (strcmp(rsc->id, "mysql-clone-group") == 0) {
for (GList *iter = rsc->children; iter != NULL; iter = iter->next) {
pcmk_resource_t *child = (pcmk_resource_t *) iter->data;
if (strcmp(child->id, "mysql-group:0") == 0) {
mysql_group_0 = child;
} else if (strcmp(child->id, "mysql-group:1") == 0) {
mysql_group_1 = child;
}
}
} else if (strcmp(rsc->id, "promotable-clone") == 0) {
for (GList *iter = rsc->children; iter != NULL; iter = iter->next) {
pcmk_resource_t *child = (pcmk_resource_t *) iter->data;
if (strcmp(child->id, "promotable-rsc:0") == 0) {
promotable_0 = child;
} else if (strcmp(child->id, "promotable-rsc:1") == 0) {
promotable_1 = child;
}
}
}
}
return 0;
}
static int
teardown(void **state) {
pe_free_working_set(scheduler);
return 0;
}
static void
bad_args(void **state) {
char *id = dummy->id;
assert_false(pe_base_name_eq(NULL, "dummy"));
assert_false(pe_base_name_eq(dummy, NULL));
dummy->id = NULL;
assert_false(pe_base_name_eq(dummy, "dummy"));
dummy->id = id;
}
static void
primitive_rsc(void **state) {
assert_true(pe_base_name_eq(dummy, "dummy"));
assert_false(pe_base_name_eq(dummy, "DUMMY"));
assert_false(pe_base_name_eq(dummy, "dUmMy"));
assert_false(pe_base_name_eq(dummy, "dummy0"));
assert_false(pe_base_name_eq(dummy, "dummy:0"));
}
static void
group_rsc(void **state) {
assert_true(pe_base_name_eq(exim_group, "exim-group"));
assert_false(pe_base_name_eq(exim_group, "EXIM-GROUP"));
assert_false(pe_base_name_eq(exim_group, "exim-group0"));
assert_false(pe_base_name_eq(exim_group, "exim-group:0"));
assert_false(pe_base_name_eq(exim_group, "Public-IP"));
}
static void
clone_rsc(void **state) {
assert_true(pe_base_name_eq(promotable_0, "promotable-rsc"));
assert_true(pe_base_name_eq(promotable_1, "promotable-rsc"));
assert_false(pe_base_name_eq(promotable_0, "promotable-rsc:0"));
assert_false(pe_base_name_eq(promotable_1, "promotable-rsc:1"));
assert_false(pe_base_name_eq(promotable_0, "PROMOTABLE-RSC"));
assert_false(pe_base_name_eq(promotable_1, "PROMOTABLE-RSC"));
assert_false(pe_base_name_eq(promotable_0, "Promotable-rsc"));
assert_false(pe_base_name_eq(promotable_1, "Promotable-rsc"));
}
static void
bundle_rsc(void **state) {
assert_true(pe_base_name_eq(httpd_bundle, "httpd-bundle"));
assert_false(pe_base_name_eq(httpd_bundle, "HTTPD-BUNDLE"));
assert_false(pe_base_name_eq(httpd_bundle, "httpd"));
assert_false(pe_base_name_eq(httpd_bundle, "httpd-docker-0"));
}
PCMK__UNIT_TEST(setup, teardown,
cmocka_unit_test(bad_args),
cmocka_unit_test(primitive_rsc),
cmocka_unit_test(group_rsc),
cmocka_unit_test(clone_rsc),
cmocka_unit_test(bundle_rsc))
diff --git a/lib/pengine/unpack.c b/lib/pengine/unpack.c
index f15abd399c..8f641826ed 100644
--- a/lib/pengine/unpack.c
+++ b/lib/pengine/unpack.c
@@ -1,5131 +1,5131 @@
/*
* 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 pe__set_working_set_flags()/pe__clear_working_set_flags() so that the
+ * 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 { \
const char *scf_value = pe_pref((scheduler)->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",
pe__node_name(node), reason, rsc->id);
} else {
pcmk__sched_warn("Guest node %s will be fenced "
"(by recovering its guest resource %s): %s",
pe__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;
pe__set_resource_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",
pe__node_name(node), reason);
pe__set_resource_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",
pe__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",
pe__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",
pe__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",
pe__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 XML_CIB_TAG_NVPAIR \
"[(@" XML_NVPAIR_ATTR_NAME "='" PCMK_STONITH_PROVIDES "'" \
"or @" XML_NVPAIR_ATTR_NAME "='" PCMK_META_REQUIRES "') " \
"and @" XML_NVPAIR_ATTR_VALUE "='" PCMK__VALUE_UNFENCING "']"
// unfencing in rsc_defaults or any resource
#define XPATH_ENABLE_UNFENCING \
"/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_RESOURCES \
"//" XML_TAG_META_SETS "/" XPATH_UNFENCING_NVPAIR \
"|/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_RSCCONFIG \
"/" XML_TAG_META_SETS "/" 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)) {
- pe__set_working_set_flags(scheduler, flag);
+ 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, XML_CIB_TAG_PROPSET, &rule_data, config_hash,
CIB_OPTIONS_FIRST, FALSE, scheduler);
verify_pe_options(scheduler->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 = pe_pref(scheduler->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");
- pe__set_working_set_flags(scheduler, pcmk_sched_have_fencing);
+ 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 = pe_pref(scheduler->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 = pe_pref(scheduler->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 = pe_pref(scheduler->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__btoa(pcmk_is_set(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 = pe_pref(scheduler->config_hash, PCMK_OPT_NO_QUORUM_POLICY);
if (pcmk__str_eq(value, "ignore", pcmk__str_casei)) {
scheduler->no_quorum_policy = pcmk_no_quorum_ignore;
} else if (pcmk__str_eq(value, "freeze", pcmk__str_casei)) {
scheduler->no_quorum_policy = pcmk_no_quorum_freeze;
} else if (pcmk__str_eq(value, "demote", pcmk__str_casei)) {
scheduler->no_quorum_policy = pcmk_no_quorum_demote;
} else if (pcmk__str_eq(value, "suicide", 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 = pe_pref(scheduler->config_hash, PCMK__OPT_REMOVE_AFTER_STOP);
if (value != NULL) {
if (crm_is_true(value)) {
- pe__set_working_set_flags(scheduler, pcmk_sched_remove_after_stop);
+ 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 {
- pe__clear_working_set_flags(scheduler,
+ 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__btoa(pcmk_is_set(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 = pe_pref(scheduler->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 = pe_pref(scheduler->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 = pe_pref(scheduler->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, "member", pcmk__str_null_matches | pcmk__str_casei)) {
new_node->details->type = pcmk_node_variant_cluster;
} else if (pcmk__str_eq(type, "remote", pcmk__str_casei)) {
new_node->details->type = pcmk_node_variant_remote;
- pe__set_working_set_flags(scheduler, pcmk_sched_have_remote_nodes);
+ 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, "ping", pcmk__str_casei)) {
pcmk__config_warn("Node %s has unrecognized type '%s', "
"assuming 'ping'", pcmk__s(uname, "without name"),
type);
}
pcmk__warn_once(pcmk__wo_ping_node,
"Support for nodes of type '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, XML_TAG_META_SETS,
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, XML_NVPAIR_ATTR_VALUE);
const char *name = crm_element_value(attr, XML_NVPAIR_ATTR_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, XML_CIB_TAG_NODE, pcmk__str_none)) {
new_node = NULL;
id = crm_element_value(xml_obj, PCMK_XA_ID);
uname = crm_element_value(xml_obj, XML_ATTR_UNAME);
type = crm_element_value(xml_obj, XML_ATTR_TYPE);
score = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE);
crm_trace("Processing node %s/%s", uname, id);
if (id == NULL) {
pcmk__config_err("Ignoring <" XML_CIB_TAG_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, XML_ATTR_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;
pe__set_resource_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, "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, XML_CIB_TAG_RESOURCE, 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, "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, XML_CIB_TAG_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, "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, pe__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, XML_CIB_TAG_RSC_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, XML_CIB_TAG_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, XML_CIB_TAG_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, "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, "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, "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, XML_CIB_TAG_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 *resource_discovery_enabled = NULL;
const xmlNode *attrs = NULL;
pcmk_resource_t *rsc = NULL;
if (!pcmk__str_eq((const char *)state->name, XML_CIB_TAG_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", pe__node_name(this_node));
pcmk__scan_min_int(crm_element_value(state, XML_NODE_IS_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, XML_TAG_TRANSIENT_NODEATTRS, FALSE);
add_node_attrs(attrs, this_node, TRUE, scheduler);
if (pe__shutdown_requested(this_node)) {
crm_info("%s is shutting down", pe__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", pe__node_name(this_node));
this_node->details->standby = TRUE;
}
if (crm_is_true(pe_node_attribute_raw(this_node, "maintenance")) ||
((rsc != NULL) && !pcmk_is_set(rsc->flags, pcmk_rsc_managed))) {
crm_info("%s is in maintenance mode", pe__node_name(this_node));
this_node->details->maintenance = TRUE;
}
resource_discovery_enabled = pe_node_attribute_raw(this_node, XML_NODE_ATTR_RSC_DISCOVERY);
if (resource_discovery_enabled && !crm_is_true(resource_discovery_enabled)) {
if (pe__is_remote_node(this_node)
&& !pcmk_is_set(scheduler->flags, pcmk_sched_fencing_enabled)) {
pcmk__config_warn("Ignoring " XML_NODE_ATTR_RSC_DISCOVERY
" attribute on Pacemaker Remote node %s"
" because fencing is disabled",
pe__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",
pe__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, XML_TAG_TRANSIENT_NODEATTRS,
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", pe__node_name(node));
node->details->standby = TRUE;
}
if (crm_is_true(pe_node_attribute_raw(node, "maintenance"))) {
crm_info("%s is in maintenance mode", pe__node_name(node));
node->details->maintenance = TRUE;
}
discovery = pe_node_attribute_raw(node, XML_NODE_ATTR_RSC_DISCOVERY);
if ((discovery != NULL) && !crm_is_true(discovery)) {
pcmk__config_warn("Ignoring " XML_NODE_ATTR_RSC_DISCOVERY " attribute "
"for %s because disabling resource discovery is "
"not allowed for cluster nodes", pe__node_name(node));
}
}
/*!
* \internal
* \brief Unpack a node state entry (first pass)
*
* Unpack one node state entry from status. This unpacks information from the
* 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 " XML_CIB_TAG_STATE " entry without "
PCMK_XA_ID);
crm_log_xml_info(state, "missing-id");
return;
}
uname = crm_element_value(state, XML_ATTR_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 node_state entry doesn't have an uname
* yet. We should recognize the node as `pending` and wait for it to
* join CPG.
*/
crm_trace("Handling " XML_CIB_TAG_STATE " entry with id=\"%s\" without "
XML_ATTR_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, XML_NODE_IS_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)",
pe__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 node_state entries in CIB status
for (const xmlNode *state = first_named_child(status, XML_CIB_TAG_STATE);
state != NULL; state = crm_next_same_xml(state)) {
const char *id = ID(state);
const char *uname = crm_element_value(state, XML_ATTR_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 "
XML_CIB_TAG_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, XML_CIB_TAG_TICKETS, pcmk__str_none)) {
unpack_tickets_state((xmlNode *) state, scheduler);
} else if (pcmk__str_eq((const char *)state->name, XML_CIB_TAG_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 = pe__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 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 " XML_CIB_TAG_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 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);
// @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 "
XML_CIB_TAG_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 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, pe__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", pe__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",
pe__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",
pe__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",
pe__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
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
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",
pe__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", pe__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", pe__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",
pe__node_name(this_node));
this_node->details->standby = TRUE;
this_node->details->pending = TRUE;
} else {
crm_trace("%s is down or still coming up",
pe__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", pe__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", pe__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", pe__node_name(this_node));
} else if (this_node->details->unclean) {
pcmk__sched_warn("%s is unclean", pe__node_name(this_node));
} else if (this_node->details->online) {
crm_info("%s is %s", pe__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", pe__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, XML_CIB_TAG_RESOURCE);
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, "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);
pe__set_resource_flags(rsc, pcmk_rsc_removed_filler);
}
pe__set_resource_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, pe__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, pe__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_node_t *) locations->data)->details == node->details) {
/* 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, pe__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 && inactive_instance->pending_node
&& (inactive_instance->pending_node->details != node->details)) {
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, pe__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, pe__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)) {
pe__clear_resource_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, role2text(rsc->role), pe__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),
pe__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)) {
pe__set_resource_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:
pe__set_resource_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
*/
pe__clear_resource_flags(rsc, pcmk_rsc_managed);
pe__set_resource_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, "on-fail=stop");
break;
case pcmk_on_fail_restart:
if ((rsc->role != pcmk_role_stopped)
&& (rsc->role != pcmk_role_unknown)) {
pe__set_resource_flags(rsc,
pcmk_rsc_failed|pcmk_rsc_stop_if_failed);
stop_action(rsc, node, FALSE);
}
break;
case pcmk_on_fail_restart_container:
pe__set_resource_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:
pe__set_resource_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, pe__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, pe__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:
pe__set_resource_flags(rsc, pcmk_rsc_failed);
break;
default:
pe__set_resource_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;
pe__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, pe__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, pe__node_name(node));
continue;
} else if (counter < start_index) {
pcmk__rsc_trace(rsc, "Skipping %s on %s: old %d",
id, pe__node_name(node), counter);
continue;
}
crm_element_value_ms(rsc_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
if (interval_ms == 0) {
pcmk__rsc_trace(rsc, "Skipping %s on %s: non-recurring",
id, pe__node_name(node));
continue;
}
status = crm_element_value(rsc_op, XML_LRM_ATTR_OPSTATUS);
if (pcmk__str_eq(status, "-1", pcmk__str_casei)) {
pcmk__rsc_trace(rsc, "Skipping %s on %s: status",
id, pe__node_name(node));
continue;
}
task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK);
/* create the action */
key = pcmk__op_key(rsc->id, task, interval_ms);
pcmk__rsc_trace(rsc, "Creating %s on %s", key, pe__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, XML_LRM_ATTR_TASK);
status = crm_element_value(rsc_op, XML_LRM_ATTR_OPSTATUS);
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, XML_LRM_ATTR_RC);
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, pe__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 lrm_resource entry from a node's CIB status
*
* \param[in,out] node Node whose status is being unpacked
* \param[in] rsc_entry 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 " XML_LRM_TAG_RESOURCE
" entry: No " PCMK_XA_ID);
crm_log_xml_info(lrm_resource, "missing-id");
return NULL;
}
crm_trace("Unpacking " XML_LRM_TAG_RESOURCE " for %s on %s",
rsc_id, pe__node_name(node));
// Build a list of individual lrm_rsc_op entries, so we can sort them
for (rsc_op = first_named_child(lrm_resource, XML_LRM_TAG_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, role2text(rsc->next_role),
role2text(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, XML_LRM_TAG_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 lrm_resources section
xml = find_xml_node(xml, XML_CIB_TAG_LRM, FALSE);
if (xml == NULL) {
return;
}
xml = find_xml_node(xml, XML_LRM_TAG_RESOURCES, FALSE);
if (xml == NULL) {
return;
}
// Unpack each lrm_resource entry
for (const xmlNode *rsc_entry = first_named_child(xml, XML_LRM_TAG_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 "/" XML_TAG_CIB "/" XML_CIB_TAG_STATUS \
"/" XML_CIB_TAG_STATE
#define SUB_XPATH_LRM_RESOURCE "/" XML_CIB_TAG_LRM \
"/" XML_LRM_TAG_RESOURCES \
"/" XML_LRM_TAG_RESOURCE
#define SUB_XPATH_LRM_RSC_OP "/" XML_LRM_TAG_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 "[@" XML_ATTR_UNAME "='", node, "']"
SUB_XPATH_LRM_RESOURCE "[@" PCMK_XA_ID "='", resource, "']"
SUB_XPATH_LRM_RSC_OP "[@" XML_LRM_ATTR_TASK "='", op, "'",
NULL);
/* Need to check against transition_magic too? */
if ((source != NULL) && (strcmp(op, PCMK_ACTION_MIGRATE_TO) == 0)) {
pcmk__g_strcat(xpath,
" and @" XML_LRM_ATTR_MIGRATE_TARGET "='", source, "']",
NULL);
} else if ((source != NULL)
&& (strcmp(op, PCMK_ACTION_MIGRATE_FROM) == 0)) {
pcmk__g_strcat(xpath,
" and @" XML_LRM_ATTR_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, XML_LRM_ATTR_RC, &rc);
crm_element_value_int(xml, XML_LRM_ATTR_OPSTATUS, &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 "[@" XML_ATTR_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;
GString *xpath = g_string_sized_new(256);
pcmk__g_strcat(xpath,
XPATH_NODE_STATE "[@" XML_ATTR_UNAME "='", node_name, "']"
SUB_XPATH_LRM_RESOURCE "[@" PCMK_XA_ID "='", rsc->id, "']"
SUB_XPATH_LRM_RSC_OP "[@" XML_LRM_ATTR_RC "!='193']",
NULL);
search = xpath_search(rsc->cluster->input, (const char *) xpath->str);
result = (numXpathResults(search) == 0);
freeXpathObject(search);
g_string_free(xpath, TRUE);
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, XML_LRM_TAG_RSC_OP);
op != NULL; op = crm_next_same_xml(op)) {
const char * task = NULL;
if (op == xml_op) {
continue;
}
task = crm_element_value(op, XML_LRM_ATTR_TASK);
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, XML_LRM_ATTR_MIGRATE_SOURCE);
target = crm_element_value(xml_op, XML_LRM_ATTR_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, XML_LRM_ATTR_MIGRATE_SOURCE);
*target_name = crm_element_value(entry, XML_LRM_ATTR_MIGRATE_TARGET);
if ((*source_name == NULL) || (*target_name == NULL)) {
pcmk__config_err("Ignoring resource history entry %s without "
XML_LRM_ATTR_MIGRATE_SOURCE " and "
XML_LRM_ATTR_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 "
XML_LRM_ATTR_MIGRATE_SOURCE "='%s' does not match %s",
ID(entry), *source_name, pe__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 "
XML_LRM_ATTR_MIGRATE_TARGET "='%s' does not match %s",
ID(entry), *target_name, pe__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, pe__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, XML_LRM_ATTR_RC, &from_rc);
crm_element_value_int(migrate_from, XML_LRM_ATTR_OPSTATUS,
&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
pe__set_resource_flags(history->rsc,
pcmk_rsc_failed|pcmk_rsc_stop_if_failed);
pe__clear_resource_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
pe__set_resource_flags(history->rsc,
pcmk_rsc_failed|pcmk_rsc_stop_if_failed);
pe__clear_resource_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 = pe__xe_history_key(xIter);
const char *uname = crm_element_value(xIter, XML_ATTR_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, pe__node_name(history->node));
return;
}
}
crm_trace("Adding entry for %s on %s to failed action list",
history->key, pe__node_name(history->node));
crm_xml_add(history->xml, XML_ATTR_UNAME, history->node->details->uname);
crm_xml_add(history->xml, XML_LRM_ATTR_RSCID, 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, XML_RSC_OP_LAST_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 on-fail=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, XML_OP_ATTR_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,
pe__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,
pe__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 resource-discovery option for location constraints",
history->rsc->id, pe__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, role2text(history->rsc->role),
pcmk__btoa(history->node->details->unclean),
fail2text(config_on_fail), role2text(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,
pe__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);
pe__clear_resource_flags(history->rsc, pcmk_rsc_managed);
pe__set_resource_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, pe__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, pe__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, pe__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, pe__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, XML_LRM_ATTR_INTERVAL_MS,
&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,
pe__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, pe__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, pe__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, XML_RSC_OP_LAST_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, pe__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, pe__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, XML_ATTR_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, pe__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 on-fail=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, pe__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, XML_LRM_ATTR_OPSTATUS,
&(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 " XML_LRM_ATTR_OPSTATUS " '%s'",
history->id, history->rsc->id,
pe__node_name(history->node),
pcmk__s(crm_element_value(history->xml,
XML_LRM_ATTR_OPSTATUS),
""));
return pcmk_rc_unpack_error;
}
if ((crm_element_value_int(history->xml, XML_LRM_ATTR_RC,
&(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 " XML_LRM_ATTR_RC " '%s'",
history->id, history->rsc->id,
pe__node_name(history->node),
pcmk__s(crm_element_value(history->xml,
XML_LRM_ATTR_RC),
""));
return pcmk_rc_unpack_error;
#else
history->exit_status = CRM_EX_ERROR;
#endif
}
history->exit_reason = crm_element_value(history->xml,
XML_LRM_ATTR_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,
pe__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,
pe__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 an 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, pe__node_name(history->node));
crm_xml_add(history->xml, XML_LRM_ATTR_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,
pe__node_name(history->node));
update_resource_state(history, history->expected_exit_status, last_failure,
on_fail);
crm_xml_add(history->xml, XML_ATTR_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
* 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, XML_LRM_ATTR_TASK),
pcmk__str_none)) {
return false; // last_failure is for different action
}
if ((crm_element_value_ms(last_failure, XML_LRM_ATTR_INTERVAL_MS,
&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, XML_RSC_OP_LAST_CHANGE),
&this_change, 0LL) != pcmk_rc_ok)
|| (pcmk__scan_ll(crm_element_value(last_failure,
XML_RSC_OP_LAST_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) {
pe__set_resource_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,
XML_LRM_ATTR_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, pe__node_name(node));
return;
}
// Task and interval
history.task = crm_element_value(xml_op, XML_LRM_ATTR_TASK);
if (history.task == NULL) {
pcmk__config_err("Ignoring resource history entry %s for %s on %s "
"without " XML_LRM_ATTR_TASK,
history.id, rsc->id, pe__node_name(node));
return;
}
crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS,
&(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, pe__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 = pe__xe_history_key(xml_op);
crm_element_value_int(xml_op, XML_LRM_ATTR_CALLID, &(history.call_id));
pcmk__rsc_trace(rsc, "Unpacking %s (%s call %d on %s): %s (%s)",
history.id, history.task, history.call_id,
pe__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, pe__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, pe__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.
*/
pe__set_resource_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, pe__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, XML_ATTR_UNAME, node->details->uname);
pe__set_resource_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, pe__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, pe__node_name(node), history.id,
role2text(rsc->role), role2text(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(XML_BOOLEAN_TRUE));
} else {
g_hash_table_insert(node->details->attrs,
strdup(CRM_ATTR_IS_DC), strdup(XML_BOOLEAN_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, XML_TAG_ATTR_SETS, &rule_data,
node->details->attrs, NULL, overwrite,
scheduler);
pe__unpack_dataset_nvpairs(xml_obj, XML_TAG_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, XML_LRM_TAG_RSC_OP,
pcmk__str_none)) {
crm_xml_add(rsc_op, "resource", rsc);
crm_xml_add(rsc_op, XML_ATTR_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, XML_CIB_TAG_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, XML_CIB_TAG_STATE, pcmk__str_none)) {
const char *uname = crm_element_value(node_state, XML_ATTR_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, XML_CIB_TAG_LRM, FALSE);
tmp = find_xml_node(tmp, XML_LRM_TAG_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,
XML_LRM_TAG_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;
}
diff --git a/tools/crm_resource.c b/tools/crm_resource.c
index afb71cc299..de3766d5f8 100644
--- a/tools/crm_resource.c
+++ b/tools/crm_resource.c
@@ -1,2086 +1,2086 @@
/*
* 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 <crm_resource.h>
#include <crm/lrmd_internal.h>
#include <crm/common/cmdline_internal.h>
#include <crm/common/ipc_attrd_internal.h>
#include <crm/common/lists_internal.h>
#include <crm/common/output.h>
#include <pacemaker-internal.h>
#include <sys/param.h>
#include <stdio.h>
#include <sys/types.h>
#include <unistd.h>
#include <stdlib.h>
#include <errno.h>
#include <fcntl.h>
#include <libgen.h>
#include <time.h>
#include <crm/crm.h>
#include <crm/stonith-ng.h>
#include <crm/common/ipc_controld.h>
#include <crm/cib/internal.h>
#define SUMMARY "crm_resource - perform tasks related to Pacemaker cluster resources"
enum rsc_command {
cmd_none = 0, // No command option given (yet)
cmd_ban,
cmd_cleanup,
cmd_clear,
cmd_colocations,
cmd_cts,
cmd_delete,
cmd_delete_param,
cmd_digests,
cmd_execute_agent,
cmd_fail,
cmd_get_param,
cmd_get_property,
cmd_list_active_ops,
cmd_list_agents,
cmd_list_all_ops,
cmd_list_alternatives,
cmd_list_instances,
cmd_list_providers,
cmd_list_resources,
cmd_list_standards,
cmd_locate,
cmd_metadata,
cmd_move,
cmd_query_raw_xml,
cmd_query_xml,
cmd_refresh,
cmd_restart,
cmd_set_param,
cmd_set_property,
cmd_wait,
cmd_why,
};
struct {
enum rsc_command rsc_cmd; // crm_resource command to perform
// Infrastructure that given command needs to work
gboolean require_cib; // Whether command requires CIB IPC
int cib_options; // Options to use with CIB IPC calls
gboolean require_crmd; // Whether command requires controller IPC
gboolean require_scheduler; // Whether command requires scheduler data
gboolean require_resource; // Whether command requires resource specified
gboolean require_node; // Whether command requires node specified
int find_flags; // Flags to use when searching for resource
// Command-line option values
gchar *rsc_id; // Value of --resource
gchar *rsc_type; // Value of --resource-type
gboolean force; // --force was given
gboolean clear_expired; // --expired was given
gboolean recursive; // --recursive was given
gboolean promoted_role_only; // --promoted was given
gchar *host_uname; // Value of --node
gchar *interval_spec; // Value of --interval
gchar *move_lifetime; // Value of --lifetime
gchar *operation; // Value of --operation
const char *attr_set_type; // Instance, meta, utilization, or element attribute
gchar *prop_id; // --nvpair (attribute XML ID)
char *prop_name; // Attribute name
gchar *prop_set; // --set-name (attribute block XML ID)
gchar *prop_value; // --parameter-value (attribute value)
int timeout_ms; // Parsed from --timeout value
char *agent_spec; // Standard and/or provider and/or agent
gchar *xml_file; // Value of (deprecated) --xml-file
int check_level; // Optional value of --validate or --force-check
// Resource configuration specified via command-line arguments
gboolean cmdline_config; // Resource configuration was via arguments
char *v_agent; // Value of --agent
char *v_class; // Value of --class
char *v_provider; // Value of --provider
GHashTable *cmdline_params; // Resource parameters specified
// Positional command-line arguments
gchar **remainder; // Positional arguments as given
GHashTable *override_params; // Resource parameter values that override config
} options = {
.attr_set_type = XML_TAG_ATTR_SETS,
.check_level = -1,
.cib_options = cib_sync_call,
.require_cib = TRUE,
.require_scheduler = TRUE,
.require_resource = TRUE,
};
#if 0
// @COMPAT @TODO enable this at next backward compatibility break
#define SET_COMMAND(cmd) do { \
if (options.rsc_cmd != cmd_none) { \
g_set_error(error, PCMK__EXITC_ERROR, CRM_EX_USAGE, \
"Only one command option may be specified"); \
return FALSE; \
} \
options.rsc_cmd = (cmd); \
} while (0)
#else
#define SET_COMMAND(cmd) do { \
if (options.rsc_cmd != cmd_none) { \
reset_options(); \
} \
options.rsc_cmd = (cmd); \
} while (0)
#endif
gboolean agent_provider_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
gboolean attr_set_type_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
gboolean class_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
gboolean cleanup_refresh_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
gboolean delete_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
gboolean expired_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
gboolean list_agents_cb(const gchar *option_name, const gchar *optarg,
gpointer data, GError **error);
gboolean list_providers_cb(const gchar *option_name, const gchar *optarg,
gpointer data, GError **error);
gboolean list_standards_cb(const gchar *option_name, const gchar *optarg,
gpointer data, GError **error);
gboolean list_alternatives_cb(const gchar *option_name, const gchar *optarg,
gpointer data, GError **error);
gboolean metadata_cb(const gchar *option_name, const gchar *optarg,
gpointer data, GError **error);
gboolean option_cb(const gchar *option_name, const gchar *optarg,
gpointer data, GError **error);
gboolean fail_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
gboolean flag_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
gboolean get_param_prop_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
gboolean list_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
gboolean set_delete_param_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
gboolean set_prop_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
gboolean timeout_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
gboolean validate_or_force_cb(const gchar *option_name, const gchar *optarg,
gpointer data, GError **error);
gboolean restart_cb(const gchar *option_name, const gchar *optarg,
gpointer data, GError **error);
gboolean digests_cb(const gchar *option_name, const gchar *optarg,
gpointer data, GError **error);
gboolean wait_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
gboolean why_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
static crm_exit_t exit_code = CRM_EX_OK;
static pcmk__output_t *out = NULL;
static pcmk__common_args_t *args = NULL;
// Things that should be cleaned up on exit
static GError *error = NULL;
static GMainLoop *mainloop = NULL;
static cib_t *cib_conn = NULL;
static pcmk_ipc_api_t *controld_api = NULL;
static pcmk_scheduler_t *scheduler = NULL;
#define MESSAGE_TIMEOUT_S 60
#define INDENT " "
static pcmk__supported_format_t formats[] = {
PCMK__SUPPORTED_FORMAT_NONE,
PCMK__SUPPORTED_FORMAT_TEXT,
PCMK__SUPPORTED_FORMAT_XML,
{ NULL, NULL, NULL }
};
// Clean up and exit
static crm_exit_t
bye(crm_exit_t ec)
{
pcmk__output_and_clear_error(&error, out);
if (out != NULL) {
out->finish(out, ec, true, NULL);
pcmk__output_free(out);
}
pcmk__unregister_formats();
if (cib_conn != NULL) {
cib_t *save_cib_conn = cib_conn;
cib_conn = NULL; // Ensure we can't free this twice
cib__clean_up_connection(&save_cib_conn);
}
if (controld_api != NULL) {
pcmk_ipc_api_t *save_controld_api = controld_api;
controld_api = NULL; // Ensure we can't free this twice
pcmk_free_ipc_api(save_controld_api);
}
if (mainloop != NULL) {
g_main_loop_unref(mainloop);
mainloop = NULL;
}
pe_free_working_set(scheduler);
scheduler = NULL;
crm_exit(ec);
return ec;
}
static void
quit_main_loop(crm_exit_t ec)
{
exit_code = ec;
if (mainloop != NULL) {
GMainLoop *mloop = mainloop;
mainloop = NULL; // Don't re-enter this block
pcmk_quit_main_loop(mloop, 10);
g_main_loop_unref(mloop);
}
}
static gboolean
resource_ipc_timeout(gpointer data)
{
// Start with newline because "Waiting for ..." message doesn't have one
if (error != NULL) {
g_clear_error(&error);
}
g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_TIMEOUT,
_("Aborting because no messages received in %d seconds"), MESSAGE_TIMEOUT_S);
quit_main_loop(CRM_EX_TIMEOUT);
return FALSE;
}
static void
controller_event_callback(pcmk_ipc_api_t *api, enum pcmk_ipc_event event_type,
crm_exit_t status, void *event_data, void *user_data)
{
switch (event_type) {
case pcmk_ipc_event_disconnect:
if (exit_code == CRM_EX_DISCONNECT) { // Unexpected
crm_info("Connection to controller was terminated");
}
quit_main_loop(exit_code);
break;
case pcmk_ipc_event_reply:
if (status != CRM_EX_OK) {
out->err(out, "Error: bad reply from controller: %s",
crm_exit_str(status));
pcmk_disconnect_ipc(api);
quit_main_loop(status);
} else {
if ((pcmk_controld_api_replies_expected(api) == 0)
&& mainloop && g_main_loop_is_running(mainloop)) {
out->info(out, "... got reply (done)");
crm_debug("Got all the replies we expected");
pcmk_disconnect_ipc(api);
quit_main_loop(CRM_EX_OK);
} else {
out->info(out, "... got reply");
}
}
break;
default:
break;
}
}
static void
start_mainloop(pcmk_ipc_api_t *capi)
{
unsigned int count = pcmk_controld_api_replies_expected(capi);
if (count > 0) {
out->info(out, "Waiting for %u %s from the controller",
count, pcmk__plural_alt(count, "reply", "replies"));
exit_code = CRM_EX_DISCONNECT; // For unexpected disconnects
mainloop = g_main_loop_new(NULL, FALSE);
g_timeout_add(MESSAGE_TIMEOUT_S * 1000, resource_ipc_timeout, NULL);
g_main_loop_run(mainloop);
}
}
static int
compare_id(gconstpointer a, gconstpointer b)
{
return strcmp((const char *)a, (const char *)b);
}
static GList *
build_constraint_list(xmlNode *root)
{
GList *retval = NULL;
xmlNode *cib_constraints = NULL;
xmlXPathObjectPtr xpathObj = NULL;
int ndx = 0;
cib_constraints = pcmk_find_cib_element(root, XML_CIB_TAG_CONSTRAINTS);
xpathObj = xpath_search(cib_constraints, "//" XML_CONS_TAG_RSC_LOCATION);
for (ndx = 0; ndx < numXpathResults(xpathObj); ndx++) {
xmlNode *match = getXpathResult(xpathObj, ndx);
retval = g_list_insert_sorted(retval, (gpointer) ID(match), compare_id);
}
freeXpathObject(xpathObj);
return retval;
}
/* short option letters still available: eEJkKXyYZ */
static GOptionEntry query_entries[] = {
{ "list", 'L', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, list_cb,
"List all cluster resources with status",
NULL },
{ "list-raw", 'l', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, list_cb,
"List IDs of all instantiated resources (individual members\n"
INDENT "rather than groups etc.)",
NULL },
{ "list-cts", 'c', G_OPTION_FLAG_HIDDEN|G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, list_cb,
NULL,
NULL },
{ "list-operations", 'O', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, list_cb,
"List active resource operations, optionally filtered by\n"
INDENT "--resource and/or --node",
NULL },
{ "list-all-operations", 'o', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, list_cb,
"List all resource operations, optionally filtered by\n"
INDENT "--resource and/or --node",
NULL },
{ "list-standards", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK,
list_standards_cb,
"List supported standards",
NULL },
{ "list-ocf-providers", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK,
list_providers_cb,
"List all available OCF providers",
NULL },
{ "list-agents", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK,
list_agents_cb,
"List all agents available for the named standard and/or provider",
"STD:PROV" },
{ "list-ocf-alternatives", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK,
list_alternatives_cb,
"List all available providers for the named OCF agent",
"AGENT" },
{ "show-metadata", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK,
metadata_cb,
"Show the metadata for the named class:provider:agent",
"SPEC" },
{ "query-xml", 'q', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, flag_cb,
"Show XML configuration of resource (after any template expansion)",
NULL },
{ "query-xml-raw", 'w', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, flag_cb,
"Show XML configuration of resource (before any template expansion)",
NULL },
{ "get-parameter", 'g', G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, get_param_prop_cb,
"Display named parameter for resource (use instance attribute\n"
INDENT "unless --element, --meta, or --utilization is specified)",
"PARAM" },
{ "get-property", 'G', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_CALLBACK, get_param_prop_cb,
"Display named property of resource ('class', 'type', or 'provider') "
"(requires --resource)",
"PROPERTY" },
{ "locate", 'W', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, flag_cb,
"Show node(s) currently running resource",
NULL },
{ "constraints", 'a', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, flag_cb,
"Display the location and colocation constraints that apply to a\n"
INDENT "resource, and if --recursive is specified, to the resources\n"
INDENT "directly or indirectly involved in those colocations.\n"
INDENT "If the named resource is part of a group, or a clone or\n"
INDENT "bundle instance, constraints for the collective resource\n"
INDENT "will be shown unless --force is given.",
NULL },
{ "stack", 'A', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, flag_cb,
"Equivalent to --constraints --recursive",
NULL },
{ "why", 'Y', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, why_cb,
"Show why resources are not running, optionally filtered by\n"
INDENT "--resource and/or --node",
NULL },
{ NULL }
};
static GOptionEntry command_entries[] = {
{ "validate", 0, G_OPTION_FLAG_OPTIONAL_ARG, G_OPTION_ARG_CALLBACK,
validate_or_force_cb,
"Validate resource configuration by calling agent's validate-all\n"
INDENT "action. The configuration may be specified either by giving an\n"
INDENT "existing resource name with -r, or by specifying --class,\n"
INDENT "--agent, and --provider arguments, along with any number of\n"
INDENT "--option arguments. An optional LEVEL argument can be given\n"
INDENT "to control the level of checking performed.",
"LEVEL" },
{ "cleanup", 'C', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, cleanup_refresh_cb,
"If resource has any past failures, clear its history and fail\n"
INDENT "count. Optionally filtered by --resource, --node, --operation\n"
INDENT "and --interval (otherwise all). --operation and --interval\n"
INDENT "apply to fail counts, but entire history is always clear, to\n"
INDENT "allow current state to be rechecked. If the named resource is\n"
INDENT "part of a group, or one numbered instance of a clone or bundled\n"
INDENT "resource, the clean-up applies to the whole collective resource\n"
INDENT "unless --force is given.",
NULL },
{ "refresh", 'R', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, cleanup_refresh_cb,
"Delete resource's history (including failures) so its current state\n"
INDENT "is rechecked. Optionally filtered by --resource and --node\n"
INDENT "(otherwise all). If the named resource is part of a group, or one\n"
INDENT "numbered instance of a clone or bundled resource, the refresh\n"
INDENT "applies to the whole collective resource unless --force is given.",
NULL },
{ "set-parameter", 'p', G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, set_delete_param_cb,
"Set named parameter for resource (requires -v). Use instance\n"
INDENT "attribute unless --element, --meta, or --utilization is "
"specified.",
"PARAM" },
{ "delete-parameter", 'd', G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, set_delete_param_cb,
"Delete named parameter for resource. Use instance attribute\n"
INDENT "unless --element, --meta or, --utilization is specified.",
"PARAM" },
{ "set-property", 'S', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_CALLBACK, set_prop_cb,
"Set named property of resource ('class', 'type', or 'provider') "
"(requires -r, -t, -v)",
"PROPERTY" },
{ NULL }
};
static GOptionEntry location_entries[] = {
{ "move", 'M', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, flag_cb,
"Create a constraint to move resource. If --node is specified,\n"
INDENT "the constraint will be to move to that node, otherwise it\n"
INDENT "will be to ban the current node. Unless --force is specified\n"
INDENT "this will return an error if the resource is already running\n"
INDENT "on the specified node. If --force is specified, this will\n"
INDENT "always ban the current node.\n"
INDENT "Optional: --lifetime, --promoted. NOTE: This may prevent the\n"
INDENT "resource from running on its previous location until the\n"
INDENT "implicit constraint expires or is removed with --clear.",
NULL },
{ "ban", 'B', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, flag_cb,
"Create a constraint to keep resource off a node.\n"
INDENT "Optional: --node, --lifetime, --promoted.\n"
INDENT "NOTE: This will prevent the resource from running on the\n"
INDENT "affected node until the implicit constraint expires or is\n"
INDENT "removed with --clear. If --node is not specified, it defaults\n"
INDENT "to the node currently running the resource for primitives\n"
INDENT "and groups, or the promoted instance of promotable clones with\n"
INDENT PCMK_META_PROMOTED_MAX "=1 (all other situations result in an\n"
INDENT "error as there is no sane default).",
NULL },
{ "clear", 'U', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, flag_cb,
"Remove all constraints created by the --ban and/or --move\n"
INDENT "commands. Requires: --resource. Optional: --node, --promoted,\n"
INDENT "--expired. If --node is not specified, all constraints created\n"
INDENT "by --ban and --move will be removed for the named resource. If\n"
INDENT "--node and --force are specified, any constraint created by\n"
INDENT "--move will be cleared, even if it is not for the specified\n"
INDENT "node. If --expired is specified, only those constraints whose\n"
INDENT "lifetimes have expired will be removed.",
NULL },
{ "expired", 'e', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, expired_cb,
"Modifies the --clear argument to remove constraints with\n"
INDENT "expired lifetimes.",
NULL },
{ "lifetime", 'u', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.move_lifetime,
"Lifespan (as ISO 8601 duration) of created constraints (with\n"
INDENT "-B, -M) see https://en.wikipedia.org/wiki/ISO_8601#Durations)",
"TIMESPEC" },
{ "promoted", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE,
&options.promoted_role_only,
"Limit scope of command to promoted role (with -B, -M, -U). For\n"
INDENT "-B and -M, previously promoted instances may remain\n"
INDENT "active in the unpromoted role.",
NULL },
// Deprecated since 2.1.0
{ "master", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE,
&options.promoted_role_only,
"Deprecated: Use --promoted instead", NULL },
{ NULL }
};
static GOptionEntry advanced_entries[] = {
{ "delete", 'D', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, delete_cb,
"(Advanced) Delete a resource from the CIB. Required: -t",
NULL },
{ "fail", 'F', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, fail_cb,
"(Advanced) Tell the cluster this resource has failed",
NULL },
{ "restart", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, restart_cb,
"(Advanced) Tell the cluster to restart this resource and\n"
INDENT "anything that depends on it",
NULL },
{ "wait", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, wait_cb,
"(Advanced) Wait until the cluster settles into a stable state",
NULL },
{ "digests", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, digests_cb,
"(Advanced) Show parameter hashes that Pacemaker uses to detect\n"
INDENT "configuration changes (only accurate if there is resource\n"
INDENT "history on the specified node). Required: --resource, --node.\n"
INDENT "Optional: any NAME=VALUE parameters will be used to override\n"
INDENT "the configuration (to see what the hash would be with those\n"
INDENT "changes).",
NULL },
{ "force-demote", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK,
validate_or_force_cb,
"(Advanced) Bypass the cluster and demote a resource on the local\n"
INDENT "node. Unless --force is specified, this will refuse to do so if\n"
INDENT "the cluster believes the resource is a clone instance already\n"
INDENT "running on the local node.",
NULL },
{ "force-stop", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK,
validate_or_force_cb,
"(Advanced) Bypass the cluster and stop a resource on the local node",
NULL },
{ "force-start", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK,
validate_or_force_cb,
"(Advanced) Bypass the cluster and start a resource on the local\n"
INDENT "node. Unless --force is specified, this will refuse to do so if\n"
INDENT "the cluster believes the resource is a clone instance already\n"
INDENT "running on the local node.",
NULL },
{ "force-promote", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK,
validate_or_force_cb,
"(Advanced) Bypass the cluster and promote a resource on the local\n"
INDENT "node. Unless --force is specified, this will refuse to do so if\n"
INDENT "the cluster believes the resource is a clone instance already\n"
INDENT "running on the local node.",
NULL },
{ "force-check", 0, G_OPTION_FLAG_OPTIONAL_ARG, G_OPTION_ARG_CALLBACK,
validate_or_force_cb,
"(Advanced) Bypass the cluster and check the state of a resource on\n"
INDENT "the local node. An optional LEVEL argument can be given\n"
INDENT "to control the level of checking performed.",
"LEVEL" },
{ NULL }
};
static GOptionEntry addl_entries[] = {
{ "node", 'N', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.host_uname,
"Node name",
"NAME" },
{ "recursive", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE, &options.recursive,
"Follow colocation chains when using --set-parameter or --constraints",
NULL },
{ "resource-type", 't', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.rsc_type,
"Resource XML element (primitive, group, etc.) (with -D)",
"ELEMENT" },
{ "parameter-value", 'v', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.prop_value,
"Value to use with -p",
"PARAM" },
{ "meta", 'm', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, attr_set_type_cb,
"Use resource meta-attribute instead of instance attribute\n"
INDENT "(with -p, -g, -d)",
NULL },
{ "utilization", 'z', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, attr_set_type_cb,
"Use resource utilization attribute instead of instance attribute\n"
INDENT "(with -p, -g, -d)",
NULL },
{ "element", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, attr_set_type_cb,
"Use resource element attribute instead of instance attribute\n"
INDENT "(with -p, -g, -d)",
NULL },
{ "operation", 'n', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.operation,
"Operation to clear instead of all (with -C -r)",
"OPERATION" },
{ "interval", 'I', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.interval_spec,
"Interval of operation to clear (default 0) (with -C -r -n)",
"N" },
{ "class", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, class_cb,
"The standard the resource agent conforms to (for example, ocf).\n"
INDENT "Use with --agent, --provider, --option, and --validate.",
"CLASS" },
{ "agent", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, agent_provider_cb,
"The agent to use (for example, IPaddr). Use with --class,\n"
INDENT "--provider, --option, and --validate.",
"AGENT" },
{ "provider", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, agent_provider_cb,
"The vendor that supplies the resource agent (for example,\n"
INDENT "heartbeat). Use with --class, --agent, --option, and --validate.",
"PROVIDER" },
{ "option", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, option_cb,
"Specify a device configuration parameter as NAME=VALUE (may be\n"
INDENT "specified multiple times). Use with --validate and without the\n"
INDENT "-r option.",
"PARAM" },
{ "set-name", 's', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.prop_set,
"(Advanced) XML ID of attributes element to use (with -p, -d)",
"ID" },
{ "nvpair", 'i', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.prop_id,
"(Advanced) XML ID of nvpair element to use (with -p, -d)",
"ID" },
{ "timeout", 'T', G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, timeout_cb,
"(Advanced) Abort if command does not finish in this time (with\n"
INDENT "--restart, --wait, --force-*)",
"N" },
{ "force", 'f', G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE, &options.force,
"Force the action to be performed. See help for individual commands for\n"
INDENT "additional behavior.",
NULL },
{ "xml-file", 'x', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_FILENAME, &options.xml_file,
NULL,
"FILE" },
{ "host-uname", 'H', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_STRING, &options.host_uname,
NULL,
"HOST" },
{ NULL }
};
static void
reset_options(void) {
options.require_crmd = FALSE;
options.require_node = FALSE;
options.require_cib = TRUE;
options.require_scheduler = TRUE;
options.require_resource = TRUE;
options.find_flags = 0;
}
gboolean
agent_provider_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.cmdline_config = TRUE;
options.require_resource = FALSE;
if (pcmk__str_eq(option_name, "--provider", pcmk__str_casei)) {
pcmk__str_update(&options.v_provider, optarg);
} else {
pcmk__str_update(&options.v_agent, optarg);
}
return TRUE;
}
gboolean
attr_set_type_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
if (pcmk__str_any_of(option_name, "-m", "--meta", NULL)) {
options.attr_set_type = XML_TAG_META_SETS;
} else if (pcmk__str_any_of(option_name, "-z", "--utilization", NULL)) {
options.attr_set_type = XML_TAG_UTILIZATION;
} else if (pcmk__str_eq(option_name, "--element", pcmk__str_casei)) {
options.attr_set_type = ATTR_SET_ELEMENT;
}
return TRUE;
}
gboolean
class_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
pcmk__str_update(&options.v_class, optarg);
options.cmdline_config = TRUE;
options.require_resource = FALSE;
return TRUE;
}
gboolean
cleanup_refresh_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
if (pcmk__str_any_of(option_name, "-C", "--cleanup", NULL)) {
SET_COMMAND(cmd_cleanup);
} else {
SET_COMMAND(cmd_refresh);
}
options.require_resource = FALSE;
if (getenv("CIB_file") == NULL) {
options.require_crmd = TRUE;
}
options.find_flags = pcmk_rsc_match_history|pcmk_rsc_match_anon_basename;
return TRUE;
}
gboolean
delete_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
SET_COMMAND(cmd_delete);
options.require_scheduler = FALSE;
options.find_flags = pcmk_rsc_match_history|pcmk_rsc_match_basename;
return TRUE;
}
gboolean
expired_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.clear_expired = TRUE;
options.require_resource = FALSE;
return TRUE;
}
static void
get_agent_spec(const gchar *optarg)
{
options.require_cib = FALSE;
options.require_scheduler = FALSE;
options.require_resource = FALSE;
pcmk__str_update(&options.agent_spec, optarg);
}
gboolean
list_agents_cb(const gchar *option_name, const gchar *optarg, gpointer data,
GError **error)
{
SET_COMMAND(cmd_list_agents);
get_agent_spec(optarg);
return TRUE;
}
gboolean
list_providers_cb(const gchar *option_name, const gchar *optarg, gpointer data,
GError **error)
{
SET_COMMAND(cmd_list_providers);
get_agent_spec(optarg);
return TRUE;
}
gboolean
list_standards_cb(const gchar *option_name, const gchar *optarg, gpointer data,
GError **error)
{
SET_COMMAND(cmd_list_standards);
options.require_cib = FALSE;
options.require_scheduler = FALSE;
options.require_resource = FALSE;
return TRUE;
}
gboolean
list_alternatives_cb(const gchar *option_name, const gchar *optarg,
gpointer data, GError **error)
{
SET_COMMAND(cmd_list_alternatives);
get_agent_spec(optarg);
return TRUE;
}
gboolean
metadata_cb(const gchar *option_name, const gchar *optarg, gpointer data,
GError **error)
{
SET_COMMAND(cmd_metadata);
get_agent_spec(optarg);
return TRUE;
}
gboolean
option_cb(const gchar *option_name, const gchar *optarg, gpointer data,
GError **error)
{
char *name = NULL;
char *value = NULL;
if (pcmk__scan_nvpair(optarg, &name, &value) != 2) {
return FALSE;
}
if (options.cmdline_params == NULL) {
options.cmdline_params = pcmk__strkey_table(free, free);
}
g_hash_table_replace(options.cmdline_params, name, value);
return TRUE;
}
gboolean
fail_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
SET_COMMAND(cmd_fail);
options.require_crmd = TRUE;
options.require_node = TRUE;
return TRUE;
}
gboolean
flag_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
if (pcmk__str_any_of(option_name, "-U", "--clear", NULL)) {
SET_COMMAND(cmd_clear);
options.find_flags = pcmk_rsc_match_history
|pcmk_rsc_match_anon_basename;
} else if (pcmk__str_any_of(option_name, "-B", "--ban", NULL)) {
SET_COMMAND(cmd_ban);
options.find_flags = pcmk_rsc_match_history
|pcmk_rsc_match_anon_basename;
} else if (pcmk__str_any_of(option_name, "-M", "--move", NULL)) {
SET_COMMAND(cmd_move);
options.find_flags = pcmk_rsc_match_history
|pcmk_rsc_match_anon_basename;
} else if (pcmk__str_any_of(option_name, "-q", "--query-xml", NULL)) {
SET_COMMAND(cmd_query_xml);
options.find_flags = pcmk_rsc_match_history|pcmk_rsc_match_basename;
} else if (pcmk__str_any_of(option_name, "-w", "--query-xml-raw", NULL)) {
SET_COMMAND(cmd_query_raw_xml);
options.find_flags = pcmk_rsc_match_history|pcmk_rsc_match_basename;
} else if (pcmk__str_any_of(option_name, "-W", "--locate", NULL)) {
SET_COMMAND(cmd_locate);
options.find_flags = pcmk_rsc_match_history
|pcmk_rsc_match_anon_basename;
} else if (pcmk__str_any_of(option_name, "-a", "--constraints", NULL)) {
SET_COMMAND(cmd_colocations);
options.find_flags = pcmk_rsc_match_history
|pcmk_rsc_match_anon_basename;
} else if (pcmk__str_any_of(option_name, "-A", "--stack", NULL)) {
SET_COMMAND(cmd_colocations);
options.find_flags = pcmk_rsc_match_history
|pcmk_rsc_match_anon_basename;
options.recursive = TRUE;
}
return TRUE;
}
gboolean
get_param_prop_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
if (pcmk__str_any_of(option_name, "-g", "--get-parameter", NULL)) {
SET_COMMAND(cmd_get_param);
} else {
SET_COMMAND(cmd_get_property);
}
pcmk__str_update(&options.prop_name, optarg);
options.find_flags = pcmk_rsc_match_history|pcmk_rsc_match_basename;
return TRUE;
}
gboolean
list_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
if (pcmk__str_any_of(option_name, "-c", "--list-cts", NULL)) {
SET_COMMAND(cmd_cts);
} else if (pcmk__str_any_of(option_name, "-L", "--list", NULL)) {
SET_COMMAND(cmd_list_resources);
} else if (pcmk__str_any_of(option_name, "-l", "--list-raw", NULL)) {
SET_COMMAND(cmd_list_instances);
} else if (pcmk__str_any_of(option_name, "-O", "--list-operations", NULL)) {
SET_COMMAND(cmd_list_active_ops);
} else {
SET_COMMAND(cmd_list_all_ops);
}
options.require_resource = FALSE;
return TRUE;
}
gboolean
set_delete_param_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
if (pcmk__str_any_of(option_name, "-p", "--set-parameter", NULL)) {
SET_COMMAND(cmd_set_param);
} else {
SET_COMMAND(cmd_delete_param);
}
pcmk__str_update(&options.prop_name, optarg);
options.find_flags = pcmk_rsc_match_history|pcmk_rsc_match_basename;
return TRUE;
}
gboolean
set_prop_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
SET_COMMAND(cmd_set_property);
options.require_scheduler = FALSE;
pcmk__str_update(&options.prop_name, optarg);
options.find_flags = pcmk_rsc_match_history|pcmk_rsc_match_basename;
return TRUE;
}
gboolean
timeout_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.timeout_ms = crm_get_msec(optarg);
return TRUE;
}
gboolean
validate_or_force_cb(const gchar *option_name, const gchar *optarg,
gpointer data, GError **error)
{
SET_COMMAND(cmd_execute_agent);
if (options.operation) {
g_free(options.operation);
}
options.operation = g_strdup(option_name + 2); // skip "--"
options.find_flags = pcmk_rsc_match_history|pcmk_rsc_match_anon_basename;
if (options.override_params == NULL) {
options.override_params = pcmk__strkey_table(free, free);
}
if (optarg != NULL) {
if (pcmk__scan_min_int(optarg, &options.check_level, 0) != pcmk_rc_ok) {
g_set_error(error, G_OPTION_ERROR, CRM_EX_INVALID_PARAM,
_("Invalid check level setting: %s"), optarg);
return FALSE;
}
}
return TRUE;
}
gboolean
restart_cb(const gchar *option_name, const gchar *optarg, gpointer data,
GError **error)
{
SET_COMMAND(cmd_restart);
options.find_flags = pcmk_rsc_match_history|pcmk_rsc_match_anon_basename;
return TRUE;
}
gboolean
digests_cb(const gchar *option_name, const gchar *optarg, gpointer data,
GError **error)
{
SET_COMMAND(cmd_digests);
options.find_flags = pcmk_rsc_match_history|pcmk_rsc_match_anon_basename;
if (options.override_params == NULL) {
options.override_params = pcmk__strkey_table(free, free);
}
options.require_node = TRUE;
options.require_scheduler = TRUE;
return TRUE;
}
gboolean
wait_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
SET_COMMAND(cmd_wait);
options.require_resource = FALSE;
options.require_scheduler = FALSE;
return TRUE;
}
gboolean
why_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
SET_COMMAND(cmd_why);
options.require_resource = FALSE;
options.find_flags = pcmk_rsc_match_history|pcmk_rsc_match_anon_basename;
return TRUE;
}
static int
ban_or_move(pcmk__output_t *out, pcmk_resource_t *rsc,
const char *move_lifetime)
{
int rc = pcmk_rc_ok;
pcmk_node_t *current = NULL;
unsigned int nactive = 0;
CRM_CHECK(rsc != NULL, return EINVAL);
current = pe__find_active_requires(rsc, &nactive);
if (nactive == 1) {
rc = cli_resource_ban(out, options.rsc_id, current->details->uname, move_lifetime,
cib_conn, options.cib_options, options.promoted_role_only,
PCMK__ROLE_PROMOTED);
} else if (pcmk_is_set(rsc->flags, pcmk_rsc_promotable)) {
int count = 0;
GList *iter = NULL;
current = NULL;
for(iter = rsc->children; iter; iter = iter->next) {
pcmk_resource_t *child = (pcmk_resource_t *)iter->data;
enum rsc_role_e child_role = child->fns->state(child, TRUE);
if (child_role == pcmk_role_promoted) {
count++;
current = pe__current_node(child);
}
}
if(count == 1 && current) {
rc = cli_resource_ban(out, options.rsc_id, current->details->uname, move_lifetime,
cib_conn, options.cib_options, options.promoted_role_only,
PCMK__ROLE_PROMOTED);
} else {
rc = EINVAL;
g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE,
_("Resource '%s' not moved: active in %d locations (promoted in %d).\n"
"To prevent '%s' from running on a specific location, "
"specify a node."
"To prevent '%s' from being promoted at a specific "
"location, specify a node and the --promoted option."),
options.rsc_id, nactive, count, options.rsc_id, options.rsc_id);
}
} else {
rc = EINVAL;
g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE,
_("Resource '%s' not moved: active in %d locations.\n"
"To prevent '%s' from running on a specific location, "
"specify a node."),
options.rsc_id, nactive, options.rsc_id);
}
return rc;
}
static void
cleanup(pcmk__output_t *out, pcmk_resource_t *rsc, pcmk_node_t *node)
{
int rc = pcmk_rc_ok;
if (options.force == FALSE) {
rsc = uber_parent(rsc);
}
crm_debug("Erasing failures of %s (%s requested) on %s",
rsc->id, options.rsc_id, (options.host_uname? options.host_uname: "all nodes"));
rc = cli_resource_delete(controld_api, options.host_uname, rsc,
options.operation, options.interval_spec, TRUE,
scheduler, options.force);
if ((rc == pcmk_rc_ok) && !out->is_quiet(out)) {
// Show any reasons why resource might stay stopped
cli_resource_check(out, rsc, node);
}
if (rc == pcmk_rc_ok) {
start_mainloop(controld_api);
}
}
static int
clear_constraints(pcmk__output_t *out, xmlNodePtr *cib_xml_copy)
{
GList *before = NULL;
GList *after = NULL;
GList *remaining = NULL;
GList *ele = NULL;
pcmk_node_t *dest = NULL;
int rc = pcmk_rc_ok;
if (!out->is_quiet(out)) {
before = build_constraint_list(scheduler->input);
}
if (options.clear_expired) {
rc = cli_resource_clear_all_expired(scheduler->input, cib_conn,
options.cib_options, options.rsc_id,
options.host_uname,
options.promoted_role_only);
} else if (options.host_uname) {
dest = pe_find_node(scheduler->nodes, options.host_uname);
if (dest == NULL) {
rc = pcmk_rc_node_unknown;
if (!out->is_quiet(out)) {
g_list_free(before);
}
return rc;
}
rc = cli_resource_clear(options.rsc_id, dest->details->uname, NULL,
cib_conn, options.cib_options, TRUE, options.force);
} else {
rc = cli_resource_clear(options.rsc_id, NULL, scheduler->nodes,
cib_conn, options.cib_options, TRUE, options.force);
}
if (!out->is_quiet(out)) {
rc = cib_conn->cmds->query(cib_conn, NULL, cib_xml_copy, cib_scope_local | cib_sync_call);
rc = pcmk_legacy2rc(rc);
if (rc != pcmk_rc_ok) {
g_set_error(&error, PCMK__RC_ERROR, rc,
_("Could not get modified CIB: %s\n"), pcmk_rc_str(rc));
g_list_free(before);
free_xml(*cib_xml_copy);
*cib_xml_copy = NULL;
return rc;
}
scheduler->input = *cib_xml_copy;
cluster_status(scheduler);
after = build_constraint_list(scheduler->input);
remaining = pcmk__subtract_lists(before, after, (GCompareFunc) strcmp);
for (ele = remaining; ele != NULL; ele = ele->next) {
out->info(out, "Removing constraint: %s", (char *) ele->data);
}
g_list_free(before);
g_list_free(after);
g_list_free(remaining);
}
return rc;
}
static int
initialize_scheduler_data(xmlNodePtr *cib_xml_copy)
{
int rc = pcmk_rc_ok;
if (options.xml_file != NULL) {
*cib_xml_copy = filename2xml(options.xml_file);
if (*cib_xml_copy == NULL) {
rc = pcmk_rc_cib_corrupt;
}
} else {
rc = cib_conn->cmds->query(cib_conn, NULL, cib_xml_copy, cib_scope_local | cib_sync_call);
rc = pcmk_legacy2rc(rc);
}
if (rc == pcmk_rc_ok) {
scheduler = pe_new_working_set();
if (scheduler == NULL) {
rc = ENOMEM;
} else {
- pe__set_working_set_flags(scheduler,
+ pcmk__set_scheduler_flags(scheduler,
pcmk_sched_no_counts
|pcmk_sched_no_compat);
scheduler->priv = out;
rc = update_scheduler_input(scheduler, cib_xml_copy);
}
}
if (rc != pcmk_rc_ok) {
free_xml(*cib_xml_copy);
*cib_xml_copy = NULL;
return rc;
}
cluster_status(scheduler);
return pcmk_rc_ok;
}
static int
refresh(pcmk__output_t *out)
{
int rc = pcmk_rc_ok;
const char *router_node = options.host_uname;
int attr_options = pcmk__node_attr_none;
if (options.host_uname) {
pcmk_node_t *node = pe_find_node(scheduler->nodes, options.host_uname);
if (pe__is_guest_or_remote_node(node)) {
node = pe__current_node(node->details->remote_rsc);
if (node == NULL) {
rc = ENXIO;
g_set_error(&error, PCMK__RC_ERROR, rc,
_("No cluster connection to Pacemaker Remote node %s detected"),
options.host_uname);
return rc;
}
router_node = node->details->uname;
attr_options |= pcmk__node_attr_remote;
}
}
if (controld_api == NULL) {
out->info(out, "Dry run: skipping clean-up of %s due to CIB_file",
options.host_uname? options.host_uname : "all nodes");
rc = pcmk_rc_ok;
return rc;
}
crm_debug("Re-checking the state of all resources on %s", options.host_uname?options.host_uname:"all nodes");
rc = pcmk__attrd_api_clear_failures(NULL, options.host_uname, NULL,
NULL, NULL, NULL, attr_options);
if (pcmk_controld_api_reprobe(controld_api, options.host_uname,
router_node) == pcmk_rc_ok) {
start_mainloop(controld_api);
}
return rc;
}
static void
refresh_resource(pcmk__output_t *out, pcmk_resource_t *rsc, pcmk_node_t *node)
{
int rc = pcmk_rc_ok;
if (options.force == FALSE) {
rsc = uber_parent(rsc);
}
crm_debug("Re-checking the state of %s (%s requested) on %s",
rsc->id, options.rsc_id, (options.host_uname? options.host_uname: "all nodes"));
rc = cli_resource_delete(controld_api, options.host_uname, rsc, NULL, 0,
FALSE, scheduler, options.force);
if ((rc == pcmk_rc_ok) && !out->is_quiet(out)) {
// Show any reasons why resource might stay stopped
cli_resource_check(out, rsc, node);
}
if (rc == pcmk_rc_ok) {
start_mainloop(controld_api);
}
}
static int
set_property(void)
{
int rc = pcmk_rc_ok;
xmlNode *msg_data = NULL;
if (pcmk__str_empty(options.rsc_type)) {
g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE,
_("Must specify -t with resource type"));
rc = ENXIO;
return rc;
} else if (pcmk__str_empty(options.prop_value)) {
g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE,
_("Must supply -v with new value"));
rc = ENXIO;
return rc;
}
CRM_LOG_ASSERT(options.prop_name != NULL);
msg_data = create_xml_node(NULL, options.rsc_type);
crm_xml_add(msg_data, PCMK_XA_ID, options.rsc_id);
crm_xml_add(msg_data, options.prop_name, options.prop_value);
rc = cib_conn->cmds->modify(cib_conn, XML_CIB_TAG_RESOURCES, msg_data,
options.cib_options);
rc = pcmk_legacy2rc(rc);
free_xml(msg_data);
return rc;
}
static int
show_metadata(pcmk__output_t *out, const char *agent_spec)
{
int rc = pcmk_rc_ok;
char *standard = NULL;
char *provider = NULL;
char *type = NULL;
char *metadata = NULL;
lrmd_t *lrmd_conn = NULL;
rc = lrmd__new(&lrmd_conn, NULL, NULL, 0);
if (rc != pcmk_rc_ok) {
g_set_error(&error, PCMK__RC_ERROR, rc,
_("Could not create executor connection"));
lrmd_api_delete(lrmd_conn);
return rc;
}
rc = crm_parse_agent_spec(agent_spec, &standard, &provider, &type);
rc = pcmk_legacy2rc(rc);
if (rc == pcmk_rc_ok) {
rc = lrmd_conn->cmds->get_metadata(lrmd_conn, standard,
provider, type,
&metadata, 0);
rc = pcmk_legacy2rc(rc);
if (metadata) {
out->output_xml(out, "metadata", metadata);
free(metadata);
} else {
/* We were given a validly formatted spec, but it doesn't necessarily
* match up with anything that exists. Use ENXIO as the return code
* here because that maps to an exit code of CRM_EX_NOSUCH, which
* probably is the most common reason to get here.
*/
rc = ENXIO;
g_set_error(&error, PCMK__RC_ERROR, rc,
_("Metadata query for %s failed: %s"),
agent_spec, pcmk_rc_str(rc));
}
} else {
rc = ENXIO;
g_set_error(&error, PCMK__RC_ERROR, rc,
_("'%s' is not a valid agent specification"), agent_spec);
}
lrmd_api_delete(lrmd_conn);
return rc;
}
static void
validate_cmdline_config(void)
{
// Cannot use both --resource and command-line resource configuration
if (options.rsc_id != NULL) {
g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE,
_("--resource cannot be used with --class, --agent, and --provider"));
// Not all commands support command-line resource configuration
} else if (options.rsc_cmd != cmd_execute_agent) {
g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE,
_("--class, --agent, and --provider can only be used with "
"--validate and --force-*"));
// Not all of --class, --agent, and --provider need to be given. Not all
// classes support the concept of a provider. Check that what we were given
// is valid.
} else if (pcmk__str_eq(options.v_class, "stonith", pcmk__str_none)) {
if (options.v_provider != NULL) {
g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE,
_("stonith does not support providers"));
} else if (stonith_agent_exists(options.v_agent, 0) == FALSE) {
g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE,
_("%s is not a known stonith agent"), options.v_agent ? options.v_agent : "");
}
} else if (resources_agent_exists(options.v_class, options.v_provider, options.v_agent) == FALSE) {
g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE,
_("%s:%s:%s is not a known resource"),
options.v_class ? options.v_class : "",
options.v_provider ? options.v_provider : "",
options.v_agent ? options.v_agent : "");
}
if (error != NULL) {
return;
}
if (options.cmdline_params == NULL) {
options.cmdline_params = pcmk__strkey_table(free, free);
}
options.require_resource = FALSE;
options.require_scheduler = FALSE;
options.require_cib = FALSE;
}
static GOptionContext *
build_arg_context(pcmk__common_args_t *args, GOptionGroup **group) {
GOptionContext *context = NULL;
GOptionEntry extra_prog_entries[] = {
{ "quiet", 'Q', G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE, &(args->quiet),
"Be less descriptive in output.",
NULL },
{ "resource", 'r', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.rsc_id,
"Resource ID",
"ID" },
{ G_OPTION_REMAINING, 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING_ARRAY, &options.remainder,
NULL,
NULL },
{ NULL }
};
const char *description = "Examples:\n\n"
"List the available OCF agents:\n\n"
"\t# crm_resource --list-agents ocf\n\n"
"List the available OCF agents from the linux-ha project:\n\n"
"\t# crm_resource --list-agents ocf:heartbeat\n\n"
"Move 'myResource' to a specific node:\n\n"
"\t# crm_resource --resource myResource --move --node altNode\n\n"
"Allow (but not force) 'myResource' to move back to its original "
"location:\n\n"
"\t# crm_resource --resource myResource --clear\n\n"
"Stop 'myResource' (and anything that depends on it):\n\n"
"\t# crm_resource --resource myResource --set-parameter "
PCMK_META_TARGET_ROLE "--meta --parameter-value Stopped\n\n"
"Tell the cluster not to manage 'myResource' (the cluster will not "
"attempt to start or stop the\n"
"resource under any circumstances; useful when performing maintenance "
"tasks on a resource):\n\n"
"\t# crm_resource --resource myResource --set-parameter "
PCMK_META_IS_MANAGED "--meta --parameter-value false\n\n"
"Erase the operation history of 'myResource' on 'aNode' (the cluster "
"will 'forget' the existing\n"
"resource state, including any errors, and attempt to recover the"
"resource; useful when a resource\n"
"had failed permanently and has been repaired by an administrator):\n\n"
"\t# crm_resource --resource myResource --cleanup --node aNode\n\n";
context = pcmk__build_arg_context(args, "text (default), xml", group, NULL);
g_option_context_set_description(context, description);
/* Add the -Q option, which cannot be part of the globally supported options
* because some tools use that flag for something else.
*/
pcmk__add_main_args(context, extra_prog_entries);
pcmk__add_arg_group(context, "queries", "Queries:",
"Show query help", query_entries);
pcmk__add_arg_group(context, "commands", "Commands:",
"Show command help", command_entries);
pcmk__add_arg_group(context, "locations", "Locations:",
"Show location help", location_entries);
pcmk__add_arg_group(context, "advanced", "Advanced:",
"Show advanced option help", advanced_entries);
pcmk__add_arg_group(context, "additional", "Additional Options:",
"Show additional options", addl_entries);
return context;
}
int
main(int argc, char **argv)
{
xmlNode *cib_xml_copy = NULL;
pcmk_resource_t *rsc = NULL;
pcmk_node_t *node = NULL;
int rc = pcmk_rc_ok;
GOptionGroup *output_group = NULL;
gchar **processed_args = NULL;
GOptionContext *context = NULL;
/*
* Parse command line arguments
*/
args = pcmk__new_common_args(SUMMARY);
processed_args = pcmk__cmdline_preproc(argv, "GHINSTdginpstuvx");
context = build_arg_context(args, &output_group);
pcmk__register_formats(output_group, formats);
if (!g_option_context_parse_strv(context, &processed_args, &error)) {
exit_code = CRM_EX_USAGE;
goto done;
}
pcmk__cli_init_logging("crm_resource", args->verbosity);
rc = pcmk__output_new(&out, args->output_ty, args->output_dest, argv);
if (rc != pcmk_rc_ok) {
exit_code = CRM_EX_ERROR;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code, _("Error creating output format %s: %s"),
args->output_ty, pcmk_rc_str(rc));
goto done;
}
pe__register_messages(out);
crm_resource_register_messages(out);
lrmd__register_messages(out);
pcmk__register_lib_messages(out);
out->quiet = args->quiet;
crm_log_args(argc, argv);
/*
* Validate option combinations
*/
// If the user didn't explicitly specify a command, list resources
if (options.rsc_cmd == cmd_none) {
options.rsc_cmd = cmd_list_resources;
options.require_resource = FALSE;
}
// --expired without --clear/-U doesn't make sense
if (options.clear_expired && (options.rsc_cmd != cmd_clear)) {
exit_code = CRM_EX_USAGE;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code, _("--expired requires --clear or -U"));
goto done;
}
if ((options.remainder != NULL) && (options.override_params != NULL)) {
// Commands that use positional arguments will create override_params
for (gchar **s = options.remainder; *s; s++) {
char *name = calloc(1, strlen(*s));
char *value = calloc(1, strlen(*s));
int rc = sscanf(*s, "%[^=]=%s", name, value);
if (rc == 2) {
g_hash_table_replace(options.override_params, name, value);
} else {
exit_code = CRM_EX_USAGE;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
_("Error parsing '%s' as a name=value pair"),
argv[optind]);
free(value);
free(name);
goto done;
}
}
} else if (options.remainder != NULL) {
gchar **strv = NULL;
gchar *msg = NULL;
int i = 1;
int len = 0;
for (gchar **s = options.remainder; *s; s++) {
len++;
}
CRM_ASSERT(len > 0);
/* Add 1 for the strv[0] string below, and add another 1 for the NULL
* at the end of the array so g_strjoinv knows when to stop.
*/
strv = calloc(len+2, sizeof(char *));
strv[0] = strdup("non-option ARGV-elements:\n");
for (gchar **s = options.remainder; *s; s++) {
strv[i] = crm_strdup_printf("[%d of %d] %s\n", i, len, *s);
i++;
}
strv[i] = NULL;
exit_code = CRM_EX_USAGE;
msg = g_strjoinv("", strv);
g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "%s", msg);
g_free(msg);
/* Don't try to free the last element, which is just NULL. */
for(i = 0; i < len+1; i++) {
free(strv[i]);
}
free(strv);
goto done;
}
if (pcmk__str_eq(args->output_ty, "xml", pcmk__str_none)) {
/* Kind of a hack to display XML lists using a real tag instead of <list>. This just
* saves from having to write custom messages to build the lists around all these things
*/
switch (options.rsc_cmd) {
case cmd_execute_agent:
case cmd_list_resources:
case cmd_query_xml:
case cmd_query_raw_xml:
case cmd_list_active_ops:
case cmd_list_all_ops:
case cmd_colocations:
pcmk__force_args(context, &error, "%s --xml-simple-list --xml-substitute", g_get_prgname());
break;
default:
pcmk__force_args(context, &error, "%s --xml-substitute", g_get_prgname());
break;
}
} else if (pcmk__str_eq(args->output_ty, "text", pcmk__str_null_matches)) {
if ((options.rsc_cmd == cmd_colocations) ||
options.rsc_cmd == cmd_list_resources) {
pcmk__force_args(context, &error, "%s --text-fancy", g_get_prgname());
}
}
if (args->version) {
out->version(out, false);
goto done;
}
if (options.cmdline_config) {
/* A resource configuration was given on the command line. Sanity-check
* the values and set error if they don't make sense.
*/
validate_cmdline_config();
if (error != NULL) {
exit_code = CRM_EX_USAGE;
goto done;
}
} else if (options.cmdline_params != NULL) {
// @COMPAT @TODO error out here when we can break backward compatibility
g_hash_table_destroy(options.cmdline_params);
options.cmdline_params = NULL;
}
if (options.require_resource && (options.rsc_id == NULL)) {
exit_code = CRM_EX_USAGE;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
_("Must supply a resource id with -r"));
goto done;
}
if (options.require_node && (options.host_uname == NULL)) {
exit_code = CRM_EX_USAGE;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
_("Must supply a node name with -N"));
goto done;
}
/*
* Set up necessary connections
*/
if (options.find_flags && options.rsc_id) {
options.require_scheduler = TRUE;
}
// Establish a connection to the CIB if needed
if (options.require_cib) {
cib_conn = cib_new();
if ((cib_conn == NULL) || (cib_conn->cmds == NULL)) {
exit_code = CRM_EX_DISCONNECT;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
_("Could not create CIB connection"));
goto done;
}
rc = cib_conn->cmds->signon(cib_conn, crm_system_name, cib_command);
rc = pcmk_legacy2rc(rc);
if (rc != pcmk_rc_ok) {
exit_code = pcmk_rc2exitc(rc);
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
_("Could not connect to the CIB: %s"), pcmk_rc_str(rc));
goto done;
}
}
// Populate scheduler data from XML file if specified or CIB query otherwise
if (options.require_scheduler) {
rc = initialize_scheduler_data(&cib_xml_copy);
if (rc != pcmk_rc_ok) {
exit_code = pcmk_rc2exitc(rc);
goto done;
}
}
// If command requires that resource exist if specified, find it
if (options.find_flags && options.rsc_id) {
rsc = pe_find_resource_with_flags(scheduler->resources, options.rsc_id,
options.find_flags);
if (rsc == NULL) {
exit_code = CRM_EX_NOSUCH;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
_("Resource '%s' not found"), options.rsc_id);
goto done;
}
/* The --ban, --clear, --move, and --restart commands do not work with
* instances of clone resourcs.
*/
if (strchr(options.rsc_id, ':') != NULL && pe_rsc_is_clone(rsc->parent) &&
(options.rsc_cmd == cmd_ban || options.rsc_cmd == cmd_clear ||
options.rsc_cmd == cmd_move || options.rsc_cmd == cmd_restart)) {
exit_code = CRM_EX_INVALID_PARAM;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
_("Cannot operate on clone resource instance '%s'"), options.rsc_id);
goto done;
}
}
// If user supplied a node name, check whether it exists
if ((options.host_uname != NULL) && (scheduler != NULL)) {
node = pe_find_node(scheduler->nodes, options.host_uname);
if (node == NULL) {
exit_code = CRM_EX_NOSUCH;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
_("Node '%s' not found"), options.host_uname);
goto done;
}
}
// Establish a connection to the controller if needed
if (options.require_crmd) {
rc = pcmk_new_ipc_api(&controld_api, pcmk_ipc_controld);
if (rc != pcmk_rc_ok) {
exit_code = pcmk_rc2exitc(rc);
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
_("Error connecting to the controller: %s"), pcmk_rc_str(rc));
goto done;
}
pcmk_register_ipc_callback(controld_api, controller_event_callback,
NULL);
rc = pcmk__connect_ipc(controld_api, pcmk_ipc_dispatch_main, 5);
if (rc != pcmk_rc_ok) {
exit_code = pcmk_rc2exitc(rc);
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
_("Error connecting to %s: %s"),
pcmk_ipc_name(controld_api, true), pcmk_rc_str(rc));
goto done;
}
}
/*
* Handle requested command
*/
switch (options.rsc_cmd) {
case cmd_list_resources: {
GList *all = NULL;
all = g_list_prepend(all, (gpointer) "*");
rc = out->message(out, "resource-list", scheduler,
pcmk_show_inactive_rscs | pcmk_show_rsc_only | pcmk_show_pending,
true, all, all, false);
g_list_free(all);
if (rc == pcmk_rc_no_output) {
rc = ENXIO;
}
break;
}
case cmd_list_instances:
rc = out->message(out, "resource-names-list", scheduler->resources);
if (rc != pcmk_rc_ok) {
rc = ENXIO;
}
break;
case cmd_list_alternatives:
rc = pcmk__list_alternatives(out, options.agent_spec);
break;
case cmd_list_agents:
rc = pcmk__list_agents(out, options.agent_spec);
break;
case cmd_list_standards:
rc = pcmk__list_standards(out);
break;
case cmd_list_providers:
rc = pcmk__list_providers(out, options.agent_spec);
break;
case cmd_metadata:
rc = show_metadata(out, options.agent_spec);
break;
case cmd_restart:
/* We don't pass scheduler because rsc needs to stay valid for the
* entire lifetime of cli_resource_restart(), but it will reset and
* update the scheduler data multiple times, so it needs to use its
* own copy.
*/
rc = cli_resource_restart(out, rsc, node, options.move_lifetime,
options.timeout_ms, cib_conn,
options.cib_options, options.promoted_role_only,
options.force);
break;
case cmd_wait:
rc = wait_till_stable(out, options.timeout_ms, cib_conn);
break;
case cmd_execute_agent:
if (options.cmdline_config) {
exit_code = cli_resource_execute_from_params(out, NULL,
options.v_class, options.v_provider, options.v_agent,
options.operation, options.cmdline_params,
options.override_params, options.timeout_ms,
args->verbosity, options.force, options.check_level);
} else {
exit_code = cli_resource_execute(rsc, options.rsc_id,
options.operation, options.override_params,
options.timeout_ms, cib_conn, scheduler,
args->verbosity, options.force, options.check_level);
}
goto done;
case cmd_digests:
node = pe_find_node(scheduler->nodes, options.host_uname);
if (node == NULL) {
rc = pcmk_rc_node_unknown;
} else {
rc = pcmk__resource_digests(out, rsc, node,
options.override_params);
}
break;
case cmd_colocations:
rc = out->message(out, "locations-and-colocations", rsc,
options.recursive, (bool) options.force);
break;
case cmd_cts:
rc = pcmk_rc_ok;
g_list_foreach(scheduler->resources, (GFunc) cli_resource_print_cts,
out);
cli_resource_print_cts_constraints(scheduler);
break;
case cmd_fail:
rc = cli_resource_fail(controld_api, options.host_uname,
options.rsc_id, scheduler);
if (rc == pcmk_rc_ok) {
start_mainloop(controld_api);
}
break;
case cmd_list_active_ops:
rc = cli_resource_print_operations(options.rsc_id,
options.host_uname, TRUE,
scheduler);
break;
case cmd_list_all_ops:
rc = cli_resource_print_operations(options.rsc_id,
options.host_uname, FALSE,
scheduler);
break;
case cmd_locate: {
GList *nodes = cli_resource_search(rsc, options.rsc_id, scheduler);
rc = out->message(out, "resource-search-list", nodes, options.rsc_id);
g_list_free_full(nodes, free);
break;
}
case cmd_query_xml:
rc = cli_resource_print(rsc, scheduler, true);
break;
case cmd_query_raw_xml:
rc = cli_resource_print(rsc, scheduler, false);
break;
case cmd_why:
if ((options.host_uname != NULL) && (node == NULL)) {
rc = pcmk_rc_node_unknown;
} else {
rc = out->message(out, "resource-reasons-list",
scheduler->resources, rsc, node);
}
break;
case cmd_clear:
rc = clear_constraints(out, &cib_xml_copy);
break;
case cmd_move:
if (options.host_uname == NULL) {
rc = ban_or_move(out, rsc, options.move_lifetime);
} else {
rc = cli_resource_move(rsc, options.rsc_id, options.host_uname,
options.move_lifetime, cib_conn,
options.cib_options, scheduler,
options.promoted_role_only,
options.force);
}
if (rc == EINVAL) {
exit_code = CRM_EX_USAGE;
goto done;
}
break;
case cmd_ban:
if (options.host_uname == NULL) {
rc = ban_or_move(out, rsc, options.move_lifetime);
} else if (node == NULL) {
rc = pcmk_rc_node_unknown;
} else {
rc = cli_resource_ban(out, options.rsc_id, node->details->uname,
options.move_lifetime, cib_conn,
options.cib_options,
options.promoted_role_only,
PCMK__ROLE_PROMOTED);
}
if (rc == EINVAL) {
exit_code = CRM_EX_USAGE;
goto done;
}
break;
case cmd_get_property:
rc = out->message(out, "property-list", rsc, options.prop_name);
if (rc == pcmk_rc_no_output) {
rc = ENXIO;
}
break;
case cmd_set_property:
rc = set_property();
break;
case cmd_get_param: {
unsigned int count = 0;
GHashTable *params = NULL;
pcmk_node_t *current = rsc->fns->active_node(rsc, &count, NULL);
bool free_params = true;
const char* value = NULL;
if (count > 1) {
out->err(out, "%s is active on more than one node,"
" returning the default value for %s", rsc->id,
pcmk__s(options.prop_name, "unspecified property"));
current = NULL;
}
crm_debug("Looking up %s in %s", options.prop_name, rsc->id);
if (pcmk__str_eq(options.attr_set_type, XML_TAG_ATTR_SETS, pcmk__str_none)) {
params = pe_rsc_params(rsc, current, scheduler);
free_params = false;
value = g_hash_table_lookup(params, options.prop_name);
} else if (pcmk__str_eq(options.attr_set_type, XML_TAG_META_SETS, pcmk__str_none)) {
params = pcmk__strkey_table(free, free);
get_meta_attributes(params, rsc, current, scheduler);
value = g_hash_table_lookup(params, options.prop_name);
} else if (pcmk__str_eq(options.attr_set_type, ATTR_SET_ELEMENT, pcmk__str_none)) {
value = crm_element_value(rsc->xml, options.prop_name);
free_params = false;
} else {
params = pcmk__strkey_table(free, free);
pe__unpack_dataset_nvpairs(rsc->xml, XML_TAG_UTILIZATION, NULL, params,
NULL, FALSE, scheduler);
value = g_hash_table_lookup(params, options.prop_name);
}
rc = out->message(out, "attribute-list", rsc, options.prop_name, value);
if (free_params) {
g_hash_table_destroy(params);
}
break;
}
case cmd_set_param:
if (pcmk__str_empty(options.prop_value)) {
exit_code = CRM_EX_USAGE;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
_("You need to supply a value with the -v option"));
goto done;
}
/* coverity[var_deref_model] False positive */
rc = cli_resource_update_attribute(rsc, options.rsc_id,
options.prop_set,
options.attr_set_type,
options.prop_id,
options.prop_name,
options.prop_value,
options.recursive, cib_conn,
options.cib_options,
options.force);
break;
case cmd_delete_param:
/* coverity[var_deref_model] False positive */
rc = cli_resource_delete_attribute(rsc, options.rsc_id,
options.prop_set,
options.attr_set_type,
options.prop_id,
options.prop_name, cib_conn,
options.cib_options,
options.force);
break;
case cmd_cleanup:
if (rsc == NULL) {
rc = cli_cleanup_all(controld_api, options.host_uname,
options.operation, options.interval_spec,
scheduler);
if (rc == pcmk_rc_ok) {
start_mainloop(controld_api);
}
} else {
cleanup(out, rsc, node);
}
break;
case cmd_refresh:
if (rsc == NULL) {
rc = refresh(out);
} else {
refresh_resource(out, rsc, node);
}
break;
case cmd_delete:
/* rsc_id was already checked for NULL much earlier when validating
* command line arguments.
*/
if (options.rsc_type == NULL) {
// @COMPAT @TODO change this to exit_code = CRM_EX_USAGE
rc = ENXIO;
g_set_error(&error, PCMK__RC_ERROR, rc,
_("You need to specify a resource type with -t"));
} else {
rc = pcmk__resource_delete(cib_conn, options.cib_options,
options.rsc_id, options.rsc_type);
if (rc != pcmk_rc_ok) {
g_set_error(&error, PCMK__RC_ERROR, rc,
_("Could not delete resource %s: %s"),
options.rsc_id, pcmk_rc_str(rc));
}
}
break;
default:
exit_code = CRM_EX_USAGE;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
_("Unimplemented command: %d"), (int) options.rsc_cmd);
goto done;
}
/* Convert rc into an exit code. */
if (rc != pcmk_rc_ok && rc != pcmk_rc_no_output) {
exit_code = pcmk_rc2exitc(rc);
}
/*
* Clean up and exit
*/
done:
/* When we get here, exit_code has been set one of two ways - either at one of
* the spots where there's a "goto done" (which itself could have happened either
* directly or by calling pcmk_rc2exitc), or just up above after any of the break
* statements.
*
* Thus, we can use just exit_code here to decide what to do.
*/
if (exit_code != CRM_EX_OK && exit_code != CRM_EX_USAGE) {
if (error != NULL) {
char *msg = crm_strdup_printf("%s\nError performing operation: %s",
error->message, crm_exit_str(exit_code));
g_clear_error(&error);
g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "%s", msg);
free(msg);
} else {
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
_("Error performing operation: %s"), crm_exit_str(exit_code));
}
}
g_free(options.host_uname);
g_free(options.interval_spec);
g_free(options.move_lifetime);
g_free(options.operation);
g_free(options.prop_id);
free(options.prop_name);
g_free(options.prop_set);
g_free(options.prop_value);
g_free(options.rsc_id);
g_free(options.rsc_type);
free(options.agent_spec);
free(options.v_agent);
free(options.v_class);
free(options.v_provider);
g_free(options.xml_file);
g_strfreev(options.remainder);
if (options.override_params != NULL) {
g_hash_table_destroy(options.override_params);
}
/* options.cmdline_params does not need to be destroyed here. See the
* comments in cli_resource_execute_from_params.
*/
g_strfreev(processed_args);
g_option_context_free(context);
return bye(exit_code);
}
diff --git a/tools/crm_resource_runtime.c b/tools/crm_resource_runtime.c
index e4ec45d47e..64505a755f 100644
--- a/tools/crm_resource_runtime.c
+++ b/tools/crm_resource_runtime.c
@@ -1,2255 +1,2255 @@
/*
* 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 <crm_resource.h>
#include <crm/common/ipc_attrd_internal.h>
#include <crm/common/ipc_controld.h>
#include <crm/common/lists_internal.h>
#include <crm/services_internal.h>
static GList *
build_node_info_list(const pcmk_resource_t *rsc)
{
GList *retval = NULL;
for (const GList *iter = rsc->children; iter != NULL; iter = iter->next) {
const pcmk_resource_t *child = (const pcmk_resource_t *) iter->data;
for (const GList *iter2 = child->running_on;
iter2 != NULL; iter2 = iter2->next) {
const pcmk_node_t *node = (const pcmk_node_t *) iter2->data;
node_info_t *ni = calloc(1, sizeof(node_info_t));
ni->node_name = node->details->uname;
ni->promoted = pcmk_is_set(rsc->flags, pcmk_rsc_promotable) &&
child->fns->state(child, TRUE) == pcmk_role_promoted;
retval = g_list_prepend(retval, ni);
}
}
return retval;
}
GList *
cli_resource_search(pcmk_resource_t *rsc, const char *requested_name,
pcmk_scheduler_t *scheduler)
{
GList *retval = NULL;
const pcmk_resource_t *parent = pe__const_top_resource(rsc, false);
if (pe_rsc_is_clone(rsc)) {
retval = build_node_info_list(rsc);
/* The anonymous clone children's common ID is supplied */
} else if (pe_rsc_is_clone(parent)
&& !pcmk_is_set(rsc->flags, pcmk_rsc_unique)
&& rsc->clone_name
&& pcmk__str_eq(requested_name, rsc->clone_name, pcmk__str_casei)
&& !pcmk__str_eq(requested_name, rsc->id, pcmk__str_casei)) {
retval = build_node_info_list(parent);
} else if (rsc->running_on != NULL) {
for (GList *iter = rsc->running_on; iter != NULL; iter = iter->next) {
pcmk_node_t *node = (pcmk_node_t *) iter->data;
node_info_t *ni = calloc(1, sizeof(node_info_t));
ni->node_name = node->details->uname;
ni->promoted = (rsc->fns->state(rsc, TRUE) == pcmk_role_promoted);
retval = g_list_prepend(retval, ni);
}
}
return retval;
}
// \return Standard Pacemaker return code
static int
find_resource_attr(pcmk__output_t *out, cib_t * the_cib, const char *attr,
const char *rsc, const char *attr_set_type, const char *set_name,
const char *attr_id, const char *attr_name, char **value)
{
int rc = pcmk_rc_ok;
xmlNode *xml_search = NULL;
GString *xpath = NULL;
const char *xpath_base = NULL;
if(value) {
*value = NULL;
}
if(the_cib == NULL) {
return ENOTCONN;
}
xpath_base = pcmk_cib_xpath_for(XML_CIB_TAG_RESOURCES);
if (xpath_base == NULL) {
crm_err(XML_CIB_TAG_RESOURCES " CIB element not known (bug?)");
return ENOMSG;
}
xpath = g_string_sized_new(1024);
pcmk__g_strcat(xpath,
xpath_base, "//*[@" PCMK_XA_ID "=\"", rsc, "\"]", NULL);
if (attr_set_type != NULL) {
pcmk__g_strcat(xpath, "/", attr_set_type, NULL);
if (set_name != NULL) {
pcmk__g_strcat(xpath, "[@" PCMK_XA_ID "=\"", set_name, "\"]",
NULL);
}
}
g_string_append(xpath, "//" XML_CIB_TAG_NVPAIR "[");
if (attr_id != NULL) {
pcmk__g_strcat(xpath, "@" PCMK_XA_ID "=\"", attr_id, "\"", NULL);
}
if (attr_name != NULL) {
if (attr_id != NULL) {
g_string_append(xpath, " and ");
}
pcmk__g_strcat(xpath, "@" XML_NVPAIR_ATTR_NAME "=\"", attr_name, "\"",
NULL);
}
g_string_append_c(xpath, ']');
rc = the_cib->cmds->query(the_cib, (const char *) xpath->str, &xml_search,
cib_sync_call | cib_scope_local | cib_xpath);
rc = pcmk_legacy2rc(rc);
if (rc != pcmk_rc_ok) {
goto done;
}
crm_log_xml_debug(xml_search, "Match");
if (xml_search->children != NULL) {
xmlNode *child = NULL;
rc = ENOTUNIQ;
out->info(out, "Multiple attributes match name=%s", attr_name);
for (child = pcmk__xml_first_child(xml_search); child != NULL;
child = pcmk__xml_next(child)) {
out->info(out, " Value: %s \t(id=%s)",
crm_element_value(child, XML_NVPAIR_ATTR_VALUE), ID(child));
}
out->spacer(out);
} else if(value) {
pcmk__str_update(value, crm_element_value(xml_search, attr));
}
done:
g_string_free(xpath, TRUE);
free_xml(xml_search);
return rc;
}
/* PRIVATE. Use the find_matching_attr_resources instead. */
static void
find_matching_attr_resources_recursive(pcmk__output_t *out,
GList /* <pcmk_resource_t*> */ **result,
pcmk_resource_t *rsc, const char *rsc_id,
const char * attr_set, const char * attr_set_type,
const char * attr_id, const char * attr_name,
cib_t * cib, const char * cmd, int depth)
{
int rc = pcmk_rc_ok;
char *lookup_id = clone_strip(rsc->id);
char *local_attr_id = NULL;
/* visit the children */
for(GList *gIter = rsc->children; gIter; gIter = gIter->next) {
find_matching_attr_resources_recursive(out, result,
(pcmk_resource_t *) gIter->data,
rsc_id, attr_set, attr_set_type,
attr_id, attr_name, cib, cmd, depth+1);
/* do it only once for clones */
if (rsc->variant == pcmk_rsc_variant_clone) {
break;
}
}
rc = find_resource_attr(out, cib, PCMK_XA_ID, lookup_id, attr_set_type,
attr_set, attr_id, attr_name, &local_attr_id);
/* Post-order traversal.
* The root is always on the list and it is the last item. */
if((0 == depth) || (pcmk_rc_ok == rc)) {
/* push the head */
*result = g_list_append(*result, rsc);
}
free(local_attr_id);
free(lookup_id);
}
/* The result is a linearized pre-ordered tree of resources. */
static GList/*<pcmk_resource_t*>*/ *
find_matching_attr_resources(pcmk__output_t *out, pcmk_resource_t *rsc,
const char * rsc_id, const char * attr_set,
const char * attr_set_type, const char * attr_id,
const char * attr_name, cib_t * cib, const char * cmd,
gboolean force)
{
int rc = pcmk_rc_ok;
char *lookup_id = NULL;
char *local_attr_id = NULL;
GList * result = NULL;
/* If --force is used, update only the requested resource (clone or primitive).
* Otherwise, if the primitive has the attribute, use that.
* Otherwise use the clone. */
if(force == TRUE) {
return g_list_append(result, rsc);
}
if ((rsc->parent != NULL)
&& (rsc->parent->variant == pcmk_rsc_variant_clone)) {
int rc = pcmk_rc_ok;
char *local_attr_id = NULL;
rc = find_resource_attr(out, cib, PCMK_XA_ID, rsc_id, attr_set_type,
attr_set, attr_id, attr_name, &local_attr_id);
free(local_attr_id);
if(rc != pcmk_rc_ok) {
rsc = rsc->parent;
out->info(out, "Performing %s of '%s' on '%s', the parent of '%s'",
cmd, attr_name, rsc->id, rsc_id);
}
return g_list_append(result, rsc);
} else if ((rsc->parent == NULL) && (rsc->children != NULL)
&& (rsc->variant == pcmk_rsc_variant_clone)) {
pcmk_resource_t *child = rsc->children->data;
if (child->variant == pcmk_rsc_variant_primitive) {
lookup_id = clone_strip(child->id); /* Could be a cloned group! */
rc = find_resource_attr(out, cib, PCMK_XA_ID, lookup_id,
attr_set_type, attr_set, attr_id, attr_name,
&local_attr_id);
if(rc == pcmk_rc_ok) {
rsc = child;
out->info(out, "A value for '%s' already exists in child '%s', performing %s on that instead of '%s'",
attr_name, lookup_id, cmd, rsc_id);
}
free(local_attr_id);
free(lookup_id);
}
return g_list_append(result, rsc);
}
/* If the resource is a group ==> children inherit the attribute if defined. */
find_matching_attr_resources_recursive(out, &result, rsc, rsc_id, attr_set,
attr_set_type, attr_id, attr_name,
cib, cmd, 0);
return result;
}
// \return Standard Pacemaker return code
int
cli_resource_update_attribute(pcmk_resource_t *rsc, const char *requested_name,
const char *attr_set, const char *attr_set_type,
const char *attr_id, const char *attr_name,
const char *attr_value, gboolean recursive,
cib_t *cib, int cib_options, gboolean force)
{
pcmk__output_t *out = rsc->cluster->priv;
int rc = pcmk_rc_ok;
char *found_attr_id = NULL;
GList/*<pcmk_resource_t*>*/ *resources = NULL;
const char *top_id = pe__const_top_resource(rsc, false)->id;
if ((attr_id == NULL) && !force) {
find_resource_attr(out, cib, PCMK_XA_ID, top_id, NULL, NULL, NULL,
attr_name, NULL);
}
if (pcmk__str_eq(attr_set_type, XML_TAG_ATTR_SETS, pcmk__str_casei)) {
if (!force) {
rc = find_resource_attr(out, cib, PCMK_XA_ID, top_id,
XML_TAG_META_SETS, attr_set, attr_id,
attr_name, &found_attr_id);
if ((rc == pcmk_rc_ok) && !out->is_quiet(out)) {
out->err(out,
"WARNING: There is already a meta attribute "
"for '%s' called '%s' (id=%s)",
top_id, attr_name, found_attr_id);
out->err(out,
" Delete '%s' first or use the force option "
"to override", found_attr_id);
}
free(found_attr_id);
if (rc == pcmk_rc_ok) {
return ENOTUNIQ;
}
}
resources = g_list_append(resources, rsc);
} else if (pcmk__str_eq(attr_set_type, ATTR_SET_ELEMENT, pcmk__str_none)) {
crm_xml_add(rsc->xml, attr_name, attr_value);
CRM_ASSERT(cib != NULL);
rc = cib->cmds->replace(cib, XML_CIB_TAG_RESOURCES, rsc->xml,
cib_options);
rc = pcmk_legacy2rc(rc);
if (rc == pcmk_rc_ok) {
out->info(out, "Set attribute: name=%s value=%s",
attr_name, attr_value);
}
return rc;
} else {
resources = find_matching_attr_resources(out, rsc, requested_name,
attr_set, attr_set_type,
attr_id, attr_name, cib,
"update", force);
}
/* If the user specified attr_set or attr_id, the intent is to modify a
* single resource, which will be the last item in the list.
*/
if ((attr_set != NULL) || (attr_id != NULL)) {
GList *last = g_list_last(resources);
resources = g_list_remove_link(resources, last);
g_list_free(resources);
resources = last;
}
for (GList *iter = resources; iter != NULL; iter = iter->next) {
char *lookup_id = NULL;
char *local_attr_set = NULL;
const char *rsc_attr_id = attr_id;
const char *rsc_attr_set = attr_set;
xmlNode *xml_top = NULL;
xmlNode *xml_obj = NULL;
found_attr_id = NULL;
rsc = (pcmk_resource_t *) iter->data;
lookup_id = clone_strip(rsc->id); /* Could be a cloned group! */
rc = find_resource_attr(out, cib, PCMK_XA_ID, lookup_id, attr_set_type,
attr_set, attr_id, attr_name, &found_attr_id);
switch (rc) {
case pcmk_rc_ok:
crm_debug("Found a match for name=%s: id=%s",
attr_name, found_attr_id);
rsc_attr_id = found_attr_id;
break;
case ENXIO:
if (rsc_attr_set == NULL) {
local_attr_set = crm_strdup_printf("%s-%s", lookup_id,
attr_set_type);
rsc_attr_set = local_attr_set;
}
if (rsc_attr_id == NULL) {
found_attr_id = crm_strdup_printf("%s-%s",
rsc_attr_set, attr_name);
rsc_attr_id = found_attr_id;
}
xml_top = create_xml_node(NULL, (const char *) rsc->xml->name);
crm_xml_add(xml_top, PCMK_XA_ID, lookup_id);
xml_obj = create_xml_node(xml_top, attr_set_type);
crm_xml_add(xml_obj, PCMK_XA_ID, rsc_attr_set);
break;
default:
free(lookup_id);
free(found_attr_id);
g_list_free(resources);
return rc;
}
xml_obj = crm_create_nvpair_xml(xml_obj, rsc_attr_id, attr_name,
attr_value);
if (xml_top == NULL) {
xml_top = xml_obj;
}
crm_log_xml_debug(xml_top, "Update");
rc = cib->cmds->modify(cib, XML_CIB_TAG_RESOURCES, xml_top,
cib_options);
rc = pcmk_legacy2rc(rc);
if (rc == pcmk_rc_ok) {
out->info(out, "Set '%s' option: id=%s%s%s%s%s value=%s",
lookup_id, found_attr_id,
((rsc_attr_set == NULL)? "" : " set="),
pcmk__s(rsc_attr_set, ""),
((attr_name == NULL)? "" : " name="),
pcmk__s(attr_name, ""), attr_value);
}
free_xml(xml_top);
free(lookup_id);
free(found_attr_id);
free(local_attr_set);
if (recursive
&& pcmk__str_eq(attr_set_type, XML_TAG_META_SETS,
pcmk__str_casei)) {
GList *lpc = NULL;
static bool need_init = true;
if (need_init) {
need_init = false;
pcmk__unpack_constraints(rsc->cluster);
pe__clear_resource_flags_on_all(rsc->cluster,
pcmk_rsc_detect_loop);
}
/* We want to set the attribute only on resources explicitly
* colocated with this one, so we use rsc->rsc_cons_lhs directly
* rather than the with_this_colocations() method.
*/
pe__set_resource_flags(rsc, pcmk_rsc_detect_loop);
for (lpc = rsc->rsc_cons_lhs; lpc != NULL; lpc = lpc->next) {
pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data;
crm_debug("Checking %s %d", cons->id, cons->score);
if (!pcmk_is_set(cons->dependent->flags, pcmk_rsc_detect_loop)
&& (cons->score > 0)) {
crm_debug("Setting %s=%s for dependent resource %s",
attr_name, attr_value, cons->dependent->id);
cli_resource_update_attribute(cons->dependent,
cons->dependent->id, NULL,
attr_set_type, NULL,
attr_name, attr_value,
recursive, cib, cib_options,
force);
}
}
}
}
g_list_free(resources);
return rc;
}
// \return Standard Pacemaker return code
int
cli_resource_delete_attribute(pcmk_resource_t *rsc, const char *requested_name,
const char *attr_set, const char *attr_set_type,
const char *attr_id, const char *attr_name,
cib_t *cib, int cib_options, gboolean force)
{
pcmk__output_t *out = rsc->cluster->priv;
int rc = pcmk_rc_ok;
GList/*<pcmk_resource_t*>*/ *resources = NULL;
if ((attr_id == NULL) && !force) {
find_resource_attr(out, cib, PCMK_XA_ID,
pe__const_top_resource(rsc, false)->id, NULL,
NULL, NULL, attr_name, NULL);
}
if (pcmk__str_eq(attr_set_type, XML_TAG_META_SETS, pcmk__str_casei)) {
resources = find_matching_attr_resources(out, rsc, requested_name,
attr_set, attr_set_type,
attr_id, attr_name, cib,
"delete", force);
} else if (pcmk__str_eq(attr_set_type, ATTR_SET_ELEMENT, pcmk__str_none)) {
xml_remove_prop(rsc->xml, attr_name);
CRM_ASSERT(cib != NULL);
rc = cib->cmds->replace(cib, XML_CIB_TAG_RESOURCES, rsc->xml,
cib_options);
rc = pcmk_legacy2rc(rc);
if (rc == pcmk_rc_ok) {
out->info(out, "Deleted attribute: %s", attr_name);
}
return rc;
} else {
resources = g_list_append(resources, rsc);
}
for (GList *iter = resources; iter != NULL; iter = iter->next) {
char *lookup_id = NULL;
xmlNode *xml_obj = NULL;
char *found_attr_id = NULL;
const char *rsc_attr_id = attr_id;
rsc = (pcmk_resource_t *) iter->data;
lookup_id = clone_strip(rsc->id);
rc = find_resource_attr(out, cib, PCMK_XA_ID, lookup_id, attr_set_type,
attr_set, attr_id, attr_name, &found_attr_id);
switch (rc) {
case pcmk_rc_ok:
break;
case ENXIO:
free(lookup_id);
rc = pcmk_rc_ok;
continue;
default:
free(lookup_id);
g_list_free(resources);
return rc;
}
if (rsc_attr_id == NULL) {
rsc_attr_id = found_attr_id;
}
xml_obj = crm_create_nvpair_xml(NULL, rsc_attr_id, attr_name, NULL);
crm_log_xml_debug(xml_obj, "Delete");
CRM_ASSERT(cib);
rc = cib->cmds->remove(cib, XML_CIB_TAG_RESOURCES, xml_obj,
cib_options);
rc = pcmk_legacy2rc(rc);
if (rc == pcmk_rc_ok) {
out->info(out, "Deleted '%s' option: id=%s%s%s%s%s",
lookup_id, found_attr_id,
((attr_set == NULL)? "" : " set="),
pcmk__s(attr_set, ""),
((attr_name == NULL)? "" : " name="),
pcmk__s(attr_name, ""));
}
free(lookup_id);
free_xml(xml_obj);
free(found_attr_id);
}
g_list_free(resources);
return rc;
}
// \return Standard Pacemaker return code
static int
send_lrm_rsc_op(pcmk_ipc_api_t *controld_api, bool do_fail_resource,
const char *host_uname, const char *rsc_id,
pcmk_scheduler_t *scheduler)
{
pcmk__output_t *out = scheduler->priv;
const char *router_node = host_uname;
const char *rsc_api_id = NULL;
const char *rsc_long_id = NULL;
const char *rsc_class = NULL;
const char *rsc_provider = NULL;
const char *rsc_type = NULL;
bool cib_only = false;
pcmk_resource_t *rsc = pe_find_resource(scheduler->resources, rsc_id);
if (rsc == NULL) {
out->err(out, "Resource %s not found", rsc_id);
return ENXIO;
} else if (rsc->variant != pcmk_rsc_variant_primitive) {
out->err(out, "We can only process primitive resources, not %s", rsc_id);
return EINVAL;
}
rsc_class = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS);
rsc_provider = crm_element_value(rsc->xml, XML_AGENT_ATTR_PROVIDER),
rsc_type = crm_element_value(rsc->xml, XML_ATTR_TYPE);
if ((rsc_class == NULL) || (rsc_type == NULL)) {
out->err(out, "Resource %s does not have a class and type", rsc_id);
return EINVAL;
}
{
pcmk_node_t *node = pe_find_node(scheduler->nodes, host_uname);
if (node == NULL) {
out->err(out, "Node %s not found", host_uname);
return pcmk_rc_node_unknown;
}
if (!(node->details->online)) {
if (do_fail_resource) {
out->err(out, "Node %s is not online", host_uname);
return ENOTCONN;
} else {
cib_only = true;
}
}
if (!cib_only && pe__is_guest_or_remote_node(node)) {
node = pe__current_node(node->details->remote_rsc);
if (node == NULL) {
out->err(out, "No cluster connection to Pacemaker Remote node %s detected",
host_uname);
return ENOTCONN;
}
router_node = node->details->uname;
}
}
if (rsc->clone_name) {
rsc_api_id = rsc->clone_name;
rsc_long_id = rsc->id;
} else {
rsc_api_id = rsc->id;
}
if (do_fail_resource) {
return pcmk_controld_api_fail(controld_api, host_uname, router_node,
rsc_api_id, rsc_long_id,
rsc_class, rsc_provider, rsc_type);
} else {
return pcmk_controld_api_refresh(controld_api, host_uname, router_node,
rsc_api_id, rsc_long_id, rsc_class,
rsc_provider, rsc_type, cib_only);
}
}
/*!
* \internal
* \brief Get resource name as used in failure-related node attributes
*
* \param[in] rsc Resource to check
*
* \return Newly allocated string containing resource's fail name
* \note The caller is responsible for freeing the result.
*/
static inline char *
rsc_fail_name(const pcmk_resource_t *rsc)
{
const char *name = (rsc->clone_name? rsc->clone_name : rsc->id);
if (pcmk_is_set(rsc->flags, pcmk_rsc_unique)) {
return strdup(name);
}
return clone_strip(name);
}
// \return Standard Pacemaker return code
static int
clear_rsc_history(pcmk_ipc_api_t *controld_api, const char *host_uname,
const char *rsc_id, pcmk_scheduler_t *scheduler)
{
int rc = pcmk_rc_ok;
/* Erase the resource's entire LRM history in the CIB, even if we're only
* clearing a single operation's fail count. If we erased only entries for a
* single operation, we might wind up with a wrong idea of the current
* resource state, and we might not re-probe the resource.
*/
rc = send_lrm_rsc_op(controld_api, false, host_uname, rsc_id, scheduler);
if (rc != pcmk_rc_ok) {
return rc;
}
crm_trace("Processing %d mainloop inputs",
pcmk_controld_api_replies_expected(controld_api));
while (g_main_context_iteration(NULL, FALSE)) {
crm_trace("Processed mainloop input, %d still remaining",
pcmk_controld_api_replies_expected(controld_api));
}
return rc;
}
// \return Standard Pacemaker return code
static int
clear_rsc_failures(pcmk__output_t *out, pcmk_ipc_api_t *controld_api,
const char *node_name, const char *rsc_id, const char *operation,
const char *interval_spec, pcmk_scheduler_t *scheduler)
{
int rc = pcmk_rc_ok;
const char *failed_value = NULL;
const char *failed_id = NULL;
char *interval_ms_s = NULL;
GHashTable *rscs = NULL;
GHashTableIter iter;
/* Create a hash table to use as a set of resources to clean. This lets us
* clean each resource only once (per node) regardless of how many failed
* operations it has.
*/
rscs = pcmk__strkey_table(NULL, NULL);
// Normalize interval to milliseconds for comparison to history entry
if (operation) {
guint interval_ms = 0U;
pcmk_parse_interval_spec(interval_spec, &interval_ms);
interval_ms_s = crm_strdup_printf("%u", interval_ms);
}
for (xmlNode *xml_op = pcmk__xml_first_child(scheduler->failed);
xml_op != NULL;
xml_op = pcmk__xml_next(xml_op)) {
failed_id = crm_element_value(xml_op, XML_LRM_ATTR_RSCID);
if (failed_id == NULL) {
// Malformed history entry, should never happen
continue;
}
// No resource specified means all resources match
if (rsc_id) {
pcmk_resource_t *fail_rsc = NULL;
fail_rsc = pe_find_resource_with_flags(scheduler->resources,
failed_id,
pcmk_rsc_match_history
|pcmk_rsc_match_anon_basename);
if (!fail_rsc || !pcmk__str_eq(rsc_id, fail_rsc->id, pcmk__str_casei)) {
continue;
}
}
// Host name should always have been provided by this point
failed_value = crm_element_value(xml_op, XML_ATTR_UNAME);
if (!pcmk__str_eq(node_name, failed_value, pcmk__str_casei)) {
continue;
}
// No operation specified means all operations match
if (operation) {
failed_value = crm_element_value(xml_op, XML_LRM_ATTR_TASK);
if (!pcmk__str_eq(operation, failed_value, pcmk__str_casei)) {
continue;
}
// Interval (if operation was specified) defaults to 0 (not all)
failed_value = crm_element_value(xml_op, XML_LRM_ATTR_INTERVAL_MS);
if (!pcmk__str_eq(interval_ms_s, failed_value, pcmk__str_casei)) {
continue;
}
}
g_hash_table_add(rscs, (gpointer) failed_id);
}
free(interval_ms_s);
g_hash_table_iter_init(&iter, rscs);
while (g_hash_table_iter_next(&iter, (gpointer *) &failed_id, NULL)) {
crm_debug("Erasing failures of %s on %s", failed_id, node_name);
rc = clear_rsc_history(controld_api, node_name, failed_id, scheduler);
if (rc != pcmk_rc_ok) {
return rc;
}
}
g_hash_table_destroy(rscs);
return rc;
}
// \return Standard Pacemaker return code
static int
clear_rsc_fail_attrs(const pcmk_resource_t *rsc, const char *operation,
const char *interval_spec, const pcmk_node_t *node)
{
int rc = pcmk_rc_ok;
int attr_options = pcmk__node_attr_none;
char *rsc_name = rsc_fail_name(rsc);
if (pe__is_guest_or_remote_node(node)) {
attr_options |= pcmk__node_attr_remote;
}
rc = pcmk__attrd_api_clear_failures(NULL, node->details->uname, rsc_name,
operation, interval_spec, NULL,
attr_options);
free(rsc_name);
return rc;
}
// \return Standard Pacemaker return code
int
cli_resource_delete(pcmk_ipc_api_t *controld_api, const char *host_uname,
const pcmk_resource_t *rsc, const char *operation,
const char *interval_spec, bool just_failures,
pcmk_scheduler_t *scheduler, gboolean force)
{
pcmk__output_t *out = scheduler->priv;
int rc = pcmk_rc_ok;
pcmk_node_t *node = NULL;
if (rsc == NULL) {
return ENXIO;
} else if (rsc->children) {
for (const GList *lpc = rsc->children; lpc != NULL; lpc = lpc->next) {
const pcmk_resource_t *child = (const pcmk_resource_t *) lpc->data;
rc = cli_resource_delete(controld_api, host_uname, child, operation,
interval_spec, just_failures, scheduler,
force);
if (rc != pcmk_rc_ok) {
return rc;
}
}
return pcmk_rc_ok;
} else if (host_uname == NULL) {
GList *lpc = NULL;
GList *nodes = g_hash_table_get_values(rsc->known_on);
if(nodes == NULL && force) {
nodes = pcmk__copy_node_list(scheduler->nodes, false);
} else if(nodes == NULL && rsc->exclusive_discover) {
GHashTableIter iter;
pcmk_node_t *node = NULL;
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void**)&node)) {
if(node->weight >= 0) {
nodes = g_list_prepend(nodes, node);
}
}
} else if(nodes == NULL) {
nodes = g_hash_table_get_values(rsc->allowed_nodes);
}
for (lpc = nodes; lpc != NULL; lpc = lpc->next) {
node = (pcmk_node_t *) lpc->data;
if (node->details->online) {
rc = cli_resource_delete(controld_api, node->details->uname, rsc,
operation, interval_spec, just_failures,
scheduler, force);
}
if (rc != pcmk_rc_ok) {
g_list_free(nodes);
return rc;
}
}
g_list_free(nodes);
return pcmk_rc_ok;
}
node = pe_find_node(scheduler->nodes, host_uname);
if (node == NULL) {
out->err(out, "Unable to clean up %s because node %s not found",
rsc->id, host_uname);
return ENODEV;
}
if (!node->details->rsc_discovery_enabled) {
out->err(out, "Unable to clean up %s because resource discovery disabled on %s",
rsc->id, host_uname);
return EOPNOTSUPP;
}
if (controld_api == NULL) {
out->err(out, "Dry run: skipping clean-up of %s on %s due to CIB_file",
rsc->id, host_uname);
return pcmk_rc_ok;
}
rc = clear_rsc_fail_attrs(rsc, operation, interval_spec, node);
if (rc != pcmk_rc_ok) {
out->err(out, "Unable to clean up %s failures on %s: %s",
rsc->id, host_uname, pcmk_rc_str(rc));
return rc;
}
if (just_failures) {
rc = clear_rsc_failures(out, controld_api, host_uname, rsc->id, operation,
interval_spec, scheduler);
} else {
rc = clear_rsc_history(controld_api, host_uname, rsc->id, scheduler);
}
if (rc != pcmk_rc_ok) {
out->err(out, "Cleaned %s failures on %s, but unable to clean history: %s",
rsc->id, host_uname, pcmk_rc_str(rc));
} else {
out->info(out, "Cleaned up %s on %s", rsc->id, host_uname);
}
return rc;
}
// \return Standard Pacemaker return code
int
cli_cleanup_all(pcmk_ipc_api_t *controld_api, const char *node_name,
const char *operation, const char *interval_spec,
pcmk_scheduler_t *scheduler)
{
pcmk__output_t *out = scheduler->priv;
int rc = pcmk_rc_ok;
int attr_options = pcmk__node_attr_none;
const char *display_name = node_name? node_name : "all nodes";
if (controld_api == NULL) {
out->info(out, "Dry run: skipping clean-up of %s due to CIB_file",
display_name);
return rc;
}
if (node_name) {
pcmk_node_t *node = pe_find_node(scheduler->nodes, node_name);
if (node == NULL) {
out->err(out, "Unknown node: %s", node_name);
return ENXIO;
}
if (pe__is_guest_or_remote_node(node)) {
attr_options |= pcmk__node_attr_remote;
}
}
rc = pcmk__attrd_api_clear_failures(NULL, node_name, NULL, operation,
interval_spec, NULL, attr_options);
if (rc != pcmk_rc_ok) {
out->err(out, "Unable to clean up all failures on %s: %s",
display_name, pcmk_rc_str(rc));
return rc;
}
if (node_name) {
rc = clear_rsc_failures(out, controld_api, node_name, NULL,
operation, interval_spec, scheduler);
if (rc != pcmk_rc_ok) {
out->err(out, "Cleaned all resource failures on %s, but unable to clean history: %s",
node_name, pcmk_rc_str(rc));
return rc;
}
} else {
for (GList *iter = scheduler->nodes; iter; iter = iter->next) {
pcmk_node_t *node = (pcmk_node_t *) iter->data;
rc = clear_rsc_failures(out, controld_api, node->details->uname, NULL,
operation, interval_spec, scheduler);
if (rc != pcmk_rc_ok) {
out->err(out, "Cleaned all resource failures on all nodes, but unable to clean history: %s",
pcmk_rc_str(rc));
return rc;
}
}
}
out->info(out, "Cleaned up all resources on %s", display_name);
return rc;
}
static void
check_role(resource_checks_t *checks)
{
const char *role_s = g_hash_table_lookup(checks->rsc->meta,
PCMK_META_TARGET_ROLE);
if (role_s == NULL) {
return;
}
switch (text2role(role_s)) {
case pcmk_role_stopped:
checks->flags |= rsc_remain_stopped;
break;
case pcmk_role_unpromoted:
if (pcmk_is_set(pe__const_top_resource(checks->rsc, false)->flags,
pcmk_rsc_promotable)) {
checks->flags |= rsc_unpromotable;
}
break;
default:
break;
}
}
static void
check_managed(resource_checks_t *checks)
{
const char *managed_s = g_hash_table_lookup(checks->rsc->meta,
PCMK_META_IS_MANAGED);
if ((managed_s != NULL) && !crm_is_true(managed_s)) {
checks->flags |= rsc_unmanaged;
}
}
static void
check_locked(resource_checks_t *checks)
{
if (checks->rsc->lock_node != NULL) {
checks->flags |= rsc_locked;
checks->lock_node = checks->rsc->lock_node->details->uname;
}
}
static bool
node_is_unhealthy(pcmk_node_t *node)
{
switch (pe__health_strategy(node->details->data_set)) {
case pcmk__health_strategy_none:
break;
case pcmk__health_strategy_no_red:
if (pe__node_health(node) < 0) {
return true;
}
break;
case pcmk__health_strategy_only_green:
if (pe__node_health(node) <= 0) {
return true;
}
break;
case pcmk__health_strategy_progressive:
case pcmk__health_strategy_custom:
/* @TODO These are finite scores, possibly with rules, and possibly
* combining with other scores, so attributing these as a cause is
* nontrivial.
*/
break;
}
return false;
}
static void
check_node_health(resource_checks_t *checks, pcmk_node_t *node)
{
if (node == NULL) {
GHashTableIter iter;
bool allowed = false;
bool all_nodes_unhealthy = true;
g_hash_table_iter_init(&iter, checks->rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) {
allowed = true;
if (!node_is_unhealthy(node)) {
all_nodes_unhealthy = false;
break;
}
}
if (allowed && all_nodes_unhealthy) {
checks->flags |= rsc_node_health;
}
} else if (node_is_unhealthy(node)) {
checks->flags |= rsc_node_health;
}
}
int
cli_resource_check(pcmk__output_t *out, pcmk_resource_t *rsc, pcmk_node_t *node)
{
resource_checks_t checks = { .rsc = rsc };
check_role(&checks);
check_managed(&checks);
check_locked(&checks);
check_node_health(&checks, node);
return out->message(out, "resource-check-list", &checks);
}
// \return Standard Pacemaker return code
int
cli_resource_fail(pcmk_ipc_api_t *controld_api, const char *host_uname,
const char *rsc_id, pcmk_scheduler_t *scheduler)
{
crm_notice("Failing %s on %s", rsc_id, host_uname);
return send_lrm_rsc_op(controld_api, true, host_uname, rsc_id, scheduler);
}
static GHashTable *
generate_resource_params(pcmk_resource_t *rsc, pcmk_node_t *node,
pcmk_scheduler_t *scheduler)
{
GHashTable *params = NULL;
GHashTable *meta = NULL;
GHashTable *combined = NULL;
GHashTableIter iter;
char *key = NULL;
char *value = NULL;
combined = pcmk__strkey_table(free, free);
params = pe_rsc_params(rsc, node, scheduler);
if (params != NULL) {
g_hash_table_iter_init(&iter, params);
while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) {
g_hash_table_insert(combined, strdup(key), strdup(value));
}
}
meta = pcmk__strkey_table(free, free);
get_meta_attributes(meta, rsc, node, scheduler);
if (meta != NULL) {
g_hash_table_iter_init(&iter, meta);
while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) {
char *crm_name = crm_meta_name(key);
g_hash_table_insert(combined, crm_name, strdup(value));
}
g_hash_table_destroy(meta);
}
return combined;
}
bool resource_is_running_on(pcmk_resource_t *rsc, const char *host)
{
bool found = true;
GList *hIter = NULL;
GList *hosts = NULL;
if (rsc == NULL) {
return false;
}
rsc->fns->location(rsc, &hosts, TRUE);
for (hIter = hosts; host != NULL && hIter != NULL; hIter = hIter->next) {
pcmk_node_t *node = (pcmk_node_t *) hIter->data;
if (pcmk__strcase_any_of(host, node->details->uname, node->details->id, NULL)) {
crm_trace("Resource %s is running on %s\n", rsc->id, host);
goto done;
}
}
if (host != NULL) {
crm_trace("Resource %s is not running on: %s\n", rsc->id, host);
found = false;
} else if(host == NULL && hosts == NULL) {
crm_trace("Resource %s is not running\n", rsc->id);
found = false;
}
done:
g_list_free(hosts);
return found;
}
/*!
* \internal
* \brief Create a list of all resources active on host from a given list
*
* \param[in] host Name of host to check whether resources are active
* \param[in] rsc_list List of resources to check
*
* \return New list of resources from list that are active on host
*/
static GList *
get_active_resources(const char *host, GList *rsc_list)
{
GList *rIter = NULL;
GList *active = NULL;
for (rIter = rsc_list; rIter != NULL; rIter = rIter->next) {
pcmk_resource_t *rsc = (pcmk_resource_t *) rIter->data;
/* Expand groups to their members, because if we're restarting a member
* other than the first, we can't otherwise tell which resources are
* stopping and starting.
*/
if (rsc->variant == pcmk_rsc_variant_group) {
active = g_list_concat(active,
get_active_resources(host, rsc->children));
} else if (resource_is_running_on(rsc, host)) {
active = g_list_append(active, strdup(rsc->id));
}
}
return active;
}
static void dump_list(GList *items, const char *tag)
{
int lpc = 0;
GList *item = NULL;
for (item = items; item != NULL; item = item->next) {
crm_trace("%s[%d]: %s", tag, lpc, (char*)item->data);
lpc++;
}
}
static void display_list(pcmk__output_t *out, GList *items, const char *tag)
{
GList *item = NULL;
for (item = items; item != NULL; item = item->next) {
out->info(out, "%s%s", tag, (const char *)item->data);
}
}
/*!
* \internal
* \brief Upgrade XML to latest schema version and use it as scheduler input
*
* This also updates the scheduler timestamp to the current time.
*
* \param[in,out] scheduler Scheduler data to update
* \param[in,out] xml XML to use as input
*
* \return Standard Pacemaker return code
* \note On success, caller is responsible for freeing memory allocated for
* scheduler->now.
* \todo This follows the example of other callers of cli_config_update()
* and returns ENOKEY ("Required key not available") if that fails,
* but perhaps pcmk_rc_schema_validation would be better in that case.
*/
int
update_scheduler_input(pcmk_scheduler_t *scheduler, xmlNode **xml)
{
if (cli_config_update(xml, NULL, FALSE) == FALSE) {
return ENOKEY;
}
scheduler->input = *xml;
scheduler->now = crm_time_new(NULL);
return pcmk_rc_ok;
}
/*!
* \internal
* \brief Update scheduler XML input based on a CIB query
*
* \param[in] scheduler Scheduler data to initialize
* \param[in] cib Connection to the CIB manager
*
* \return Standard Pacemaker return code
* \note On success, caller is responsible for freeing memory allocated for
* scheduler->input and scheduler->now.
*/
static int
update_scheduler_input_to_cib(pcmk__output_t *out, pcmk_scheduler_t *scheduler,
cib_t *cib)
{
xmlNode *cib_xml_copy = NULL;
int rc = pcmk_rc_ok;
rc = cib->cmds->query(cib, NULL, &cib_xml_copy, cib_scope_local | cib_sync_call);
rc = pcmk_legacy2rc(rc);
if (rc != pcmk_rc_ok) {
out->err(out, "Could not obtain the current CIB: %s (%d)", pcmk_rc_str(rc), rc);
return rc;
}
rc = update_scheduler_input(scheduler, &cib_xml_copy);
if (rc != pcmk_rc_ok) {
out->err(out, "Could not upgrade the current CIB XML");
free_xml(cib_xml_copy);
return rc;
}
return rc;
}
// \return Standard Pacemaker return code
static int
update_dataset(cib_t *cib, pcmk_scheduler_t *scheduler, bool simulate)
{
char *pid = NULL;
char *shadow_file = NULL;
cib_t *shadow_cib = NULL;
int rc = pcmk_rc_ok;
pcmk__output_t *out = scheduler->priv;
pe_reset_working_set(scheduler);
- pe__set_working_set_flags(scheduler,
+ pcmk__set_scheduler_flags(scheduler,
pcmk_sched_no_counts|pcmk_sched_no_compat);
rc = update_scheduler_input_to_cib(out, scheduler, cib);
if (rc != pcmk_rc_ok) {
return rc;
}
if(simulate) {
bool prev_quiet = false;
pid = pcmk__getpid_s();
shadow_cib = cib_shadow_new(pid);
shadow_file = get_shadow_file(pid);
if (shadow_cib == NULL) {
out->err(out, "Could not create shadow cib: '%s'", pid);
rc = ENXIO;
goto done;
}
rc = write_xml_file(scheduler->input, shadow_file, FALSE);
if (rc < 0) {
out->err(out, "Could not populate shadow cib: %s (%d)", pcmk_strerror(rc), rc);
goto done;
}
rc = shadow_cib->cmds->signon(shadow_cib, crm_system_name, cib_command);
rc = pcmk_legacy2rc(rc);
if (rc != pcmk_rc_ok) {
out->err(out, "Could not connect to shadow cib: %s (%d)", pcmk_rc_str(rc), rc);
goto done;
}
pcmk__schedule_actions(scheduler->input,
pcmk_sched_no_counts|pcmk_sched_no_compat,
scheduler);
prev_quiet = out->is_quiet(out);
out->quiet = true;
pcmk__simulate_transition(scheduler, shadow_cib, NULL);
out->quiet = prev_quiet;
rc = update_dataset(shadow_cib, scheduler, false);
} else {
cluster_status(scheduler);
}
done:
// Do not free scheduler->input here, we need rsc->xml to be valid later on
cib_delete(shadow_cib);
free(pid);
if(shadow_file) {
unlink(shadow_file);
free(shadow_file);
}
return rc;
}
/*!
* \internal
* \brief Find the maximum stop timeout of a resource and its children (if any)
*
* \param[in,out] rsc Resource to get timeout for
*
* \return Maximum stop timeout for \p rsc (in milliseconds)
*/
static int
max_rsc_stop_timeout(pcmk_resource_t *rsc)
{
long long result_ll;
int max_delay = 0;
xmlNode *config = NULL;
GHashTable *meta = NULL;
if (rsc == NULL) {
return 0;
}
// If resource is collective, use maximum of its children's stop timeouts
if (rsc->children != NULL) {
for (GList *iter = rsc->children; iter; iter = iter->next) {
pcmk_resource_t *child = iter->data;
int delay = max_rsc_stop_timeout(child);
if (delay > max_delay) {
pcmk__rsc_trace(rsc,
"Maximum stop timeout for %s is now %s "
"due to %s", rsc->id,
pcmk__readable_interval(delay), child->id);
max_delay = delay;
}
}
return max_delay;
}
// Get resource's stop action configuration from CIB
config = pcmk__find_action_config(rsc, PCMK_ACTION_STOP, 0, true);
/* Get configured timeout for stop action (fully evaluated for rules,
* defaults, etc.).
*
* @TODO This currently ignores node (which might matter for rules)
*/
meta = pcmk__unpack_action_meta(rsc, NULL, PCMK_ACTION_STOP, 0, config);
if ((pcmk__scan_ll(g_hash_table_lookup(meta, XML_ATTR_TIMEOUT),
&result_ll, -1LL) == pcmk_rc_ok)
&& (result_ll >= 0) && (result_ll <= INT_MAX)) {
max_delay = (int) result_ll;
}
g_hash_table_destroy(meta);
return max_delay;
}
/*!
* \internal
* \brief Find a reasonable waiting time for stopping any one resource in a list
*
* \param[in,out] scheduler Scheduler data
* \param[in] resources List of names of resources that will be stopped
*
* \return Rough estimate of a reasonable time to wait (in seconds) to stop any
* one resource in \p resources
* \note This estimate is very rough, simply the maximum stop timeout of all
* given resources and their children, plus a small fudge factor. It does
* not account for children that must be stopped in sequence, action
* throttling, or any demotions needed. It checks the stop timeout, even
* if the resources in question are actually being started.
*/
static int
wait_time_estimate(pcmk_scheduler_t *scheduler, const GList *resources)
{
int max_delay = 0;
// Find maximum stop timeout in milliseconds
for (const GList *item = resources; item != NULL; item = item->next) {
pcmk_resource_t *rsc = pe_find_resource(scheduler->resources,
(const char *) item->data);
int delay = max_rsc_stop_timeout(rsc);
if (delay > max_delay) {
pcmk__rsc_trace(rsc,
"Wait time is now %s due to %s",
pcmk__readable_interval(delay), rsc->id);
max_delay = delay;
}
}
return (max_delay / 1000) + 5;
}
#define waiting_for_starts(d, r, h) ((d != NULL) || \
(!resource_is_running_on((r), (h))))
/*!
* \internal
* \brief Restart a resource (on a particular host if requested).
*
* \param[in,out] out Output object
* \param[in,out] rsc The resource to restart
* \param[in] node Node to restart resource on (NULL for all)
* \param[in] move_lifetime If not NULL, how long constraint should
* remain in effect (as ISO 8601 string)
* \param[in] timeout_ms Consider failed if actions do not complete
* in this time (specified in milliseconds,
* but a two-second granularity is actually
* used; if 0, it will be calculated based on
* the resource timeout)
* \param[in,out] cib Connection to the CIB manager
* \param[in] cib_options Group of enum cib_call_options flags to
* use with CIB calls
* \param[in] promoted_role_only If true, limit to promoted instances
* \param[in] force If true, apply only to requested instance
* if part of a collective resource
*
* \return Standard Pacemaker return code (exits on certain failures)
*/
int
cli_resource_restart(pcmk__output_t *out, pcmk_resource_t *rsc,
const pcmk_node_t *node, const char *move_lifetime,
int timeout_ms, cib_t *cib, int cib_options,
gboolean promoted_role_only, gboolean force)
{
int rc = pcmk_rc_ok;
int lpc = 0;
int before = 0;
int step_timeout_s = 0;
int sleep_interval = 2;
int timeout = timeout_ms / 1000;
bool stop_via_ban = false;
char *rsc_id = NULL;
char *lookup_id = NULL;
char *orig_target_role = NULL;
GList *list_delta = NULL;
GList *target_active = NULL;
GList *current_active = NULL;
GList *restart_target_active = NULL;
pcmk_scheduler_t *scheduler = NULL;
pcmk_resource_t *parent = uber_parent(rsc);
bool running = false;
const char *id = rsc->clone_name ? rsc->clone_name : rsc->id;
const char *host = node ? node->details->uname : NULL;
/* If the implicit resource or primitive resource of a bundle is given, operate on the
* bundle itself instead.
*/
if (pe_rsc_is_bundled(rsc)) {
rsc = parent->parent;
}
running = resource_is_running_on(rsc, host);
if (pe_rsc_is_clone(parent) && !running) {
if (pe_rsc_is_unique_clone(parent)) {
lookup_id = strdup(rsc->id);
} else {
lookup_id = clone_strip(rsc->id);
}
rsc = parent->fns->find_rsc(parent, lookup_id, node,
pcmk_rsc_match_basename
|pcmk_rsc_match_current_node);
free(lookup_id);
running = resource_is_running_on(rsc, host);
}
if (!running) {
if (host) {
out->err(out, "%s is not running on %s and so cannot be restarted", id, host);
} else {
out->err(out, "%s is not running anywhere and so cannot be restarted", id);
}
return ENXIO;
}
if (!pcmk_is_set(rsc->flags, pcmk_rsc_managed)) {
out->err(out, "Unmanaged resources cannot be restarted.");
return EAGAIN;
}
rsc_id = strdup(rsc->id);
if (pe_rsc_is_unique_clone(parent)) {
lookup_id = strdup(rsc->id);
} else {
lookup_id = clone_strip(rsc->id);
}
if (host) {
if (pe_rsc_is_clone(rsc) || pe_bundle_replicas(rsc)) {
stop_via_ban = true;
} else if (pe_rsc_is_clone(parent)) {
stop_via_ban = true;
free(lookup_id);
lookup_id = strdup(parent->id);
}
}
/*
grab full cib
determine originally active resources
disable or ban
poll cib and watch for affected resources to get stopped
without --timeout, calculate the stop timeout for each step and wait for that
if we hit --timeout or the service timeout, re-enable or un-ban, report failure and indicate which resources we couldn't take down
if everything stopped, re-enable or un-ban
poll cib and watch for affected resources to get started
without --timeout, calculate the start timeout for each step and wait for that
if we hit --timeout or the service timeout, report (different) failure and indicate which resources we couldn't bring back up
report success
Optimizations:
- use constraints to determine ordered list of affected resources
- Allow a --no-deps option (aka. --force-restart)
*/
scheduler = pe_new_working_set();
if (scheduler == NULL) {
rc = errno;
out->err(out, "Could not allocate scheduler data: %s", pcmk_rc_str(rc));
goto done;
}
scheduler->priv = out;
rc = update_dataset(cib, scheduler, false);
if(rc != pcmk_rc_ok) {
out->err(out, "Could not get new resource list: %s (%d)", pcmk_rc_str(rc), rc);
goto done;
}
restart_target_active = get_active_resources(host, scheduler->resources);
current_active = get_active_resources(host, scheduler->resources);
dump_list(current_active, "Origin");
if (stop_via_ban) {
/* Stop the clone or bundle instance by banning it from the host */
out->quiet = true;
rc = cli_resource_ban(out, lookup_id, host, move_lifetime, cib,
cib_options, promoted_role_only,
PCMK__ROLE_PROMOTED);
} else {
/* Stop the resource by setting PCMK_META_TARGET_ROLE to Stopped.
* Remember any existing PCMK_META_TARGET_ROLE so we can restore it
* later (though it only makes any difference if it's Unpromoted).
*/
find_resource_attr(out, cib, XML_NVPAIR_ATTR_VALUE, lookup_id, NULL, NULL,
NULL, PCMK_META_TARGET_ROLE, &orig_target_role);
rc = cli_resource_update_attribute(rsc, rsc_id, NULL, XML_TAG_META_SETS,
NULL, PCMK_META_TARGET_ROLE,
PCMK_ACTION_STOPPED, FALSE, cib,
cib_options, force);
}
if(rc != pcmk_rc_ok) {
out->err(out, "Could not set " PCMK_META_TARGET_ROLE " for %s: %s (%d)",
rsc_id, pcmk_rc_str(rc), rc);
if (current_active != NULL) {
g_list_free_full(current_active, free);
current_active = NULL;
}
if (restart_target_active != NULL) {
g_list_free_full(restart_target_active, free);
restart_target_active = NULL;
}
goto done;
}
rc = update_dataset(cib, scheduler, true);
if(rc != pcmk_rc_ok) {
out->err(out, "Could not determine which resources would be stopped");
goto failure;
}
target_active = get_active_resources(host, scheduler->resources);
dump_list(target_active, "Target");
list_delta = pcmk__subtract_lists(current_active, target_active, (GCompareFunc) strcmp);
out->info(out, "Waiting for %d resources to stop:", g_list_length(list_delta));
display_list(out, list_delta, " * ");
step_timeout_s = timeout / sleep_interval;
while (list_delta != NULL) {
before = g_list_length(list_delta);
if(timeout_ms == 0) {
step_timeout_s = wait_time_estimate(scheduler, list_delta)
/ sleep_interval;
}
/* We probably don't need the entire step timeout */
for(lpc = 0; (lpc < step_timeout_s) && (list_delta != NULL); lpc++) {
sleep(sleep_interval);
if(timeout) {
timeout -= sleep_interval;
crm_trace("%ds remaining", timeout);
}
rc = update_dataset(cib, scheduler, FALSE);
if(rc != pcmk_rc_ok) {
out->err(out, "Could not determine which resources were stopped");
goto failure;
}
if (current_active != NULL) {
g_list_free_full(current_active, free);
}
current_active = get_active_resources(host, scheduler->resources);
g_list_free(list_delta);
list_delta = pcmk__subtract_lists(current_active, target_active, (GCompareFunc) strcmp);
dump_list(current_active, "Current");
dump_list(list_delta, "Delta");
}
crm_trace("%d (was %d) resources remaining", g_list_length(list_delta), before);
if(before == g_list_length(list_delta)) {
/* aborted during stop phase, print the contents of list_delta */
out->err(out, "Could not complete shutdown of %s, %d resources remaining", rsc_id, g_list_length(list_delta));
display_list(out, list_delta, " * ");
rc = ETIME;
goto failure;
}
}
if (stop_via_ban) {
rc = cli_resource_clear(lookup_id, host, NULL, cib, cib_options, true, force);
} else if (orig_target_role) {
rc = cli_resource_update_attribute(rsc, rsc_id, NULL, XML_TAG_META_SETS,
NULL, PCMK_META_TARGET_ROLE,
orig_target_role, FALSE, cib,
cib_options, force);
free(orig_target_role);
orig_target_role = NULL;
} else {
rc = cli_resource_delete_attribute(rsc, rsc_id, NULL, XML_TAG_META_SETS,
NULL, PCMK_META_TARGET_ROLE, cib,
cib_options, force);
}
if(rc != pcmk_rc_ok) {
out->err(out,
"Could not unset " PCMK_META_TARGET_ROLE " for %s: %s (%d)",
rsc_id, pcmk_rc_str(rc), rc);
goto done;
}
if (target_active != NULL) {
g_list_free_full(target_active, free);
}
target_active = restart_target_active;
list_delta = pcmk__subtract_lists(target_active, current_active, (GCompareFunc) strcmp);
out->info(out, "Waiting for %d resources to start again:", g_list_length(list_delta));
display_list(out, list_delta, " * ");
step_timeout_s = timeout / sleep_interval;
while (waiting_for_starts(list_delta, rsc, host)) {
before = g_list_length(list_delta);
if(timeout_ms == 0) {
step_timeout_s = wait_time_estimate(scheduler, list_delta)
/ sleep_interval;
}
/* We probably don't need the entire step timeout */
for (lpc = 0; (lpc < step_timeout_s) && waiting_for_starts(list_delta, rsc, host); lpc++) {
sleep(sleep_interval);
if(timeout) {
timeout -= sleep_interval;
crm_trace("%ds remaining", timeout);
}
rc = update_dataset(cib, scheduler, false);
if(rc != pcmk_rc_ok) {
out->err(out, "Could not determine which resources were started");
goto failure;
}
/* It's OK if dependent resources moved to a different node,
* so we check active resources on all nodes.
*/
if (current_active != NULL) {
g_list_free_full(current_active, free);
}
current_active = get_active_resources(NULL, scheduler->resources);
g_list_free(list_delta);
list_delta = pcmk__subtract_lists(target_active, current_active, (GCompareFunc) strcmp);
dump_list(current_active, "Current");
dump_list(list_delta, "Delta");
}
if(before == g_list_length(list_delta)) {
/* aborted during start phase, print the contents of list_delta */
out->err(out, "Could not complete restart of %s, %d resources remaining", rsc_id, g_list_length(list_delta));
display_list(out, list_delta, " * ");
rc = ETIME;
goto failure;
}
}
rc = pcmk_rc_ok;
goto done;
failure:
if (stop_via_ban) {
cli_resource_clear(lookup_id, host, NULL, cib, cib_options, true, force);
} else if (orig_target_role) {
cli_resource_update_attribute(rsc, rsc_id, NULL, XML_TAG_META_SETS, NULL,
PCMK_META_TARGET_ROLE, orig_target_role,
FALSE, cib, cib_options, force);
free(orig_target_role);
} else {
cli_resource_delete_attribute(rsc, rsc_id, NULL, XML_TAG_META_SETS,
NULL, PCMK_META_TARGET_ROLE, cib,
cib_options, force);
}
done:
if (list_delta != NULL) {
g_list_free(list_delta);
}
if (current_active != NULL) {
g_list_free_full(current_active, free);
}
if (target_active != NULL && (target_active != restart_target_active)) {
g_list_free_full(target_active, free);
}
if (restart_target_active != NULL) {
g_list_free_full(restart_target_active, free);
}
free(rsc_id);
free(lookup_id);
pe_free_working_set(scheduler);
return rc;
}
static inline bool
action_is_pending(const pcmk_action_t *action)
{
if (pcmk_any_flags_set(action->flags,
pcmk_action_optional|pcmk_action_pseudo)
|| !pcmk_is_set(action->flags, pcmk_action_runnable)
|| pcmk__str_eq(PCMK_ACTION_NOTIFY, action->task, pcmk__str_casei)) {
return false;
}
return true;
}
/*!
* \internal
* \brief Check whether any actions in a list are pending
*
* \param[in] actions List of actions to check
*
* \return true if any actions in the list are pending, otherwise false
*/
static bool
actions_are_pending(const GList *actions)
{
for (const GList *action = actions; action != NULL; action = action->next) {
const pcmk_action_t *a = (const pcmk_action_t *) action->data;
if (action_is_pending(a)) {
crm_notice("Waiting for %s (flags=%#.8x)", a->uuid, a->flags);
return true;
}
}
return false;
}
static void
print_pending_actions(pcmk__output_t *out, GList *actions)
{
GList *action;
out->info(out, "Pending actions:");
for (action = actions; action != NULL; action = action->next) {
pcmk_action_t *a = (pcmk_action_t *) action->data;
if (!action_is_pending(a)) {
continue;
}
if (a->node) {
out->info(out, "\tAction %d: %s\ton %s",
a->id, a->uuid, pe__node_name(a->node));
} else {
out->info(out, "\tAction %d: %s", a->id, a->uuid);
}
}
}
/* For --wait, timeout (in seconds) to use if caller doesn't specify one */
#define WAIT_DEFAULT_TIMEOUT_S (60 * 60)
/* For --wait, how long to sleep between cluster state checks */
#define WAIT_SLEEP_S (2)
/*!
* \internal
* \brief Wait until all pending cluster actions are complete
*
* This waits until either the CIB's transition graph is idle or a timeout is
* reached.
*
* \param[in,out] out Output object
* \param[in] timeout_ms Consider failed if actions do not complete in
* this time (specified in milliseconds, but
* one-second granularity is actually used; if 0, a
* default will be used)
* \param[in,out] cib Connection to the CIB manager
*
* \return Standard Pacemaker return code
*/
int
wait_till_stable(pcmk__output_t *out, int timeout_ms, cib_t * cib)
{
pcmk_scheduler_t *scheduler = NULL;
xmlXPathObjectPtr search;
int rc = pcmk_rc_ok;
bool pending_unknown_state_resources;
int timeout_s = timeout_ms? ((timeout_ms + 999) / 1000) : WAIT_DEFAULT_TIMEOUT_S;
time_t expire_time = time(NULL) + timeout_s;
time_t time_diff;
bool printed_version_warning = out->is_quiet(out); // i.e. don't print if quiet
scheduler = pe_new_working_set();
if (scheduler == NULL) {
return ENOMEM;
}
do {
/* Abort if timeout is reached */
time_diff = expire_time - time(NULL);
if (time_diff > 0) {
crm_info("Waiting up to %lld seconds for cluster actions to complete", (long long) time_diff);
} else {
print_pending_actions(out, scheduler->actions);
pe_free_working_set(scheduler);
return ETIME;
}
if (rc == pcmk_rc_ok) { /* this avoids sleep on first loop iteration */
sleep(WAIT_SLEEP_S);
}
/* Get latest transition graph */
pe_reset_working_set(scheduler);
rc = update_scheduler_input_to_cib(out, scheduler, cib);
if (rc != pcmk_rc_ok) {
pe_free_working_set(scheduler);
return rc;
}
pcmk__schedule_actions(scheduler->input,
pcmk_sched_no_counts|pcmk_sched_no_compat,
scheduler);
if (!printed_version_warning) {
/* If the DC has a different version than the local node, the two
* could come to different conclusions about what actions need to be
* done. Warn the user in this case.
*
* @TODO A possible long-term solution would be to reimplement the
* wait as a new controller operation that would be forwarded to the
* DC. However, that would have potential problems of its own.
*/
const char *dc_version = g_hash_table_lookup(scheduler->config_hash,
PCMK_OPT_DC_VERSION);
if (!pcmk__str_eq(dc_version, PACEMAKER_VERSION "-" BUILD_VERSION, pcmk__str_casei)) {
out->info(out, "warning: wait option may not work properly in "
"mixed-version cluster");
printed_version_warning = true;
}
}
search = xpath_search(scheduler->input, "/cib/status/node_state/lrm/lrm_resources/lrm_resource/"
XML_LRM_TAG_RSC_OP "[@" XML_LRM_ATTR_RC "='193']");
pending_unknown_state_resources = (numXpathResults(search) > 0);
freeXpathObject(search);
} while (actions_are_pending(scheduler->actions) || pending_unknown_state_resources);
pe_free_working_set(scheduler);
return rc;
}
static const char *
get_action(const char *rsc_action) {
const char *action = NULL;
if (pcmk__str_eq(rsc_action, "validate", pcmk__str_casei)) {
action = PCMK_ACTION_VALIDATE_ALL;
} else if (pcmk__str_eq(rsc_action, "force-check", pcmk__str_casei)) {
action = PCMK_ACTION_MONITOR;
} else if (pcmk__strcase_any_of(rsc_action, "force-start", "force-stop",
"force-demote", "force-promote", NULL)) {
action = rsc_action+6;
} else {
action = rsc_action;
}
return action;
}
/*!
* \brief Set up environment variables as expected by resource agents
*
* When the cluster executes resource agents, it adds certain environment
* variables (directly or via resource meta-attributes) expected by some
* resource agents. Add the essential ones that many resource agents expect, so
* the behavior is the same for command-line execution.
*
* \param[in,out] params Resource parameters that will be passed to agent
* \param[in] timeout_ms Action timeout (in milliseconds)
* \param[in] check_level OCF check level
* \param[in] verbosity Verbosity level
*/
static void
set_agent_environment(GHashTable *params, int timeout_ms, int check_level,
int verbosity)
{
g_hash_table_insert(params, strdup("CRM_meta_timeout"),
crm_strdup_printf("%d", timeout_ms));
g_hash_table_insert(params, strdup(PCMK_XA_CRM_FEATURE_SET),
strdup(CRM_FEATURE_SET));
if (check_level >= 0) {
char *level = crm_strdup_printf("%d", check_level);
setenv("OCF_CHECK_LEVEL", level, 1);
free(level);
}
pcmk__set_env_option(PCMK__ENV_DEBUG, ((verbosity > 0)? "1" : "0"), true);
if (verbosity > 1) {
setenv("OCF_TRACE_RA", "1", 1);
}
/* A resource agent using the standard ocf-shellfuncs library will not print
* messages to stderr if it doesn't have a controlling terminal (e.g. if
* crm_resource is called via script or ssh). This forces it to do so.
*/
setenv("OCF_TRACE_FILE", "/dev/stderr", 0);
}
/*!
* \internal
* \brief Apply command-line overrides to resource parameters
*
* \param[in,out] params Parameters to be passed to agent
* \param[in] overrides Parameters to override (or NULL if none)
*/
static void
apply_overrides(GHashTable *params, GHashTable *overrides)
{
if (overrides != NULL) {
GHashTableIter iter;
char *name = NULL;
char *value = NULL;
g_hash_table_iter_init(&iter, overrides);
while (g_hash_table_iter_next(&iter, (gpointer *) &name,
(gpointer *) &value)) {
g_hash_table_replace(params, strdup(name), strdup(value));
}
}
}
crm_exit_t
cli_resource_execute_from_params(pcmk__output_t *out, const char *rsc_name,
const char *rsc_class, const char *rsc_prov,
const char *rsc_type, const char *rsc_action,
GHashTable *params, GHashTable *override_hash,
int timeout_ms, int resource_verbose, gboolean force,
int check_level)
{
const char *class = rsc_class;
const char *action = get_action(rsc_action);
crm_exit_t exit_code = CRM_EX_OK;
svc_action_t *op = NULL;
// If no timeout was provided, use the same default as the cluster
if (timeout_ms == 0) {
timeout_ms = PCMK_DEFAULT_ACTION_TIMEOUT_MS;
}
set_agent_environment(params, timeout_ms, check_level, resource_verbose);
apply_overrides(params, override_hash);
op = services__create_resource_action(rsc_name? rsc_name : "test",
rsc_class, rsc_prov, rsc_type, action,
0, timeout_ms, params, 0);
if (op == NULL) {
out->err(out, "Could not execute %s using %s%s%s:%s: %s",
action, rsc_class, (rsc_prov? ":" : ""),
(rsc_prov? rsc_prov : ""), rsc_type, strerror(ENOMEM));
g_hash_table_destroy(params);
return CRM_EX_OSERR;
}
if (pcmk__str_eq(rsc_class, PCMK_RESOURCE_CLASS_SERVICE, pcmk__str_casei)) {
class = resources_find_service_class(rsc_type);
}
if (!pcmk__strcase_any_of(class, PCMK_RESOURCE_CLASS_OCF,
PCMK_RESOURCE_CLASS_LSB, NULL)) {
services__format_result(op, CRM_EX_UNIMPLEMENT_FEATURE, PCMK_EXEC_ERROR,
"Manual execution of the %s standard is "
"unsupported", pcmk__s(class, "unspecified"));
}
if (op->rc != PCMK_OCF_UNKNOWN) {
exit_code = op->rc;
goto done;
}
services_action_sync(op);
// Map results to OCF codes for consistent reporting to user
{
enum ocf_exitcode ocf_code = services_result2ocf(class, action, op->rc);
// Cast variable instead of function return to keep compilers happy
exit_code = (crm_exit_t) ocf_code;
}
done:
out->message(out, "resource-agent-action", resource_verbose, rsc_class,
rsc_prov, rsc_type, rsc_name, rsc_action, override_hash,
exit_code, op->status, services__exit_reason(op),
op->stdout_data, op->stderr_data);
services_action_free(op);
return exit_code;
}
crm_exit_t
cli_resource_execute(pcmk_resource_t *rsc, const char *requested_name,
const char *rsc_action, GHashTable *override_hash,
int timeout_ms, cib_t *cib, pcmk_scheduler_t *scheduler,
int resource_verbose, gboolean force, int check_level)
{
pcmk__output_t *out = scheduler->priv;
crm_exit_t exit_code = CRM_EX_OK;
const char *rid = NULL;
const char *rtype = NULL;
const char *rprov = NULL;
const char *rclass = NULL;
GHashTable *params = NULL;
if (pcmk__strcase_any_of(rsc_action, "force-start", "force-demote",
"force-promote", NULL)) {
if(pe_rsc_is_clone(rsc)) {
GList *nodes = cli_resource_search(rsc, requested_name, scheduler);
if(nodes != NULL && force == FALSE) {
out->err(out, "It is not safe to %s %s here: the cluster claims it is already active",
rsc_action, rsc->id);
out->err(out,
"Try setting "
PCMK_META_TARGET_ROLE "=" PCMK__ROLE_STOPPED
" first or specifying the force option");
return CRM_EX_UNSAFE;
}
g_list_free_full(nodes, free);
}
}
if(pe_rsc_is_clone(rsc)) {
/* Grab the first child resource in the hope it's not a group */
rsc = rsc->children->data;
}
if (rsc->variant == pcmk_rsc_variant_group) {
out->err(out, "Sorry, the %s option doesn't support group resources", rsc_action);
return CRM_EX_UNIMPLEMENT_FEATURE;
} else if (pe_rsc_is_bundled(rsc)) {
out->err(out, "Sorry, the %s option doesn't support bundled resources", rsc_action);
return CRM_EX_UNIMPLEMENT_FEATURE;
}
rclass = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS);
rprov = crm_element_value(rsc->xml, XML_AGENT_ATTR_PROVIDER);
rtype = crm_element_value(rsc->xml, XML_ATTR_TYPE);
params = generate_resource_params(rsc, NULL /* @TODO use local node */,
scheduler);
if (timeout_ms == 0) {
timeout_ms = pe_get_configured_timeout(rsc, get_action(rsc_action),
scheduler);
}
rid = pe_rsc_is_anon_clone(rsc->parent)? requested_name : rsc->id;
exit_code = cli_resource_execute_from_params(out, rid, rclass, rprov, rtype, rsc_action,
params, override_hash, timeout_ms,
resource_verbose, force, check_level);
return exit_code;
}
// \return Standard Pacemaker return code
int
cli_resource_move(const pcmk_resource_t *rsc, const char *rsc_id,
const char *host_name, const char *move_lifetime, cib_t *cib,
int cib_options, pcmk_scheduler_t *scheduler,
gboolean promoted_role_only, gboolean force)
{
pcmk__output_t *out = scheduler->priv;
int rc = pcmk_rc_ok;
unsigned int count = 0;
pcmk_node_t *current = NULL;
pcmk_node_t *dest = pe_find_node(scheduler->nodes, host_name);
bool cur_is_dest = false;
if (dest == NULL) {
return pcmk_rc_node_unknown;
}
if (promoted_role_only
&& !pcmk_is_set(rsc->flags, pcmk_rsc_promotable)) {
const pcmk_resource_t *p = pe__const_top_resource(rsc, false);
if (pcmk_is_set(p->flags, pcmk_rsc_promotable)) {
out->info(out, "Using parent '%s' for move instead of '%s'.", rsc->id, rsc_id);
rsc_id = p->id;
rsc = p;
} else {
out->info(out, "Ignoring --promoted option: %s is not promotable",
rsc_id);
promoted_role_only = FALSE;
}
}
current = pe__find_active_requires(rsc, &count);
if (pcmk_is_set(rsc->flags, pcmk_rsc_promotable)) {
unsigned int promoted_count = 0;
pcmk_node_t *promoted_node = NULL;
for (const GList *iter = rsc->children; iter; iter = iter->next) {
const pcmk_resource_t *child = (const pcmk_resource_t *) iter->data;
enum rsc_role_e child_role = child->fns->state(child, TRUE);
if (child_role == pcmk_role_promoted) {
rsc = child;
promoted_node = pe__current_node(child);
promoted_count++;
}
}
if (promoted_role_only || (promoted_count != 0)) {
count = promoted_count;
current = promoted_node;
}
}
if (count > 1) {
if (pe_rsc_is_clone(rsc)) {
current = NULL;
} else {
return pcmk_rc_multiple;
}
}
if (current && (current->details == dest->details)) {
cur_is_dest = true;
if (force) {
crm_info("%s is already %s on %s, reinforcing placement with location constraint.",
rsc_id, promoted_role_only?"promoted":"active",
pe__node_name(dest));
} else {
return pcmk_rc_already;
}
}
/* Clear any previous prefer constraints across all nodes. */
cli_resource_clear(rsc_id, NULL, scheduler->nodes, cib, cib_options, false,
force);
/* Clear any previous ban constraints on 'dest'. */
cli_resource_clear(rsc_id, dest->details->uname, scheduler->nodes, cib,
cib_options, TRUE, force);
/* Record an explicit preference for 'dest' */
rc = cli_resource_prefer(out, rsc_id, dest->details->uname, move_lifetime,
cib, cib_options, promoted_role_only,
PCMK__ROLE_PROMOTED);
crm_trace("%s%s now prefers %s%s",
rsc->id, (promoted_role_only? " (promoted)" : ""),
pe__node_name(dest), force?"(forced)":"");
/* only ban the previous location if current location != destination location.
* it is possible to use -M to enforce a location without regard of where the
* resource is currently located */
if (force && !cur_is_dest) {
/* Ban the original location if possible */
if(current) {
(void)cli_resource_ban(out, rsc_id, current->details->uname, move_lifetime,
cib, cib_options, promoted_role_only,
PCMK__ROLE_PROMOTED);
} else if(count > 1) {
out->info(out, "Resource '%s' is currently %s in %d locations. "
"One may now move to %s",
rsc_id, (promoted_role_only? "promoted" : "active"),
count, pe__node_name(dest));
out->info(out, "To prevent '%s' from being %s at a specific location, "
"specify a node.",
rsc_id, (promoted_role_only? "promoted" : "active"));
} else {
crm_trace("Not banning %s from its current location: not active", rsc_id);
}
}
return rc;
}
diff --git a/tools/crm_simulate.c b/tools/crm_simulate.c
index aab4110e60..15f183361f 100644
--- a/tools/crm_simulate.c
+++ b/tools/crm_simulate.c
@@ -1,590 +1,590 @@
/*
* Copyright 2009-2023 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include <crm_internal.h>
#include <stdint.h>
#include <stdio.h>
#include <unistd.h>
#include <stdlib.h>
#include <time.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/cmdline_internal.h>
#include <crm/common/output_internal.h>
#include <crm/common/output.h>
#include <crm/common/util.h>
#include <crm/common/iso8601.h>
#include <crm/pengine/status.h>
#include <crm/pengine/internal.h>
#include <pacemaker-internal.h>
#include <pacemaker.h>
#define SUMMARY "crm_simulate - simulate a Pacemaker cluster's response to events"
struct {
char *dot_file;
char *graph_file;
gchar *input_file;
pcmk_injections_t *injections;
unsigned int flags;
gchar *output_file;
long long repeat;
gboolean store;
gchar *test_dir;
char *use_date;
char *xml_file;
} options = {
.flags = pcmk_sim_show_pending | pcmk_sim_sanitized,
.repeat = 1
};
uint32_t section_opts = 0;
char *temp_shadow = NULL;
crm_exit_t exit_code = CRM_EX_OK;
#define INDENT " "
static pcmk__supported_format_t formats[] = {
PCMK__SUPPORTED_FORMAT_NONE,
PCMK__SUPPORTED_FORMAT_TEXT,
PCMK__SUPPORTED_FORMAT_XML,
{ NULL, NULL, NULL }
};
static gboolean
all_actions_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.flags |= pcmk_sim_all_actions;
return TRUE;
}
static gboolean
attrs_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
section_opts |= pcmk_section_attributes;
return TRUE;
}
static gboolean
failcounts_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
section_opts |= pcmk_section_failcounts | pcmk_section_failures;
return TRUE;
}
static gboolean
in_place_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.store = TRUE;
options.flags |= pcmk_sim_process | pcmk_sim_simulate;
return TRUE;
}
static gboolean
live_check_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
if (options.xml_file) {
free(options.xml_file);
}
options.xml_file = NULL;
options.flags &= ~pcmk_sim_sanitized;
return TRUE;
}
static gboolean
node_down_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.injections->node_down = g_list_append(options.injections->node_down, g_strdup(optarg));
return TRUE;
}
static gboolean
node_fail_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.injections->node_fail = g_list_append(options.injections->node_fail, g_strdup(optarg));
return TRUE;
}
static gboolean
node_up_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
pcmk__simulate_node_config = true;
options.injections->node_up = g_list_append(options.injections->node_up, g_strdup(optarg));
return TRUE;
}
static gboolean
op_fail_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.flags |= pcmk_sim_process | pcmk_sim_simulate;
options.injections->op_fail = g_list_append(options.injections->op_fail, g_strdup(optarg));
return TRUE;
}
static gboolean
op_inject_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.injections->op_inject = g_list_append(options.injections->op_inject, g_strdup(optarg));
return TRUE;
}
static gboolean
pending_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.flags |= pcmk_sim_show_pending;
return TRUE;
}
static gboolean
process_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.flags |= pcmk_sim_process;
return TRUE;
}
static gboolean
quorum_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
pcmk__str_update(&options.injections->quorum, optarg);
return TRUE;
}
static gboolean
save_dotfile_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.flags |= pcmk_sim_process;
pcmk__str_update(&options.dot_file, optarg);
return TRUE;
}
static gboolean
save_graph_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.flags |= pcmk_sim_process;
pcmk__str_update(&options.graph_file, optarg);
return TRUE;
}
static gboolean
show_scores_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.flags |= pcmk_sim_process | pcmk_sim_show_scores;
return TRUE;
}
static gboolean
simulate_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.flags |= pcmk_sim_process | pcmk_sim_simulate;
return TRUE;
}
static gboolean
ticket_activate_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.injections->ticket_activate = g_list_append(options.injections->ticket_activate, g_strdup(optarg));
return TRUE;
}
static gboolean
ticket_grant_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.injections->ticket_grant = g_list_append(options.injections->ticket_grant, g_strdup(optarg));
return TRUE;
}
static gboolean
ticket_revoke_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.injections->ticket_revoke = g_list_append(options.injections->ticket_revoke, g_strdup(optarg));
return TRUE;
}
static gboolean
ticket_standby_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.injections->ticket_standby = g_list_append(options.injections->ticket_standby, g_strdup(optarg));
return TRUE;
}
static gboolean
utilization_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.flags |= pcmk_sim_process | pcmk_sim_show_utilization;
return TRUE;
}
static gboolean
watchdog_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
pcmk__str_update(&options.injections->watchdog, optarg);
return TRUE;
}
static gboolean
xml_file_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
pcmk__str_update(&options.xml_file, optarg);
options.flags |= pcmk_sim_sanitized;
return TRUE;
}
static gboolean
xml_pipe_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
pcmk__str_update(&options.xml_file, "-");
options.flags |= pcmk_sim_sanitized;
return TRUE;
}
static GOptionEntry operation_entries[] = {
{ "run", 'R', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, process_cb,
"Process the supplied input and show what actions the cluster will take in response",
NULL },
{ "simulate", 'S', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, simulate_cb,
"Like --run, but also simulate taking those actions and show the resulting new status",
NULL },
{ "in-place", 'X', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, in_place_cb,
"Like --simulate, but also store the results back to the input file",
NULL },
{ "show-attrs", 'A', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, attrs_cb,
"Show node attributes",
NULL },
{ "show-failcounts", 'c', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, failcounts_cb,
"Show resource fail counts",
NULL },
{ "show-scores", 's', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, show_scores_cb,
"Show allocation scores",
NULL },
{ "show-utilization", 'U', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, utilization_cb,
"Show utilization information",
NULL },
{ "profile", 'P', 0, G_OPTION_ARG_FILENAME, &options.test_dir,
"Process all the XML files in the named directory to create profiling data",
"DIR" },
{ "repeat", 'N', 0, G_OPTION_ARG_INT, &options.repeat,
"With --profile, repeat each test N times and print timings",
"N" },
/* Deprecated */
{ "pending", 'j', G_OPTION_FLAG_NO_ARG|G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_CALLBACK, pending_cb,
"Display pending state if 'record-pending' is enabled",
NULL },
{ NULL }
};
static GOptionEntry synthetic_entries[] = {
{ "node-up", 'u', 0, G_OPTION_ARG_CALLBACK, node_up_cb,
"Simulate bringing a node online",
"NODE" },
{ "node-down", 'd', 0, G_OPTION_ARG_CALLBACK, node_down_cb,
"Simulate taking a node offline",
"NODE" },
{ "node-fail", 'f', 0, G_OPTION_ARG_CALLBACK, node_fail_cb,
"Simulate a node failing",
"NODE" },
{ "op-inject", 'i', 0, G_OPTION_ARG_CALLBACK, op_inject_cb,
"Generate a failure for the cluster to react to in the simulation.\n"
INDENT "See `Operation Specification` help for more information.",
"OPSPEC" },
{ "op-fail", 'F', 0, G_OPTION_ARG_CALLBACK, op_fail_cb,
"If the specified task occurs during the simulation, have it fail with return code ${rc}.\n"
INDENT "The transition will normally stop at the failed action.\n"
INDENT "Save the result with --save-output and re-run with --xml-file.\n"
INDENT "See `Operation Specification` help for more information.",
"OPSPEC" },
{ "set-datetime", 't', 0, G_OPTION_ARG_STRING, &options.use_date,
"Set date/time (ISO 8601 format, see https://en.wikipedia.org/wiki/ISO_8601)",
"DATETIME" },
{ "quorum", 'q', 0, G_OPTION_ARG_CALLBACK, quorum_cb,
"Set to '1' (or 'true') to indicate cluster has quorum",
"QUORUM" },
{ "watchdog", 'w', 0, G_OPTION_ARG_CALLBACK, watchdog_cb,
"Set to '1' (or 'true') to indicate cluster has an active watchdog device",
"DEVICE" },
{ "ticket-grant", 'g', 0, G_OPTION_ARG_CALLBACK, ticket_grant_cb,
"Simulate granting a ticket",
"TICKET" },
{ "ticket-revoke", 'r', 0, G_OPTION_ARG_CALLBACK, ticket_revoke_cb,
"Simulate revoking a ticket",
"TICKET" },
{ "ticket-standby", 'b', 0, G_OPTION_ARG_CALLBACK, ticket_standby_cb,
"Simulate making a ticket standby",
"TICKET" },
{ "ticket-activate", 'e', 0, G_OPTION_ARG_CALLBACK, ticket_activate_cb,
"Simulate activating a ticket",
"TICKET" },
{ NULL }
};
static GOptionEntry artifact_entries[] = {
{ "save-input", 'I', 0, G_OPTION_ARG_FILENAME, &options.input_file,
"Save the input configuration to the named file",
"FILE" },
{ "save-output", 'O', 0, G_OPTION_ARG_FILENAME, &options.output_file,
"Save the output configuration to the named file",
"FILE" },
{ "save-graph", 'G', 0, G_OPTION_ARG_CALLBACK, save_graph_cb,
"Save the transition graph (XML format) to the named file",
"FILE" },
{ "save-dotfile", 'D', 0, G_OPTION_ARG_CALLBACK, save_dotfile_cb,
"Save the transition graph (DOT format) to the named file",
"FILE" },
{ "all-actions", 'a', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, all_actions_cb,
"Display all possible actions in DOT graph (even if not part of transition)",
NULL },
{ NULL }
};
static GOptionEntry source_entries[] = {
{ "live-check", 'L', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, live_check_cb,
"Connect to CIB manager and use the current CIB contents as input",
NULL },
{ "xml-file", 'x', 0, G_OPTION_ARG_CALLBACK, xml_file_cb,
"Retrieve XML from the named file",
"FILE" },
{ "xml-pipe", 'p', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, xml_pipe_cb,
"Retrieve XML from stdin",
NULL },
{ NULL }
};
static int
setup_input(pcmk__output_t *out, const char *input, const char *output,
GError **error)
{
int rc = pcmk_rc_ok;
xmlNode *cib_object = NULL;
char *local_output = NULL;
if (input == NULL) {
/* Use live CIB */
rc = cib__signon_query(out, NULL, &cib_object);
if (rc != pcmk_rc_ok) {
// cib__signon_query() outputs any relevant error
return rc;
}
} else if (pcmk__str_eq(input, "-", pcmk__str_casei)) {
cib_object = filename2xml(NULL);
} else {
cib_object = filename2xml(input);
}
if (pcmk_find_cib_element(cib_object, XML_CIB_TAG_STATUS) == NULL) {
create_xml_node(cib_object, XML_CIB_TAG_STATUS);
}
if (cli_config_update(&cib_object, NULL, FALSE) == FALSE) {
free_xml(cib_object);
return pcmk_rc_transform_failed;
}
if (validate_xml(cib_object, NULL, FALSE) != TRUE) {
free_xml(cib_object);
return pcmk_rc_schema_validation;
}
if (output == NULL) {
char *pid = pcmk__getpid_s();
local_output = get_shadow_file(pid);
temp_shadow = strdup(local_output);
output = local_output;
free(pid);
}
rc = write_xml_file(cib_object, output, FALSE);
free_xml(cib_object);
cib_object = NULL;
if (rc < 0) {
rc = pcmk_legacy2rc(rc);
g_set_error(error, PCMK__EXITC_ERROR, CRM_EX_CANTCREAT,
"Could not create '%s': %s", output, pcmk_rc_str(rc));
return rc;
} else {
setenv("CIB_file", output, 1);
free(local_output);
return pcmk_rc_ok;
}
}
static GOptionContext *
build_arg_context(pcmk__common_args_t *args, GOptionGroup **group) {
GOptionContext *context = NULL;
GOptionEntry extra_prog_entries[] = {
{ "quiet", 'Q', 0, G_OPTION_ARG_NONE, &(args->quiet),
"Display only essential output",
NULL },
{ NULL }
};
const char *description = "Operation Specification:\n\n"
"The OPSPEC in any command line option is of the form\n"
"${resource}_${task}_${interval_in_ms}@${node}=${rc}\n"
"(memcached_monitor_20000@bart.example.com=7, for example).\n"
"${rc} is an OCF return code. For more information on these\n"
"return codes, refer to https://clusterlabs.org/pacemaker/doc/2.1/Pacemaker_Administration/html/agents.html#ocf-return-codes\n\n"
"Examples:\n\n"
"Pretend a recurring monitor action found memcached stopped on node\n"
"fred.example.com and, during recovery, that the memcached stop\n"
"action failed:\n\n"
"\tcrm_simulate -LS --op-inject memcached:0_monitor_20000@bart.example.com=7 "
"--op-fail memcached:0_stop_0@fred.example.com=1 --save-output /tmp/memcached-test.xml\n\n"
"Now see what the reaction to the stop failed would be:\n\n"
"\tcrm_simulate -S --xml-file /tmp/memcached-test.xml\n\n";
context = pcmk__build_arg_context(args, "text (default), xml", group, NULL);
pcmk__add_main_args(context, extra_prog_entries);
g_option_context_set_description(context, description);
pcmk__add_arg_group(context, "operations", "Operations:",
"Show operations options", operation_entries);
pcmk__add_arg_group(context, "synthetic", "Synthetic Cluster Events:",
"Show synthetic cluster event options", synthetic_entries);
pcmk__add_arg_group(context, "artifact", "Artifact Options:",
"Show artifact options", artifact_entries);
pcmk__add_arg_group(context, "source", "Data Source:",
"Show data source options", source_entries);
return context;
}
int
main(int argc, char **argv)
{
int rc = pcmk_rc_ok;
pcmk_scheduler_t *scheduler = NULL;
pcmk__output_t *out = NULL;
GError *error = NULL;
GOptionGroup *output_group = NULL;
pcmk__common_args_t *args = pcmk__new_common_args(SUMMARY);
gchar **processed_args = pcmk__cmdline_preproc(argv, "bdefgiqrtuwxDFGINOP");
GOptionContext *context = build_arg_context(args, &output_group);
options.injections = calloc(1, sizeof(pcmk_injections_t));
if (options.injections == NULL) {
rc = ENOMEM;
goto done;
}
/* This must come before g_option_context_parse_strv. */
options.xml_file = strdup("-");
pcmk__register_formats(output_group, formats);
if (!g_option_context_parse_strv(context, &processed_args, &error)) {
exit_code = CRM_EX_USAGE;
goto done;
}
pcmk__cli_init_logging("crm_simulate", args->verbosity);
rc = pcmk__output_new(&out, args->output_ty, args->output_dest, argv);
if (rc != pcmk_rc_ok) {
fprintf(stderr, "Error creating output format %s: %s\n",
args->output_ty, pcmk_rc_str(rc));
exit_code = CRM_EX_ERROR;
goto done;
}
if (pcmk__str_eq(args->output_ty, "text", pcmk__str_null_matches) &&
!pcmk_is_set(options.flags, pcmk_sim_show_scores) &&
!pcmk_is_set(options.flags, pcmk_sim_show_utilization)) {
pcmk__force_args(context, &error, "%s --text-fancy", g_get_prgname());
} else if (pcmk__str_eq(args->output_ty, "xml", pcmk__str_none)) {
pcmk__force_args(context, &error, "%s --xml-simple-list --xml-substitute", g_get_prgname());
}
pe__register_messages(out);
pcmk__register_lib_messages(out);
out->quiet = args->quiet;
if (args->version) {
out->version(out, false);
goto done;
}
if (args->verbosity > 0) {
options.flags |= pcmk_sim_verbose;
#ifdef PCMK__COMPAT_2_0
/* Redirect stderr to stdout so we can grep the output */
close(STDERR_FILENO);
dup2(STDOUT_FILENO, STDERR_FILENO);
#endif
}
scheduler = pe_new_working_set();
if (scheduler == NULL) {
rc = ENOMEM;
g_set_error(&error, PCMK__RC_ERROR, rc,
"Could not allocate scheduler data");
goto done;
}
if (pcmk_is_set(options.flags, pcmk_sim_show_scores)) {
- pe__set_working_set_flags(scheduler, pcmk_sched_output_scores);
+ pcmk__set_scheduler_flags(scheduler, pcmk_sched_output_scores);
}
if (pcmk_is_set(options.flags, pcmk_sim_show_utilization)) {
- pe__set_working_set_flags(scheduler, pcmk_sched_show_utilization);
+ pcmk__set_scheduler_flags(scheduler, pcmk_sched_show_utilization);
}
- pe__set_working_set_flags(scheduler, pcmk_sched_no_compat);
+ pcmk__set_scheduler_flags(scheduler, pcmk_sched_no_compat);
if (options.test_dir != NULL) {
scheduler->priv = out;
pcmk__profile_dir(options.test_dir, options.repeat, scheduler,
options.use_date);
rc = pcmk_rc_ok;
goto done;
}
rc = setup_input(out, options.xml_file,
options.store? options.xml_file : options.output_file,
&error);
if (rc != pcmk_rc_ok) {
goto done;
}
rc = pcmk__simulate(scheduler, out, options.injections, options.flags,
section_opts, options.use_date, options.input_file,
options.graph_file, options.dot_file);
done:
pcmk__output_and_clear_error(&error, NULL);
/* There sure is a lot to free in options. */
free(options.dot_file);
free(options.graph_file);
g_free(options.input_file);
g_free(options.output_file);
g_free(options.test_dir);
free(options.use_date);
free(options.xml_file);
pcmk_free_injections(options.injections);
pcmk__free_arg_context(context);
g_strfreev(processed_args);
if (scheduler != NULL) {
pe_free_working_set(scheduler);
}
fflush(stderr);
if (temp_shadow) {
unlink(temp_shadow);
free(temp_shadow);
}
if (rc != pcmk_rc_ok) {
exit_code = pcmk_rc2exitc(rc);
}
if (out != NULL) {
out->finish(out, exit_code, true, NULL);
pcmk__output_free(out);
}
pcmk__unregister_formats();
crm_exit(exit_code);
}
diff --git a/tools/crm_ticket.c b/tools/crm_ticket.c
index 4adc5f676d..61255a66c0 100644
--- a/tools/crm_ticket.c
+++ b/tools/crm_ticket.c
@@ -1,1011 +1,1011 @@
/*
* Copyright 2012-2024 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 <crm/crm.h>
#include <stdio.h>
#include <sys/types.h>
#include <unistd.h>
#include <stdlib.h>
#include <errno.h>
#include <fcntl.h>
#include <libgen.h>
#include <crm/msg_xml.h>
#include <crm/common/xml.h>
#include <crm/common/ipc.h>
#include <crm/common/cmdline_internal.h>
#include <crm/cib.h>
#include <crm/cib/internal.h>
#include <crm/pengine/rules.h>
#include <crm/pengine/status.h>
#include <crm/pengine/internal.h>
#include <pacemaker-internal.h>
GError *error = NULL;
#define SUMMARY "Perform tasks related to cluster tickets\n\n" \
"Allows ticket attributes to be queried, modified and deleted."
struct {
gchar *attr_default;
gchar *attr_id;
char *attr_name;
char *attr_value;
gboolean force;
char *get_attr_name;
gboolean quiet;
gchar *set_name;
char ticket_cmd;
gchar *ticket_id;
gchar *xml_file;
} options = {
.ticket_cmd = 'S'
};
GList *attr_delete;
GHashTable *attr_set;
bool modified = false;
int cib_options = cib_sync_call;
#define INDENT " "
static gboolean
attr_value_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
pcmk__str_update(&options.attr_value, optarg);
if (!options.attr_name || !options.attr_value) {
return TRUE;
}
g_hash_table_insert(attr_set, strdup(options.attr_name), strdup(options.attr_value));
pcmk__str_update(&options.attr_name, NULL);
pcmk__str_update(&options.attr_value, NULL);
modified = true;
return TRUE;
}
static gboolean
command_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
if (pcmk__str_any_of(option_name, "--info", "-l", NULL)) {
options.ticket_cmd = 'l';
} else if (pcmk__str_any_of(option_name, "--details", "-L", NULL)) {
options.ticket_cmd = 'L';
} else if (pcmk__str_any_of(option_name, "--raw", "-w", NULL)) {
options.ticket_cmd = 'w';
} else if (pcmk__str_any_of(option_name, "--query-xml", "-q", NULL)) {
options.ticket_cmd = 'q';
} else if (pcmk__str_any_of(option_name, "--constraints", "-c", NULL)) {
options.ticket_cmd = 'c';
} else if (pcmk__str_any_of(option_name, "--cleanup", "-C", NULL)) {
options.ticket_cmd = 'C';
}
return TRUE;
}
static gboolean
delete_attr_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
attr_delete = g_list_append(attr_delete, strdup(optarg));
modified = true;
return TRUE;
}
static gboolean
get_attr_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
pcmk__str_update(&options.get_attr_name, optarg);
options.ticket_cmd = 'G';
return TRUE;
}
static gboolean
grant_standby_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
if (pcmk__str_any_of(option_name, "--grant", "-g", NULL)) {
g_hash_table_insert(attr_set, strdup("granted"), strdup("true"));
modified = true;
} else if (pcmk__str_any_of(option_name, "--revoke", "-r", NULL)) {
g_hash_table_insert(attr_set, strdup("granted"), strdup("false"));
modified = true;
} else if (pcmk__str_any_of(option_name, "--standby", "-s", NULL)) {
g_hash_table_insert(attr_set, strdup("standby"), strdup("true"));
modified = true;
} else if (pcmk__str_any_of(option_name, "--activate", "-a", NULL)) {
g_hash_table_insert(attr_set, strdup("standby"), strdup("false"));
modified = true;
}
return TRUE;
}
static gboolean
set_attr_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
pcmk__str_update(&options.attr_name, optarg);
if (!options.attr_name || !options.attr_value) {
return TRUE;
}
g_hash_table_insert(attr_set, strdup(options.attr_name), strdup(options.attr_value));
pcmk__str_update(&options.attr_name, NULL);
pcmk__str_update(&options.attr_value, NULL);
modified = true;
return TRUE;
}
static GOptionEntry query_entries[] = {
{ "info", 'l', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb,
"Display the information of ticket(s)",
NULL },
{ "details", 'L', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb,
"Display the details of ticket(s)",
NULL },
{ "raw", 'w', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb,
"Display the IDs of ticket(s)",
NULL },
{ "query-xml", 'q', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb,
"Query the XML of ticket(s)",
NULL },
{ "constraints", 'c', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb,
"Display the rsc_ticket constraints that apply to ticket(s)",
NULL },
{ NULL }
};
static GOptionEntry command_entries[] = {
{ "grant", 'g', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, grant_standby_cb,
"Grant a ticket to this cluster site",
NULL },
{ "revoke", 'r', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, grant_standby_cb,
"Revoke a ticket from this cluster site",
NULL },
{ "standby", 's', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, grant_standby_cb,
"Tell this cluster site this ticket is standby",
NULL },
{ "activate", 'a', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, grant_standby_cb,
"Tell this cluster site this ticket is active",
NULL },
{ NULL }
};
static GOptionEntry advanced_entries[] = {
{ "get-attr", 'G', 0, G_OPTION_ARG_CALLBACK, get_attr_cb,
"Display the named attribute for a ticket",
"ATTRIBUTE" },
{ "set-attr", 'S', 0, G_OPTION_ARG_CALLBACK, set_attr_cb,
"Set the named attribute for a ticket",
"ATTRIBUTE" },
{ "delete-attr", 'D', 0, G_OPTION_ARG_CALLBACK, delete_attr_cb,
"Delete the named attribute for a ticket",
"ATTRIBUTE" },
{ "cleanup", 'C', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb,
"Delete all state of a ticket at this cluster site",
NULL },
{ NULL}
};
static GOptionEntry addl_entries[] = {
{ "attr-value", 'v', 0, G_OPTION_ARG_CALLBACK, attr_value_cb,
"Attribute value to use with -S",
"VALUE" },
{ "default", 'd', 0, G_OPTION_ARG_STRING, &options.attr_default,
"(Advanced) Default attribute value to display if none is found\n"
INDENT "(for use with -G)",
"VALUE" },
{ "force", 'f', 0, G_OPTION_ARG_NONE, &options.force,
"(Advanced) Force the action to be performed",
NULL },
{ "ticket", 't', 0, G_OPTION_ARG_STRING, &options.ticket_id,
"Ticket ID",
"ID" },
{ "xml-file", 'x', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_STRING, &options.xml_file,
NULL,
NULL },
{ NULL }
};
static GOptionEntry deprecated_entries[] = {
{ "set-name", 'n', 0, G_OPTION_ARG_STRING, &options.set_name,
"(Advanced) ID of the instance_attributes object to change",
"ID" },
{ "nvpair", 'i', 0, G_OPTION_ARG_STRING, &options.attr_id,
"(Advanced) ID of the nvpair object to change/delete",
"ID" },
{ "quiet", 'Q', 0, G_OPTION_ARG_NONE, &options.quiet,
"Print only the value on stdout",
NULL },
{ NULL }
};
static pcmk_ticket_t *
find_ticket(gchar *ticket_id, pcmk_scheduler_t *scheduler)
{
return g_hash_table_lookup(scheduler->tickets, ticket_id);
}
static void
print_date(time_t time)
{
int lpc = 0;
char date_str[26];
asctime_r(localtime(&time), date_str);
for (; lpc < 26; lpc++) {
if (date_str[lpc] == '\n') {
date_str[lpc] = 0;
}
}
fprintf(stdout, "'%s'", date_str);
}
static void
print_ticket(pcmk_ticket_t *ticket, bool raw, bool details)
{
if (raw) {
fprintf(stdout, "%s\n", ticket->id);
return;
}
fprintf(stdout, "%s\t%s %s",
ticket->id, ticket->granted ? "granted" : "revoked",
ticket->standby ? "[standby]" : " ");
if (details && g_hash_table_size(ticket->state) > 0) {
GHashTableIter iter;
const char *name = NULL;
const char *value = NULL;
int lpc = 0;
fprintf(stdout, " (");
g_hash_table_iter_init(&iter, ticket->state);
while (g_hash_table_iter_next(&iter, (void **)&name, (void **)&value)) {
if (lpc > 0) {
fprintf(stdout, ", ");
}
fprintf(stdout, "%s=", name);
if (pcmk__str_any_of(name, "last-granted", "expires", NULL)) {
long long time_ll;
pcmk__scan_ll(value, &time_ll, 0);
print_date((time_t) time_ll);
} else {
fprintf(stdout, "%s", value);
}
lpc++;
}
fprintf(stdout, ")\n");
} else {
if (ticket->last_granted > -1) {
fprintf(stdout, " last-granted=");
print_date(ticket->last_granted);
}
fprintf(stdout, "\n");
}
return;
}
static void
print_ticket_list(pcmk_scheduler_t *scheduler, bool raw, bool details)
{
GHashTableIter iter;
pcmk_ticket_t *ticket = NULL;
g_hash_table_iter_init(&iter, scheduler->tickets);
while (g_hash_table_iter_next(&iter, NULL, (void **)&ticket)) {
print_ticket(ticket, raw, details);
}
}
static int
find_ticket_state(cib_t * the_cib, gchar *ticket_id, xmlNode ** ticket_state_xml)
{
int rc = pcmk_rc_ok;
xmlNode *xml_search = NULL;
GString *xpath = NULL;
CRM_ASSERT(ticket_state_xml != NULL);
*ticket_state_xml = NULL;
xpath = g_string_sized_new(1024);
g_string_append(xpath,
"/" XML_TAG_CIB "/" XML_CIB_TAG_STATUS
"/" XML_CIB_TAG_TICKETS);
if (ticket_id != NULL) {
pcmk__g_strcat(xpath,
"/" XML_CIB_TAG_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);
rc = pcmk_legacy2rc(rc);
g_string_free(xpath, TRUE);
if (rc != pcmk_rc_ok) {
return rc;
}
crm_log_xml_debug(xml_search, "Match");
if (xml_search->children != NULL) {
if (ticket_id) {
fprintf(stdout, "Multiple ticket_states match ticket_id=%s\n", ticket_id);
}
*ticket_state_xml = xml_search;
} else {
*ticket_state_xml = xml_search;
}
return rc;
}
static int
find_ticket_constraints(cib_t * the_cib, gchar *ticket_id, xmlNode ** ticket_cons_xml)
{
int rc = pcmk_rc_ok;
xmlNode *xml_search = NULL;
GString *xpath = NULL;
const char *xpath_base = NULL;
CRM_ASSERT(ticket_cons_xml != NULL);
*ticket_cons_xml = NULL;
xpath_base = pcmk_cib_xpath_for(XML_CIB_TAG_CONSTRAINTS);
if (xpath_base == NULL) {
crm_err(XML_CIB_TAG_CONSTRAINTS " CIB element not known (bug?)");
return -ENOMSG;
}
xpath = g_string_sized_new(1024);
pcmk__g_strcat(xpath, xpath_base, "/" XML_CONS_TAG_RSC_TICKET, NULL);
if (ticket_id != NULL) {
pcmk__g_strcat(xpath,
"[@" XML_TICKET_ATTR_TICKET "=\"", ticket_id, "\"]",
NULL);
}
rc = the_cib->cmds->query(the_cib, (const char *) xpath->str, &xml_search,
cib_sync_call | cib_scope_local | cib_xpath);
rc = pcmk_legacy2rc(rc);
g_string_free(xpath, TRUE);
if (rc != pcmk_rc_ok) {
return rc;
}
crm_log_xml_debug(xml_search, "Match");
*ticket_cons_xml = xml_search;
return rc;
}
static int
dump_ticket_xml(cib_t * the_cib, gchar *ticket_id)
{
int rc = pcmk_rc_ok;
xmlNode *state_xml = NULL;
rc = find_ticket_state(the_cib, ticket_id, &state_xml);
if (state_xml == NULL) {
return rc;
}
fprintf(stdout, "State XML:\n");
if (state_xml) {
char *state_xml_str = NULL;
state_xml_str = dump_xml_formatted(state_xml);
fprintf(stdout, "\n%s", state_xml_str);
free_xml(state_xml);
free(state_xml_str);
}
return rc;
}
static int
dump_constraints(cib_t * the_cib, gchar *ticket_id)
{
int rc = pcmk_rc_ok;
xmlNode *cons_xml = NULL;
char *cons_xml_str = NULL;
rc = find_ticket_constraints(the_cib, ticket_id, &cons_xml);
if (cons_xml == NULL) {
return rc;
}
cons_xml_str = dump_xml_formatted(cons_xml);
fprintf(stdout, "Constraints XML:\n\n%s", cons_xml_str);
free_xml(cons_xml);
free(cons_xml_str);
return rc;
}
static int
get_ticket_state_attr(gchar *ticket_id, const char *attr_name, const char **attr_value,
pcmk_scheduler_t *scheduler)
{
pcmk_ticket_t *ticket = NULL;
CRM_ASSERT(attr_value != NULL);
*attr_value = NULL;
ticket = g_hash_table_lookup(scheduler->tickets, ticket_id);
if (ticket == NULL) {
return ENXIO;
}
*attr_value = g_hash_table_lookup(ticket->state, attr_name);
if (*attr_value == NULL) {
return ENXIO;
}
return pcmk_rc_ok;
}
static void
ticket_warning(gchar *ticket_id, const char *action)
{
GString *warning = g_string_sized_new(1024);
const char *word = NULL;
CRM_ASSERT(action != NULL);
if (strcmp(action, "grant") == 0) {
pcmk__g_strcat(warning,
"This command cannot help you verify whether '",
ticket_id,
"' has been already granted elsewhere.\n", NULL);
word = "to";
} else {
pcmk__g_strcat(warning,
"Revoking '", ticket_id, "' can trigger the specified "
"'loss-policy'(s) relating to '", ticket_id, "'.\n\n"
"You can check that with:\n"
"crm_ticket --ticket ", ticket_id, " --constraints\n\n"
"Otherwise before revoking '", ticket_id, "', "
"you may want to make '", ticket_id, "' "
"standby with:\n"
"crm_ticket --ticket ", ticket_id, " --standby\n\n",
NULL);
word = "from";
}
pcmk__g_strcat(warning,
"If you really want to ", action, " '", ticket_id, "' ",
word, " this site now, and you know what you are doing,\n"
"please specify --force.", NULL);
fprintf(stdout, "%s\n", (const char *) warning->str);
g_string_free(warning, TRUE);
}
static bool
allow_modification(gchar *ticket_id)
{
const char *value = NULL;
GList *list_iter = NULL;
if (options.force) {
return true;
}
if (g_hash_table_lookup_extended(attr_set, "granted", NULL, (gpointer *) & value)) {
if (crm_is_true(value)) {
ticket_warning(ticket_id, "grant");
return false;
} else {
ticket_warning(ticket_id, "revoke");
return false;
}
}
for(list_iter = attr_delete; list_iter; list_iter = list_iter->next) {
const char *key = (const char *)list_iter->data;
if (pcmk__str_eq(key, "granted", pcmk__str_casei)) {
ticket_warning(ticket_id, "revoke");
return false;
}
}
return true;
}
static int
modify_ticket_state(gchar *ticket_id, cib_t *cib, pcmk_scheduler_t *scheduler)
{
int rc = pcmk_rc_ok;
xmlNode *xml_top = NULL;
xmlNode *ticket_state_xml = NULL;
bool found = false;
GList *list_iter = NULL;
GHashTableIter hash_iter;
char *key = NULL;
char *value = NULL;
pcmk_ticket_t *ticket = NULL;
rc = find_ticket_state(cib, ticket_id, &ticket_state_xml);
if (rc == pcmk_rc_ok) {
crm_debug("Found a match state for ticket: id=%s", ticket_id);
xml_top = ticket_state_xml;
found = true;
} else if (rc != ENXIO) {
return rc;
} else if (g_hash_table_size(attr_set) == 0){
return pcmk_rc_ok;
} else {
xmlNode *xml_obj = NULL;
xml_top = create_xml_node(NULL, XML_CIB_TAG_STATUS);
xml_obj = create_xml_node(xml_top, XML_CIB_TAG_TICKETS);
ticket_state_xml = create_xml_node(xml_obj, XML_CIB_TAG_TICKET_STATE);
crm_xml_add(ticket_state_xml, PCMK_XA_ID, ticket_id);
}
for(list_iter = attr_delete; list_iter; list_iter = list_iter->next) {
const char *key = (const char *)list_iter->data;
xml_remove_prop(ticket_state_xml, key);
}
ticket = find_ticket(ticket_id, scheduler);
g_hash_table_iter_init(&hash_iter, attr_set);
while (g_hash_table_iter_next(&hash_iter, (gpointer *) & key, (gpointer *) & value)) {
crm_xml_add(ticket_state_xml, key, value);
if (pcmk__str_eq(key, "granted", pcmk__str_casei)
&& (ticket == NULL || ticket->granted == FALSE)
&& crm_is_true(value)) {
char *now = pcmk__ttoa(time(NULL));
crm_xml_add(ticket_state_xml, "last-granted", now);
free(now);
}
}
if (found && (attr_delete != NULL)) {
crm_log_xml_debug(xml_top, "Replace");
rc = cib->cmds->replace(cib, XML_CIB_TAG_STATUS, ticket_state_xml, cib_options);
rc = pcmk_legacy2rc(rc);
} else {
crm_log_xml_debug(xml_top, "Update");
rc = cib->cmds->modify(cib, XML_CIB_TAG_STATUS, xml_top, cib_options);
rc = pcmk_legacy2rc(rc);
}
free_xml(xml_top);
return rc;
}
static int
delete_ticket_state(gchar *ticket_id, cib_t * cib)
{
xmlNode *ticket_state_xml = NULL;
int rc = pcmk_rc_ok;
rc = find_ticket_state(cib, ticket_id, &ticket_state_xml);
if (rc == ENXIO) {
return pcmk_rc_ok;
} else if (rc != pcmk_rc_ok) {
return rc;
}
crm_log_xml_debug(ticket_state_xml, "Delete");
rc = cib->cmds->remove(cib, XML_CIB_TAG_STATUS, ticket_state_xml, cib_options);
rc = pcmk_legacy2rc(rc);
if (rc == pcmk_rc_ok) {
fprintf(stdout, "Cleaned up %s\n", ticket_id);
}
free_xml(ticket_state_xml);
return rc;
}
static GOptionContext *
build_arg_context(pcmk__common_args_t *args) {
GOptionContext *context = NULL;
const char *description = "Examples:\n\n"
"Display the info of tickets:\n\n"
"\tcrm_ticket --info\n\n"
"Display the detailed info of tickets:\n\n"
"\tcrm_ticket --details\n\n"
"Display the XML of 'ticketA':\n\n"
"\tcrm_ticket --ticket ticketA --query-xml\n\n"
"Display the rsc_ticket constraints that apply to 'ticketA':\n\n"
"\tcrm_ticket --ticket ticketA --constraints\n\n"
"Grant 'ticketA' to this cluster site:\n\n"
"\tcrm_ticket --ticket ticketA --grant\n\n"
"Revoke 'ticketA' from this cluster site:\n\n"
"\tcrm_ticket --ticket ticketA --revoke\n\n"
"Make 'ticketA' standby (the cluster site will treat a granted\n"
"'ticketA' as 'standby', and the dependent resources will be\n"
"stopped or demoted gracefully without triggering loss-policies):\n\n"
"\tcrm_ticket --ticket ticketA --standby\n\n"
"Activate 'ticketA' from being standby:\n\n"
"\tcrm_ticket --ticket ticketA --activate\n\n"
"Get the value of the 'granted' attribute for 'ticketA':\n\n"
"\tcrm_ticket --ticket ticketA --get-attr granted\n\n"
"Set the value of the 'standby' attribute for 'ticketA':\n\n"
"\tcrm_ticket --ticket ticketA --set-attr standby --attr-value true\n\n"
"Delete the 'granted' attribute for 'ticketA':\n\n"
"\tcrm_ticket --ticket ticketA --delete-attr granted\n\n"
"Erase the operation history of 'ticketA' at this cluster site,\n"
"causing the cluster site to 'forget' the existing ticket state:\n\n"
"\tcrm_ticket --ticket ticketA --cleanup\n\n";
context = pcmk__build_arg_context(args, NULL, NULL, NULL);
g_option_context_set_description(context, description);
pcmk__add_arg_group(context, "queries", "Queries:",
"Show queries", query_entries);
pcmk__add_arg_group(context, "commands", "Commands:",
"Show command options", command_entries);
pcmk__add_arg_group(context, "advanced", "Advanced Options:",
"Show advanced options", advanced_entries);
pcmk__add_arg_group(context, "additional", "Additional Options:",
"Show additional options", addl_entries);
pcmk__add_arg_group(context, "deprecated", "Deprecated Options:",
"Show deprecated options", deprecated_entries);
return context;
}
int
main(int argc, char **argv)
{
pcmk_scheduler_t *scheduler = NULL;
xmlNode *cib_xml_copy = NULL;
cib_t *cib_conn = NULL;
crm_exit_t exit_code = CRM_EX_OK;
int rc = pcmk_rc_ok;
pcmk__common_args_t *args = NULL;
GOptionContext *context = NULL;
gchar **processed_args = NULL;
attr_set = pcmk__strkey_table(free, free);
attr_delete = NULL;
args = pcmk__new_common_args(SUMMARY);
context = build_arg_context(args);
processed_args = pcmk__cmdline_preproc(argv, "dintvxCDGS");
if (!g_option_context_parse_strv(context, &processed_args, &error)) {
exit_code = CRM_EX_USAGE;
goto done;
}
pcmk__cli_init_logging("crm_ticket", args->verbosity);
if (args->version) {
g_strfreev(processed_args);
pcmk__free_arg_context(context);
/* FIXME: When crm_ticket is converted to use formatted output, this can go. */
pcmk__cli_help('v');
}
scheduler = pe_new_working_set();
if (scheduler == NULL) {
rc = errno;
exit_code = pcmk_rc2exitc(rc);
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Could not allocate scheduler data: %s", pcmk_rc_str(rc));
goto done;
}
- pe__set_working_set_flags(scheduler,
+ pcmk__set_scheduler_flags(scheduler,
pcmk_sched_no_counts|pcmk_sched_no_compat);
cib_conn = cib_new();
if (cib_conn == NULL) {
exit_code = CRM_EX_DISCONNECT;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Could not connect to the CIB manager");
goto done;
}
rc = cib_conn->cmds->signon(cib_conn, crm_system_name, cib_command);
rc = pcmk_legacy2rc(rc);
if (rc != pcmk_rc_ok) {
exit_code = pcmk_rc2exitc(rc);
g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Could not connect to the CIB: %s",
pcmk_rc_str(rc));
goto done;
}
if (options.xml_file != NULL) {
cib_xml_copy = filename2xml(options.xml_file);
} else {
rc = cib_conn->cmds->query(cib_conn, NULL, &cib_xml_copy, cib_scope_local | cib_sync_call);
rc = pcmk_legacy2rc(rc);
if (rc != pcmk_rc_ok) {
exit_code = pcmk_rc2exitc(rc);
g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Could not get local CIB: %s",
pcmk_rc_str(rc));
goto done;
}
}
if (!cli_config_update(&cib_xml_copy, NULL, FALSE)) {
exit_code = CRM_EX_CONFIG;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Could not update local CIB to latest schema version");
goto done;
}
scheduler->input = cib_xml_copy;
scheduler->now = crm_time_new(NULL);
cluster_status(scheduler);
/* For recording the tickets that are referenced in rsc_ticket constraints
* but have never been granted yet. */
pcmk__unpack_constraints(scheduler);
if (options.ticket_cmd == 'l' || options.ticket_cmd == 'L' || options.ticket_cmd == 'w') {
bool raw = false;
bool details = false;
if (options.ticket_cmd == 'L') {
details = true;
} else if (options.ticket_cmd == 'w') {
raw = true;
}
if (options.ticket_id) {
pcmk_ticket_t *ticket = find_ticket(options.ticket_id, scheduler);
if (ticket == NULL) {
exit_code = CRM_EX_NOSUCH;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"No such ticket '%s'", options.ticket_id);
goto done;
}
print_ticket(ticket, raw, details);
} else {
print_ticket_list(scheduler, raw, details);
}
} else if (options.ticket_cmd == 'q') {
rc = dump_ticket_xml(cib_conn, options.ticket_id);
exit_code = pcmk_rc2exitc(rc);
if (rc != pcmk_rc_ok) {
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Could not query ticket XML: %s", pcmk_rc_str(rc));
}
} else if (options.ticket_cmd == 'c') {
rc = dump_constraints(cib_conn, options.ticket_id);
exit_code = pcmk_rc2exitc(rc);
if (rc != pcmk_rc_ok) {
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Could not show ticket constraints: %s", pcmk_rc_str(rc));
}
} else if (options.ticket_cmd == 'G') {
const char *value = NULL;
if (options.ticket_id == NULL) {
exit_code = CRM_EX_NOSUCH;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Must supply ticket ID with -t");
goto done;
}
rc = get_ticket_state_attr(options.ticket_id, options.get_attr_name,
&value, scheduler);
if (rc == pcmk_rc_ok) {
fprintf(stdout, "%s\n", value);
} else if (rc == ENXIO && options.attr_default) {
fprintf(stdout, "%s\n", options.attr_default);
rc = pcmk_rc_ok;
}
exit_code = pcmk_rc2exitc(rc);
} else if (options.ticket_cmd == 'C') {
if (options.ticket_id == NULL) {
exit_code = CRM_EX_USAGE;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Must supply ticket ID with -t");
goto done;
}
if (options.force == FALSE) {
pcmk_ticket_t *ticket = NULL;
ticket = find_ticket(options.ticket_id, scheduler);
if (ticket == NULL) {
exit_code = CRM_EX_NOSUCH;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"No such ticket '%s'", options.ticket_id);
goto done;
}
if (ticket->granted) {
ticket_warning(options.ticket_id, "revoke");
exit_code = CRM_EX_INSUFFICIENT_PRIV;
goto done;
}
}
rc = delete_ticket_state(options.ticket_id, cib_conn);
exit_code = pcmk_rc2exitc(rc);
if (rc != pcmk_rc_ok) {
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Could not clean up ticket: %s", pcmk_rc_str(rc));
}
} else if (modified) {
if (options.ticket_id == NULL) {
exit_code = CRM_EX_USAGE;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Must supply ticket ID with -t");
goto done;
}
if (options.attr_value
&& (pcmk__str_empty(options.attr_name))) {
exit_code = CRM_EX_USAGE;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Must supply attribute name with -S for -v %s", options.attr_value);
goto done;
}
if (options.attr_name
&& (pcmk__str_empty(options.attr_value))) {
exit_code = CRM_EX_USAGE;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Must supply attribute value with -v for -S %s", options.attr_value);
goto done;
}
if (!allow_modification(options.ticket_id)) {
exit_code = CRM_EX_INSUFFICIENT_PRIV;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Ticket modification not allowed");
goto done;
}
rc = modify_ticket_state(options.ticket_id, cib_conn, scheduler);
exit_code = pcmk_rc2exitc(rc);
if (rc != pcmk_rc_ok) {
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Could not modify ticket: %s", pcmk_rc_str(rc));
}
} else if (options.ticket_cmd == 'S') {
/* Correct usage was handled in the "if (modified)" block above, so
* this is just for reporting usage errors
*/
if (pcmk__str_empty(options.attr_name)) {
// We only get here if ticket_cmd was left as default
exit_code = CRM_EX_USAGE;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Must supply a command");
goto done;
}
if (options.ticket_id == NULL) {
exit_code = CRM_EX_USAGE;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Must supply ticket ID with -t");
goto done;
}
if (pcmk__str_empty(options.attr_value)) {
exit_code = CRM_EX_USAGE;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Must supply value with -v for -S %s", options.attr_name);
goto done;
}
} else {
exit_code = CRM_EX_USAGE;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Unknown command: %c", options.ticket_cmd);
}
done:
if (attr_set) {
g_hash_table_destroy(attr_set);
}
attr_set = NULL;
if (attr_delete) {
g_list_free_full(attr_delete, free);
}
attr_delete = NULL;
pe_free_working_set(scheduler);
scheduler = NULL;
cib__clean_up_connection(&cib_conn);
g_strfreev(processed_args);
pcmk__free_arg_context(context);
g_free(options.attr_default);
g_free(options.attr_id);
free(options.attr_name);
free(options.attr_value);
free(options.get_attr_name);
g_free(options.set_name);
g_free(options.ticket_id);
g_free(options.xml_file);
pcmk__output_and_clear_error(&error, NULL);
crm_exit(exit_code);
}

File Metadata

Mime Type
text/x-diff
Expires
Sun, Jul 20, 9:16 PM (16 h, 59 m)
Storage Engine
blob
Storage Format
Raw Data
Storage Handle
2081517
Default Alt Text
(651 KB)

Event Timeline