diff --git a/tools/Makefile.am b/tools/Makefile.am index c822a8c287..4609b0f581 100644 --- a/tools/Makefile.am +++ b/tools/Makefile.am @@ -1,168 +1,167 @@ # # Copyright 2004-2019 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 $(top_srcdir)/mk/common.mk if BUILD_SYSTEMD systemdsystemunit_DATA = crm_mon.service endif -noinst_HEADERS = crm_mon.h crm_resource.h crm_resource_controller.h +noinst_HEADERS = crm_mon.h crm_resource.h pcmkdir = $(datadir)/$(PACKAGE) pcmk_DATA = report.common report.collector sbin_SCRIPTS = crm_report crm_standby crm_master crm_failcount if BUILD_CIBSECRETS sbin_SCRIPTS += cibsecret endif noinst_SCRIPTS = pcmk_simtimes EXTRA_DIST = crm_diff.8.inc \ crm_error.8.inc \ crm_mon.sysconfig \ crm_mon.8.inc \ crm_node.8.inc \ crm_rule.8.inc \ crm_simulate.8.inc \ fix-manpages \ stonith_admin.8.inc sbin_PROGRAMS = attrd_updater \ cibadmin \ crmadmin \ crm_simulate \ crm_attribute \ crm_diff \ crm_error \ crm_mon \ crm_node \ crm_resource \ crm_rule \ crm_shadow \ crm_verify \ crm_ticket \ iso8601 \ stonith_admin if BUILD_SERVICELOG sbin_PROGRAMS += notifyServicelogEvent endif if BUILD_OPENIPMI_SERVICELOG sbin_PROGRAMS += ipmiservicelogd endif ## SOURCES # A few tools are just thin wrappers around crm_attribute. # This makes their help get updated when crm_attribute changes # (see mk/common.mk). MAN8DEPS = crm_attribute crmadmin_SOURCES = crmadmin.c crmadmin_LDADD = $(top_builddir)/lib/pengine/libpe_status.la \ $(top_builddir)/lib/cib/libcib.la \ $(top_builddir)/lib/common/libcrmcommon.la crm_error_SOURCES = crm_error.c crm_error_LDADD = $(top_builddir)/lib/common/libcrmcommon.la cibadmin_SOURCES = cibadmin.c cibadmin_LDADD = $(top_builddir)/lib/cib/libcib.la \ $(top_builddir)/lib/common/libcrmcommon.la crm_shadow_SOURCES = crm_shadow.c crm_shadow_LDADD = $(top_builddir)/lib/cib/libcib.la \ $(top_builddir)/lib/common/libcrmcommon.la crm_node_SOURCES = crm_node.c crm_node_LDADD = $(top_builddir)/lib/cib/libcib.la \ $(top_builddir)/lib/common/libcrmcommon.la crm_simulate_SOURCES = crm_simulate.c crm_simulate_LDADD = $(top_builddir)/lib/pengine/libpe_status.la \ $(top_builddir)/lib/pacemaker/libpacemaker.la \ $(top_builddir)/lib/cib/libcib.la \ $(top_builddir)/lib/common/libcrmcommon.la crm_diff_SOURCES = crm_diff.c crm_diff_LDADD = $(top_builddir)/lib/common/libcrmcommon.la crm_mon_SOURCES = crm_mon.c crm_mon_curses.c crm_mon_print.c crm_mon_runtime.c crm_mon_xml.c crm_mon_LDADD = $(top_builddir)/lib/pengine/libpe_status.la \ $(top_builddir)/lib/fencing/libstonithd.la \ $(top_builddir)/lib/pacemaker/libpacemaker.la \ $(top_builddir)/lib/cib/libcib.la \ $(top_builddir)/lib/common/libcrmcommon.la \ $(CURSESLIBS) crm_verify_SOURCES = crm_verify.c crm_verify_LDADD = $(top_builddir)/lib/pengine/libpe_status.la \ $(top_builddir)/lib/pacemaker/libpacemaker.la \ $(top_builddir)/lib/cib/libcib.la \ $(top_builddir)/lib/common/libcrmcommon.la crm_attribute_SOURCES = crm_attribute.c crm_attribute_LDADD = $(top_builddir)/lib/cluster/libcrmcluster.la \ $(top_builddir)/lib/cib/libcib.la \ $(top_builddir)/lib/common/libcrmcommon.la crm_resource_SOURCES = crm_resource.c \ crm_resource_ban.c \ - crm_resource_controller.c \ crm_resource_print.c \ crm_resource_runtime.c crm_resource_LDADD = $(top_builddir)/lib/pengine/libpe_rules.la \ $(top_builddir)/lib/fencing/libstonithd.la \ $(top_builddir)/lib/lrmd/liblrmd.la \ $(top_builddir)/lib/services/libcrmservice.la \ $(top_builddir)/lib/pengine/libpe_status.la \ $(top_builddir)/lib/pacemaker/libpacemaker.la \ $(top_builddir)/lib/cib/libcib.la \ $(top_builddir)/lib/common/libcrmcommon.la crm_rule_SOURCES = crm_rule.c crm_rule_LDADD = $(top_builddir)/lib/cib/libcib.la \ $(top_builddir)/lib/pengine/libpe_rules.la \ $(top_builddir)/lib/pengine/libpe_status.la \ $(top_builddir)/lib/common/libcrmcommon.la iso8601_SOURCES = iso8601.c iso8601_LDADD = $(top_builddir)/lib/common/libcrmcommon.la attrd_updater_SOURCES = attrd_updater.c attrd_updater_LDADD = $(top_builddir)/lib/common/libcrmcommon.la crm_ticket_SOURCES = crm_ticket.c crm_ticket_LDADD = $(top_builddir)/lib/pengine/libpe_rules.la \ $(top_builddir)/lib/pengine/libpe_status.la \ $(top_builddir)/lib/pacemaker/libpacemaker.la \ $(top_builddir)/lib/cib/libcib.la \ $(top_builddir)/lib/common/libcrmcommon.la stonith_admin_SOURCES = stonith_admin.c stonith_admin_LDADD = $(top_builddir)/lib/pacemaker/libpacemaker.la \ $(top_builddir)/lib/cib/libcib.la \ $(top_builddir)/lib/pengine/libpe_status.la \ $(top_builddir)/lib/fencing/libstonithd.la \ $(top_builddir)/lib/common/libcrmcommon.la if BUILD_SERVICELOG notifyServicelogEvent_SOURCES = notifyServicelogEvent.c notifyServicelogEvent_CFLAGS = $(SERVICELOG_CFLAGS) notifyServicelogEvent_LDADD = $(top_builddir)/lib/common/libcrmcommon.la $(SERVICELOG_LIBS) endif if BUILD_OPENIPMI_SERVICELOG ipmiservicelogd_SOURCES = ipmiservicelogd.c ipmiservicelogd_CFLAGS = $(OPENIPMI_SERVICELOG_CFLAGS) $(SERVICELOG_CFLAGS) ipmiservicelogd_LDFLAGS = $(top_builddir)/lib/common/libcrmcommon.la $(OPENIPMI_SERVICELOG_LIBS) $(SERVICELOG_LIBS) endif CLEANFILES = $(man8_MANS) diff --git a/tools/crm_resource.c b/tools/crm_resource.c index 6853ad5330..c8c1cfacef 100644 --- a/tools/crm_resource.c +++ b/tools/crm_resource.c @@ -1,1568 +1,1584 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include - -#include -#include - #include #include #include - #include #include #include #include #include +#include +#include +#include + bool BE_QUIET = FALSE; bool scope_master = FALSE; int cib_options = cib_sync_call; - -static GMainLoop *mainloop = NULL; +static crm_exit_t exit_code = CRM_EX_OK; // Things that should be cleaned up on exit +static GMainLoop *mainloop = NULL; static cib_t *cib_conn = NULL; -static pcmk_controld_api_t *controld_api = NULL; +static pcmk_ipc_api_t *controld_api = NULL; static pe_working_set_t *data_set = NULL; #define MESSAGE_TIMEOUT_S 60 // Clean up and exit static crm_exit_t -bye(crm_exit_t exit_code) +bye(crm_exit_t ec) { - static bool crm_resource_shutdown_flag = false; - - if (crm_resource_shutdown_flag) { - // Allow usage like "return bye(...);" - return exit_code; - } - crm_resource_shutdown_flag = true; - if (cib_conn != NULL) { cib_t *save_cib_conn = cib_conn; - cib_conn = NULL; + cib_conn = NULL; // Ensure we can't free this twice save_cib_conn->cmds->signoff(save_cib_conn); cib_delete(save_cib_conn); } if (controld_api != NULL) { - pcmk_controld_api_t *save_controld_api = controld_api; + pcmk_ipc_api_t *save_controld_api = controld_api; - controld_api = NULL; - pcmk_free_controld_api(save_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(data_set); data_set = NULL; - crm_exit(exit_code); - return exit_code; + 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 fprintf(stderr, "\nAborting because no messages received in %d seconds\n", MESSAGE_TIMEOUT_S); crm_err("No messages received in %d seconds", MESSAGE_TIMEOUT_S); - bye(CRM_EX_TIMEOUT); + quit_main_loop(CRM_EX_TIMEOUT); return FALSE; } static void -handle_controller_reply(pcmk_controld_api_t *capi, void *api_data, - void *user_data) +controller_event_callback(pcmk_ipc_api_t *api, enum pcmk_ipc_event event_type, + crm_exit_t status, void *event_data, void *user_data) { - fprintf(stderr, "."); - if ((capi->replies_expected(capi) == 0) - && mainloop && g_main_loop_is_running(mainloop)) { - fprintf(stderr, " OK\n"); - crm_debug("Got all the replies we expected"); - bye(CRM_EX_OK); - } -} + 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; -static void -handle_controller_drop(pcmk_controld_api_t *capi, void *api_data, - void *user_data) -{ - crm_info("Connection to controller was terminated"); - bye(CRM_EX_DISCONNECT); + case pcmk_ipc_event_reply: + if (status != CRM_EX_OK) { + fprintf(stderr, "\nError: bad reply from controller: %s\n", + crm_exit_str(status)); + pcmk_disconnect_ipc(api); + quit_main_loop(status); + } else { + fprintf(stderr, "."); + if ((pcmk_controld_api_replies_expected(api) == 0) + && mainloop && g_main_loop_is_running(mainloop)) { + fprintf(stderr, " OK\n"); + crm_debug("Got all the replies we expected"); + pcmk_disconnect_ipc(api); + quit_main_loop(CRM_EX_OK); + } + } + break; + + default: + break; + } } static void -start_mainloop(pcmk_controld_api_t *capi) +start_mainloop(pcmk_ipc_api_t *capi) { - if (capi->replies_expected(capi) > 0) { - unsigned int count = capi->replies_expected(capi); + unsigned int count = pcmk_controld_api_replies_expected(capi); + if (count > 0) { fprintf(stderr, "Waiting for %d %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 GListPtr build_constraint_list(xmlNode *root) { GListPtr retval = NULL; xmlNode *cib_constraints = NULL; xmlXPathObjectPtr xpathObj = NULL; int ndx = 0; cib_constraints = get_object_root(XML_CIB_TAG_CONSTRAINTS, root); 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 pcmk__cli_option_t long_options[] = { // long option, argument type, storage, short option, description, flags { "help", no_argument, NULL, '?', "\t\tDisplay this text and exit", pcmk__option_default }, { "version", no_argument, NULL, '$', "\t\tDisplay version information and exit", pcmk__option_default }, { "verbose", no_argument, NULL, 'V', "\t\tIncrease debug output (may be specified multiple times)", pcmk__option_default }, { "quiet", no_argument, NULL, 'Q', "\t\tBe less descriptive in results", pcmk__option_default }, { "resource", required_argument, NULL, 'r', "\tResource ID", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\nQueries:", pcmk__option_default }, { "list", no_argument, NULL, 'L', "\t\tList all cluster resources with status", pcmk__option_default }, { "list-raw", no_argument, NULL, 'l', "\t\tList IDs of all instantiated resources (individual members rather " "than groups etc.)", pcmk__option_default }, { "list-cts", no_argument, NULL, 'c', NULL, pcmk__option_hidden }, { "list-operations", no_argument, NULL, 'O', "\tList active resource operations, optionally filtered by --resource " "and/or --node", pcmk__option_default }, { "list-all-operations", no_argument, NULL, 'o', "List all resource operations, optionally filtered by --resource " "and/or --node", pcmk__option_default }, { "list-standards", no_argument, NULL, 0, "\tList supported standards", pcmk__option_default }, { "list-ocf-providers", no_argument, NULL, 0, "List all available OCF providers", pcmk__option_default }, { "list-agents", required_argument, NULL, 0, "List all agents available for the named standard and/or provider", pcmk__option_default }, { "list-ocf-alternatives", required_argument, NULL, 0, "List all available providers for the named OCF agent", pcmk__option_default }, { "show-metadata", required_argument, NULL, 0, "Show the metadata for the named class:provider:agent", pcmk__option_default }, { "query-xml", no_argument, NULL, 'q', "\tShow XML configuration of resource (after any template expansion)", pcmk__option_default }, { "query-xml-raw", no_argument, NULL, 'w', "\tShow XML configuration of resource (before any template expansion)", pcmk__option_default }, { "get-parameter", required_argument, NULL, 'g', "Display named parameter for resource (use instance attribute unless " "--meta or --utilization is specified)", pcmk__option_default }, { "get-property", required_argument, NULL, 'G', "Display named property of resource ('class', 'type', or 'provider') " "(requires --resource)", pcmk__option_hidden }, { "locate", no_argument, NULL, 'W', "\t\tShow node(s) currently running resource", pcmk__option_default }, { "stack", no_argument, NULL, 'A', "\t\tDisplay the prerequisites and dependents of a resource", pcmk__option_default }, { "constraints", no_argument, NULL, 'a', "\tDisplay the (co)location constraints that apply to a resource", pcmk__option_default }, { "why", no_argument, NULL, 'Y', "\t\tShow why resources are not running, optionally filtered by " "--resource and/or --node", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\nCommands:", pcmk__option_default }, { "validate", no_argument, NULL, 0, "\t\tValidate resource configuration by calling agent's validate-all " "action. The configuration may be specified either by giving an " "existing resource name with -r, or by specifying --class, " "--agent, and --provider arguments, along with any number of " "--option arguments.", pcmk__option_default }, { "cleanup", no_argument, NULL, 'C', "\t\tIf resource has any past failures, clear its history and " "fail count. Optionally filtered by --resource, --node, " "--operation, and --interval (otherwise all). --operation and " "--interval apply to fail counts, but entire history is always " "cleared, to allow current state to be rechecked. If the named " "resource is part of a group, or one numbered instance of a clone " "or bundled resource, the clean-up applies to the whole collective " "resource unless --force is given.", pcmk__option_default }, { "refresh", no_argument, NULL, 'R', "\t\tDelete resource's history (including failures) so its current " "state is rechecked. Optionally filtered by --resource and --node " "(otherwise all). If the named resource is part of a group, or one " "numbered instance of a clone or bundled resource, the refresh " "applies to the whole collective resource unless --force is given.", pcmk__option_default }, { "set-parameter", required_argument, NULL, 'p', "Set named parameter for resource (requires -v). Use instance " "attribute unless --meta or --utilization is specified.", pcmk__option_default }, { "delete-parameter", required_argument, NULL, 'd', "Delete named parameter for resource. Use instance attribute unless " "--meta or --utilization is specified.", pcmk__option_default }, { "set-property", required_argument, NULL, 'S', "Set named property of resource ('class', 'type', or 'provider') " "(requires -r, -t, -v)", pcmk__option_hidden }, { "-spacer-", no_argument, NULL, '-', "\nResource location:", pcmk__option_default }, { "move", no_argument, NULL, 'M', "\t\tCreate a constraint to move resource. If --node is specified, the " "constraint will be to move to that node, otherwise it will be to " "ban the current node. Unless --force is specified, this will " "return an error if the resource is already running on the " "specified node. If --force is specified, this will always ban the " "current node. Optional: --lifetime, --master. NOTE: This may " "prevent the resource from running on its previous location until " "the implicit constraint expires or is removed with --clear.", pcmk__option_default }, { "ban", no_argument, NULL, 'B', "\t\tCreate a constraint to keep resource off a node. Optional: " "--node, --lifetime, --master. NOTE: This will prevent the " "resource from running on the affected node until the implicit " "constraint expires or is removed with --clear. If --node is not " "specified, it defaults to the node currently running the resource " "for primitives and groups, or the master for promotable clones " "with promoted-max=1 (all other situations result in an error as " "there is no sane default).", pcmk__option_default }, { "clear", no_argument, NULL, 'U', "\t\tRemove all constraints created by the --ban and/or --move " "commands. Requires: --resource. Optional: --node, --master, " "--expired. If --node is not specified, all constraints created " "by --ban and --move will be removed for the named resource. If " "--node and --force are specified, any constraint created by " "--move will be cleared, even if it is not for the specified node. " "If --expired is specified, only those constraints whose lifetimes " "have expired will be removed.", pcmk__option_default }, { "expired", no_argument, NULL, 'e', "\t\tModifies the --clear argument to remove constraints with " "expired lifetimes.", pcmk__option_default }, { "lifetime", required_argument, NULL, 'u', "\tLifespan (as ISO 8601 duration) of created constraints (with -B, " "-M) (see https://en.wikipedia.org/wiki/ISO_8601#Durations)", pcmk__option_default }, { "master", no_argument, NULL, 0, "\t\tLimit scope of command to Master role (with -B, -M, -U). For -B " "and -M, the previous master may remain active in the Slave role.", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\nAdvanced Commands:", pcmk__option_default }, { "delete", no_argument, NULL, 'D', "\t\t(Advanced) Delete a resource from the CIB. Required: -t", pcmk__option_default }, { "fail", no_argument, NULL, 'F', "\t\t(Advanced) Tell the cluster this resource has failed", pcmk__option_default }, { "restart", no_argument, NULL, 0, "\t\t(Advanced) Tell the cluster to restart this resource and " "anything that depends on it", pcmk__option_default }, { "wait", no_argument, NULL, 0, "\t\t(Advanced) Wait until the cluster settles into a stable state", pcmk__option_default }, { "force-demote", no_argument, NULL, 0, "\t(Advanced) Bypass the cluster and demote a resource on the local " "node. Unless --force is specified, this will refuse to do so if " "the cluster believes the resource is a clone instance already " "running on the local node.", pcmk__option_default }, { "force-stop", no_argument, NULL, 0, "\t(Advanced) Bypass the cluster and stop a resource on the local node", pcmk__option_default }, { "force-start", no_argument, NULL, 0, "\t(Advanced) Bypass the cluster and start a resource on the local " "node. Unless --force is specified, this will refuse to do so if " "the cluster believes the resource is a clone instance already " "running on the local node.", pcmk__option_default }, { "force-promote", no_argument, NULL, 0, "\t(Advanced) Bypass the cluster and promote a resource on the local " "node. Unless --force is specified, this will refuse to do so if " "the cluster believes the resource is a clone instance already " "running on the local node.", pcmk__option_default }, { "force-check", no_argument, NULL, 0, "\t(Advanced) Bypass the cluster and check the state of a resource on " "the local node", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\nValidate Options:", pcmk__option_default }, { "class", required_argument, NULL, 0, "\tThe standard the resource agent confirms to (for example, ocf). " "Use with --agent, --provider, --option, and --validate.", pcmk__option_default }, { "agent", required_argument, NULL, 0, "\tThe agent to use (for example, IPaddr). Use with --class, " "--provider, --option, and --validate.", pcmk__option_default }, { "provider", required_argument, NULL, 0, "\tThe vendor that supplies the resource agent (for example, " "heartbeat). Use with --class, --agent, --option, and --validate.", pcmk__option_default }, { "option", required_argument, NULL, 0, "\tSpecify a device configuration parameter as NAME=VALUE (may be " "specified multiple times). Use with --validate and without the " "-r option.", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\nAdditional Options:", pcmk__option_default }, { "node", required_argument, NULL, 'N', "\tNode name", pcmk__option_default }, { "recursive", no_argument, NULL, 0, "\tFollow colocation chains when using --set-parameter", pcmk__option_default }, { "resource-type", required_argument, NULL, 't', "Resource XML element (primitive, group, etc.) (with -D)", pcmk__option_default }, { "parameter-value", required_argument, NULL, 'v', "Value to use with -p", pcmk__option_default }, { "meta", no_argument, NULL, 'm', "\t\tUse resource meta-attribute instead of instance attribute " "(with -p, -g, -d)", pcmk__option_default }, { "utilization", no_argument, NULL, 'z', "\tUse resource utilization attribute instead of instance attribute " "(with -p, -g, -d)", pcmk__option_default }, { "operation", required_argument, NULL, 'n', "\tOperation to clear instead of all (with -C -r)", pcmk__option_default }, { "interval", required_argument, NULL, 'I', "\tInterval of operation to clear (default 0) (with -C -r -n)", pcmk__option_default }, { "set-name", required_argument, NULL, 's', "\t(Advanced) XML ID of attributes element to use (with -p, -d)", pcmk__option_default }, { "nvpair", required_argument, NULL, 'i', "\t(Advanced) XML ID of nvpair element to use (with -p, -d)", pcmk__option_default }, { "timeout", required_argument, NULL, 'T', "\t(Advanced) Abort if command does not finish in this time (with " "--restart, --wait, --force-*)", pcmk__option_default }, { "force", no_argument, NULL, 'f', "\t\tIf making CIB changes, do so regardless of quorum. See help for " "individual commands for additional behavior.", pcmk__option_default }, { "xml-file", required_argument, NULL, 'x', NULL, pcmk__option_hidden }, /* legacy options */ { "host-uname", required_argument, NULL, 'H', NULL, pcmk__option_hidden }, { "-spacer-", 1, NULL, '-', "\nExamples:", pcmk__option_paragraph }, { "-spacer-", 1, NULL, '-', "List the available OCF agents:", pcmk__option_paragraph }, { "-spacer-", 1, NULL, '-', " crm_resource --list-agents ocf", pcmk__option_example }, { "-spacer-", 1, NULL, '-', "List the available OCF agents from the linux-ha project:", pcmk__option_paragraph }, { "-spacer-", 1, NULL, '-', " crm_resource --list-agents ocf:heartbeat", pcmk__option_example }, { "-spacer-", 1, NULL, '-', "Move 'myResource' to a specific node:", pcmk__option_paragraph }, { "-spacer-", 1, NULL, '-', " crm_resource --resource myResource --move --node altNode", pcmk__option_example }, { "-spacer-", 1, NULL, '-', "Allow (but not force) 'myResource' to move back to its original " "location:", pcmk__option_paragraph }, { "-spacer-", 1, NULL, '-', " crm_resource --resource myResource --clear", pcmk__option_example }, { "-spacer-", 1, NULL, '-', "Stop 'myResource' (and anything that depends on it):", pcmk__option_paragraph }, { "-spacer-", 1, NULL, '-', " crm_resource --resource myResource --set-parameter target-role " "--meta --parameter-value Stopped", pcmk__option_example }, { "-spacer-", 1, NULL, '-', "Tell the cluster not to manage 'myResource' (the cluster will not " "attempt to start or stop the resource under any circumstances; " "useful when performing maintenance tasks on a resource):", pcmk__option_paragraph }, { "-spacer-", 1, NULL, '-', " crm_resource --resource myResource --set-parameter is-managed " "--meta --parameter-value false", pcmk__option_example }, { "-spacer-", 1, NULL, '-', "Erase the operation history of 'myResource' on 'aNode' (the cluster " "will 'forget' the existing resource state, including any " "errors, and attempt to recover the resource; useful when a " "resource had failed permanently and has been repaired " "by an administrator):", pcmk__option_paragraph }, { "-spacer-", 1, NULL, '-', " crm_resource --resource myResource --cleanup --node aNode", pcmk__option_example }, { 0, 0, 0, 0 } }; int main(int argc, char **argv) { char rsc_cmd = 'L'; const char *v_class = NULL; const char *v_agent = NULL; const char *v_provider = NULL; char *name = NULL; char *value = NULL; GHashTable *validate_options = NULL; const char *rsc_id = NULL; const char *host_uname = NULL; const char *prop_name = NULL; const char *prop_value = NULL; const char *rsc_type = NULL; const char *prop_id = NULL; const char *prop_set = NULL; const char *rsc_long_cmd = NULL; const char *longname = NULL; const char *operation = NULL; const char *interval_spec = NULL; const char *cib_file = getenv("CIB_file"); const char *xml_file = NULL; GHashTable *override_params = NULL; xmlNode *cib_xml_copy = NULL; pe_resource_t *rsc = NULL; bool recursive = FALSE; bool validate_cmdline = FALSE; /* whether we are just validating based on command line options */ bool require_resource = TRUE; /* whether command requires that resource be specified */ bool require_dataset = TRUE; /* whether command requires populated dataset instance */ bool require_crmd = FALSE; // whether command requires controller connection bool clear_expired = FALSE; int rc = pcmk_ok; int is_ocf_rc = 0; int option_index = 0; int timeout_ms = 0; int argerr = 0; int flag; int find_flags = 0; // Flags to use when searching for resource - crm_exit_t exit_code = CRM_EX_OK; crm_log_cli_init("crm_resource"); pcmk__set_cli_options(NULL, "| [options]", long_options, "perform tasks related to Pacemaker " "cluster resources"); validate_options = crm_str_table_new(); while (1) { flag = pcmk__next_cli_option(argc, argv, &option_index, &longname); if (flag == -1) break; switch (flag) { case 0: /* long options with no short equivalent */ if (safe_str_eq("master", longname)) { scope_master = TRUE; } else if(safe_str_eq(longname, "recursive")) { recursive = TRUE; } else if (safe_str_eq("wait", longname)) { rsc_cmd = flag; rsc_long_cmd = longname; require_resource = FALSE; require_dataset = FALSE; } else if ( safe_str_eq("validate", longname) || safe_str_eq("restart", longname) || safe_str_eq("force-demote", longname) || safe_str_eq("force-stop", longname) || safe_str_eq("force-start", longname) || safe_str_eq("force-promote", longname) || safe_str_eq("force-check", longname)) { rsc_cmd = flag; rsc_long_cmd = longname; find_flags = pe_find_renamed|pe_find_anon; crm_log_args(argc, argv); } else if (safe_str_eq("list-ocf-providers", longname) || safe_str_eq("list-ocf-alternatives", longname) || safe_str_eq("list-standards", longname)) { const char *text = NULL; lrmd_list_t *list = NULL; lrmd_list_t *iter = NULL; lrmd_t *lrmd_conn = lrmd_api_new(); if (safe_str_eq("list-ocf-providers", longname) || safe_str_eq("list-ocf-alternatives", longname)) { rc = lrmd_conn->cmds->list_ocf_providers(lrmd_conn, optarg, &list); text = "OCF providers"; } else if (safe_str_eq("list-standards", longname)) { rc = lrmd_conn->cmds->list_standards(lrmd_conn, &list); text = "standards"; } if (rc > 0) { for (iter = list; iter != NULL; iter = iter->next) { printf("%s\n", iter->val); } lrmd_list_freeall(list); } else if (optarg) { fprintf(stderr, "No %s found for %s\n", text, optarg); exit_code = CRM_EX_NOSUCH; } else { fprintf(stderr, "No %s found\n", text); exit_code = CRM_EX_NOSUCH; } lrmd_api_delete(lrmd_conn); return bye(exit_code); } else if (safe_str_eq("show-metadata", longname)) { char *standard = NULL; char *provider = NULL; char *type = NULL; char *metadata = NULL; lrmd_t *lrmd_conn = lrmd_api_new(); rc = crm_parse_agent_spec(optarg, &standard, &provider, &type); if (rc == pcmk_ok) { rc = lrmd_conn->cmds->get_metadata(lrmd_conn, standard, provider, type, &metadata, 0); } else { fprintf(stderr, "'%s' is not a valid agent specification\n", optarg); rc = -ENXIO; } if (metadata) { printf("%s\n", metadata); } else { fprintf(stderr, "Metadata query for %s failed: %s\n", optarg, pcmk_strerror(rc)); exit_code = crm_errno2exit(rc); } lrmd_api_delete(lrmd_conn); return bye(exit_code); } else if (safe_str_eq("list-agents", longname)) { lrmd_list_t *list = NULL; lrmd_list_t *iter = NULL; char *provider = strchr (optarg, ':'); lrmd_t *lrmd_conn = lrmd_api_new(); if (provider) { *provider++ = 0; } rc = lrmd_conn->cmds->list_agents(lrmd_conn, &list, optarg, provider); if (rc > 0) { for (iter = list; iter != NULL; iter = iter->next) { printf("%s\n", iter->val); } lrmd_list_freeall(list); } else { fprintf(stderr, "No agents found for standard=%s, provider=%s\n", optarg, (provider? provider : "*")); exit_code = CRM_EX_NOSUCH; } lrmd_api_delete(lrmd_conn); return bye(exit_code); } else if (safe_str_eq("class", longname)) { if (!(pcmk_get_ra_caps(optarg) & pcmk_ra_cap_params)) { if (BE_QUIET == FALSE) { fprintf(stdout, "Standard %s does not support parameters\n", optarg); } return bye(exit_code); } else { v_class = optarg; } validate_cmdline = TRUE; require_resource = FALSE; } else if (safe_str_eq("agent", longname)) { validate_cmdline = TRUE; require_resource = FALSE; v_agent = optarg; } else if (safe_str_eq("provider", longname)) { validate_cmdline = TRUE; require_resource = FALSE; v_provider = optarg; } else if (safe_str_eq("option", longname)) { crm_info("Scanning: --option %s", optarg); rc = pcmk_scan_nvpair(optarg, &name, &value); if (rc != 2) { fprintf(stderr, "Invalid option: --option %s: %s", optarg, pcmk_strerror(rc)); argerr++; } else { crm_info("Got: '%s'='%s'", name, value); } g_hash_table_replace(validate_options, name, value); } else { crm_err("Unhandled long option: %s", longname); } break; case 'V': resource_verbose++; crm_bump_log_level(argc, argv); break; case '$': case '?': pcmk__cli_help(flag, CRM_EX_OK); break; case 'x': xml_file = optarg; break; case 'Q': BE_QUIET = TRUE; break; case 'm': attr_set_type = XML_TAG_META_SETS; break; case 'z': attr_set_type = XML_TAG_UTILIZATION; break; case 'u': move_lifetime = strdup(optarg); break; case 'f': do_force = TRUE; crm_log_args(argc, argv); break; case 'i': prop_id = optarg; break; case 's': prop_set = optarg; break; case 'r': rsc_id = optarg; break; case 'v': prop_value = optarg; break; case 't': rsc_type = optarg; break; case 'T': timeout_ms = crm_get_msec(optarg); break; case 'e': clear_expired = TRUE; require_resource = FALSE; break; case 'C': case 'R': crm_log_args(argc, argv); require_resource = FALSE; if (cib_file == NULL) { require_crmd = TRUE; } rsc_cmd = flag; find_flags = pe_find_renamed|pe_find_anon; break; case 'n': operation = optarg; break; case 'I': interval_spec = optarg; break; case 'D': require_dataset = FALSE; crm_log_args(argc, argv); rsc_cmd = flag; find_flags = pe_find_renamed|pe_find_any; break; case 'F': require_crmd = TRUE; crm_log_args(argc, argv); rsc_cmd = flag; break; case 'U': case 'B': case 'M': crm_log_args(argc, argv); rsc_cmd = flag; find_flags = pe_find_renamed|pe_find_anon; break; case 'c': case 'L': case 'l': case 'O': case 'o': require_resource = FALSE; rsc_cmd = flag; break; case 'Y': require_resource = FALSE; rsc_cmd = flag; find_flags = pe_find_renamed|pe_find_anon; break; case 'q': case 'w': rsc_cmd = flag; find_flags = pe_find_renamed|pe_find_any; break; case 'W': case 'A': case 'a': rsc_cmd = flag; find_flags = pe_find_renamed|pe_find_anon; break; case 'S': require_dataset = FALSE; crm_log_args(argc, argv); prop_name = optarg; rsc_cmd = flag; find_flags = pe_find_renamed|pe_find_any; break; case 'p': case 'd': crm_log_args(argc, argv); prop_name = optarg; rsc_cmd = flag; find_flags = pe_find_renamed|pe_find_any; break; case 'G': case 'g': prop_name = optarg; rsc_cmd = flag; find_flags = pe_find_renamed|pe_find_any; break; case 'H': case 'N': crm_trace("Option %c => %s", flag, optarg); host_uname = optarg; break; default: CMD_ERR("Argument code 0%o (%c) is not (?yet?) supported", flag, flag); ++argerr; break; } } // Catch the case where the user didn't specify a command if (rsc_cmd == 'L') { require_resource = FALSE; } // --expired without --clear/-U doesn't make sense if (clear_expired == TRUE && rsc_cmd != 'U') { CMD_ERR("--expired requires --clear or -U"); argerr++; } if (optind < argc && argv[optind] != NULL && rsc_cmd == 0 && rsc_long_cmd) { override_params = crm_str_table_new(); while (optind < argc && argv[optind] != NULL) { char *name = calloc(1, strlen(argv[optind])); char *value = calloc(1, strlen(argv[optind])); int rc = sscanf(argv[optind], "%[^=]=%s", name, value); if(rc == 2) { g_hash_table_replace(override_params, name, value); } else { CMD_ERR("Error parsing '%s' as a name=value pair for --%s", argv[optind], rsc_long_cmd); free(value); free(name); argerr++; } optind++; } } else if (optind < argc && argv[optind] != NULL && rsc_cmd == 0) { CMD_ERR("non-option ARGV-elements: "); while (optind < argc && argv[optind] != NULL) { CMD_ERR("[%d of %d] %s ", optind, argc, argv[optind]); optind++; argerr++; } } if (optind > argc) { ++argerr; } // Sanity check validating from command line parameters. If everything checks out, // go ahead and run the validation. This way we don't need a CIB connection. if (validate_cmdline == TRUE) { // -r cannot be used with any of --class, --agent, or --provider if (rsc_id != NULL) { CMD_ERR("--resource cannot be used with --class, --agent, and --provider"); argerr++; // If --class, --agent, or --provider are given, --validate must also be given. } else if (!safe_str_eq(rsc_long_cmd, "validate")) { CMD_ERR("--class, --agent, and --provider require --validate"); argerr++; // 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 (crm_str_eq(v_class, "stonith", TRUE)) { if (v_provider != NULL) { CMD_ERR("stonith does not support providers"); argerr++; } else if (stonith_agent_exists(v_agent, 0) == FALSE) { CMD_ERR("%s is not a known stonith agent", v_agent ? v_agent : ""); argerr++; } } else if (resources_agent_exists(v_class, v_provider, v_agent) == FALSE) { CMD_ERR("%s:%s:%s is not a known resource", v_class ? v_class : "", v_provider ? v_provider : "", v_agent ? v_agent : ""); argerr++; } if (argerr == 0) { rc = cli_resource_execute_from_params("test", v_class, v_provider, v_agent, "validate-all", validate_options, override_params, timeout_ms); exit_code = crm_errno2exit(rc); return bye(exit_code); } } if (argerr) { CMD_ERR("Invalid option(s) supplied, use --help for valid usage"); return bye(CRM_EX_USAGE); } if (do_force) { crm_debug("Forcing..."); cib_options |= cib_quorum_override; } if (require_resource && !rsc_id) { CMD_ERR("Must supply a resource id with -r"); rc = -ENXIO; goto bail; } if (find_flags && rsc_id) { require_dataset = TRUE; } // Establish a connection to the CIB cib_conn = cib_new(); if ((cib_conn == NULL) || (cib_conn->cmds == NULL)) { CMD_ERR("Could not create CIB connection: %s", pcmk_strerror(rc)); goto bail; } rc = cib_conn->cmds->signon(cib_conn, crm_system_name, cib_command); if (rc != pcmk_ok) { CMD_ERR("Could not connect to the CIB: %s", pcmk_strerror(rc)); goto bail; } /* Populate working set from XML file if specified or CIB query otherwise */ if (require_dataset) { if (xml_file != NULL) { cib_xml_copy = filename2xml(xml_file); } else { rc = cib_conn->cmds->query(cib_conn, NULL, &cib_xml_copy, cib_scope_local | cib_sync_call); } if(rc != pcmk_ok) { goto bail; } /* Populate the working set instance */ data_set = pe_new_working_set(); if (data_set == NULL) { rc = -ENOMEM; goto bail; } set_bit(data_set->flags, pe_flag_no_counts); set_bit(data_set->flags, pe_flag_no_compat); rc = update_working_set_xml(data_set, &cib_xml_copy); if (rc != pcmk_ok) { goto bail; } cluster_status(data_set); } // If command requires that resource exist if specified, find it if (find_flags && rsc_id) { rsc = pe_find_resource_with_flags(data_set->resources, rsc_id, find_flags); if (rsc == NULL) { CMD_ERR("Resource '%s' not found", rsc_id); rc = -ENXIO; goto bail; } } // Establish a connection to the controller if needed if (require_crmd) { - char *client_uuid; - pcmk_controld_api_cb_t dispatch_cb = { - handle_controller_reply, NULL - }; - pcmk_controld_api_cb_t destroy_cb = { - handle_controller_drop, NULL - }; - - - client_uuid = pcmk__getpid_s(); - controld_api = pcmk_new_controld_api(crm_system_name, client_uuid); - free(client_uuid); - - rc = controld_api->connect(controld_api, true, &dispatch_cb, - &destroy_cb); + rc = pcmk_new_ipc_api(&controld_api, pcmk_ipc_controld); + if (rc != pcmk_rc_ok) { + CMD_ERR("Error connecting to the controller: %s", pcmk_rc_str(rc)); + rc = pcmk_rc2legacy(rc); + goto bail; + } + pcmk_register_ipc_callback(controld_api, controller_event_callback, + NULL); + rc = pcmk_connect_ipc(controld_api, pcmk_ipc_dispatch_main); if (rc != pcmk_rc_ok) { CMD_ERR("Error connecting to the controller: %s", pcmk_rc_str(rc)); rc = pcmk_rc2legacy(rc); goto bail; } } /* Handle rsc_cmd appropriately */ if (rsc_cmd == 'L') { rc = pcmk_ok; cli_resource_print_list(data_set, FALSE); } else if (rsc_cmd == 'l') { int found = 0; GListPtr lpc = NULL; rc = pcmk_ok; for (lpc = data_set->resources; lpc != NULL; lpc = lpc->next) { rsc = (pe_resource_t *) lpc->data; found++; cli_resource_print_raw(rsc); } if (found == 0) { printf("NO resources configured\n"); rc = -ENXIO; } } else if (rsc_cmd == 0 && rsc_long_cmd && safe_str_eq(rsc_long_cmd, "restart")) { /* We don't pass data_set because rsc needs to stay valid for the entire * lifetime of cli_resource_restart(), but it will reset and update the * working set multiple times, so it needs to use its own copy. */ rc = cli_resource_restart(rsc, host_uname, timeout_ms, cib_conn); } else if (rsc_cmd == 0 && rsc_long_cmd && safe_str_eq(rsc_long_cmd, "wait")) { rc = wait_till_stable(timeout_ms, cib_conn); } else if (rsc_cmd == 0 && rsc_long_cmd) { // validate, force-(stop|start|demote|promote|check) rc = cli_resource_execute(rsc, rsc_id, rsc_long_cmd, override_params, timeout_ms, cib_conn, data_set); if (rc >= 0) { is_ocf_rc = 1; } } else if (rsc_cmd == 'A' || rsc_cmd == 'a') { GListPtr lpc = NULL; xmlNode *cib_constraints = get_object_root(XML_CIB_TAG_CONSTRAINTS, data_set->input); unpack_constraints(cib_constraints, data_set); // Constraints apply to group/clone, not member/instance rsc = uber_parent(rsc); for (lpc = data_set->resources; lpc != NULL; lpc = lpc->next) { pe_resource_t *r = (pe_resource_t *) lpc->data; clear_bit(r->flags, pe_rsc_allocating); } cli_resource_print_colocation(rsc, TRUE, rsc_cmd == 'A', 1); fprintf(stdout, "* %s\n", rsc->id); cli_resource_print_location(rsc, NULL); for (lpc = data_set->resources; lpc != NULL; lpc = lpc->next) { pe_resource_t *r = (pe_resource_t *) lpc->data; clear_bit(r->flags, pe_rsc_allocating); } cli_resource_print_colocation(rsc, FALSE, rsc_cmd == 'A', 1); } else if (rsc_cmd == 'c') { GListPtr lpc = NULL; rc = pcmk_ok; for (lpc = data_set->resources; lpc != NULL; lpc = lpc->next) { rsc = (pe_resource_t *) lpc->data; cli_resource_print_cts(rsc); } cli_resource_print_cts_constraints(data_set); } else if (rsc_cmd == 'F') { rc = cli_resource_fail(controld_api, host_uname, rsc_id, data_set); if (rc == pcmk_rc_ok) { start_mainloop(controld_api); } rc = pcmk_rc2legacy(rc); } else if (rsc_cmd == 'O') { rc = cli_resource_print_operations(rsc_id, host_uname, TRUE, data_set); } else if (rsc_cmd == 'o') { rc = cli_resource_print_operations(rsc_id, host_uname, FALSE, data_set); } else if (rsc_cmd == 'W') { rc = cli_resource_search(rsc, rsc_id, data_set); if (rc >= 0) { rc = pcmk_ok; } } else if (rsc_cmd == 'q') { rc = cli_resource_print(rsc, data_set, TRUE); } else if (rsc_cmd == 'w') { rc = cli_resource_print(rsc, data_set, FALSE); } else if (rsc_cmd == 'Y') { pe_node_t *dest = NULL; if (host_uname) { dest = pe_find_node(data_set->nodes, host_uname); if (dest == NULL) { rc = -pcmk_err_node_unknown; goto bail; } } cli_resource_why(cib_conn, data_set->resources, rsc, dest); rc = pcmk_ok; } else if (rsc_cmd == 'U') { GListPtr before = NULL; GListPtr after = NULL; GListPtr remaining = NULL; GListPtr ele = NULL; pe_node_t *dest = NULL; if (BE_QUIET == FALSE) { before = build_constraint_list(data_set->input); } if (clear_expired == TRUE) { rc = cli_resource_clear_all_expired(data_set->input, cib_conn, rsc_id, host_uname, scope_master); } else if (host_uname) { dest = pe_find_node(data_set->nodes, host_uname); if (dest == NULL) { rc = -pcmk_err_node_unknown; if (BE_QUIET == FALSE) { g_list_free(before); } goto bail; } rc = cli_resource_clear(rsc_id, dest->details->uname, NULL, cib_conn, TRUE); } else { rc = cli_resource_clear(rsc_id, NULL, data_set->nodes, cib_conn, TRUE); } if (BE_QUIET == FALSE) { rc = cib_conn->cmds->query(cib_conn, NULL, &cib_xml_copy, cib_scope_local | cib_sync_call); if (rc != pcmk_ok) { CMD_ERR("Could not get modified CIB: %s\n", pcmk_strerror(rc)); g_list_free(before); goto bail; } data_set->input = cib_xml_copy; cluster_status(data_set); after = build_constraint_list(data_set->input); remaining = subtract_lists(before, after, (GCompareFunc) strcmp); for (ele = remaining; ele != NULL; ele = ele->next) { printf("Removing constraint: %s\n", (char *) ele->data); } g_list_free(before); g_list_free(after); g_list_free(remaining); } } else if (rsc_cmd == 'M' && host_uname) { rc = cli_resource_move(rsc, rsc_id, host_uname, cib_conn, data_set); } else if (rsc_cmd == 'B' && host_uname) { pe_node_t *dest = pe_find_node(data_set->nodes, host_uname); if (dest == NULL) { rc = -pcmk_err_node_unknown; goto bail; } rc = cli_resource_ban(rsc_id, dest->details->uname, NULL, cib_conn); } else if (rsc_cmd == 'B' || rsc_cmd == 'M') { pe_node_t *current = NULL; unsigned int nactive = 0; current = pe__find_active_requires(rsc, &nactive); if (nactive == 1) { rc = cli_resource_ban(rsc_id, current->details->uname, NULL, cib_conn); } else if (is_set(rsc->flags, pe_rsc_promotable)) { int count = 0; GListPtr iter = NULL; current = NULL; for(iter = rsc->children; iter; iter = iter->next) { pe_resource_t *child = (pe_resource_t *)iter->data; enum rsc_role_e child_role = child->fns->state(child, TRUE); if(child_role == RSC_ROLE_MASTER) { count++; current = pe__current_node(child); } } if(count == 1 && current) { rc = cli_resource_ban(rsc_id, current->details->uname, NULL, cib_conn); } else { rc = -EINVAL; exit_code = CRM_EX_USAGE; CMD_ERR("Resource '%s' not moved: active in %d locations (promoted in %d).", rsc_id, nactive, count); CMD_ERR("To prevent '%s' from running on a specific location, " "specify a node.", rsc_id); CMD_ERR("To prevent '%s' from being promoted at a specific " "location, specify a node and the master option.", rsc_id); } } else { rc = -EINVAL; exit_code = CRM_EX_USAGE; CMD_ERR("Resource '%s' not moved: active in %d locations.", rsc_id, nactive); CMD_ERR("To prevent '%s' from running on a specific location, " "specify a node.", rsc_id); } } else if (rsc_cmd == 'G') { rc = cli_resource_print_property(rsc, prop_name, data_set); } else if (rsc_cmd == 'S') { xmlNode *msg_data = NULL; if ((rsc_type == NULL) || !strlen(rsc_type)) { CMD_ERR("Must specify -t with resource type"); rc = -ENXIO; goto bail; } else if ((prop_value == NULL) || !strlen(prop_value)) { CMD_ERR("Must supply -v with new value"); rc = -EINVAL; goto bail; } CRM_LOG_ASSERT(prop_name != NULL); msg_data = create_xml_node(NULL, rsc_type); crm_xml_add(msg_data, XML_ATTR_ID, rsc_id); crm_xml_add(msg_data, prop_name, prop_value); rc = cib_conn->cmds->modify(cib_conn, XML_CIB_TAG_RESOURCES, msg_data, cib_options); free_xml(msg_data); } else if (rsc_cmd == 'g') { rc = cli_resource_print_attribute(rsc, prop_name, data_set); } else if (rsc_cmd == 'p') { if (prop_value == NULL || strlen(prop_value) == 0) { CMD_ERR("You need to supply a value with the -v option"); rc = -EINVAL; goto bail; } /* coverity[var_deref_model] False positive */ rc = cli_resource_update_attribute(rsc, rsc_id, prop_set, prop_id, prop_name, prop_value, recursive, cib_conn, data_set); } else if (rsc_cmd == 'd') { /* coverity[var_deref_model] False positive */ rc = cli_resource_delete_attribute(rsc, rsc_id, prop_set, prop_id, prop_name, cib_conn, data_set); } else if ((rsc_cmd == 'C') && rsc) { if (do_force == FALSE) { rsc = uber_parent(rsc); } crm_debug("Erasing failures of %s (%s requested) on %s", rsc->id, rsc_id, (host_uname? host_uname: "all nodes")); rc = cli_resource_delete(controld_api, host_uname, rsc, operation, interval_spec, TRUE, data_set); if ((rc == pcmk_ok) && !BE_QUIET) { // Show any reasons why resource might stay stopped cli_resource_check(cib_conn, rsc); } if (rc == pcmk_ok) { start_mainloop(controld_api); } } else if (rsc_cmd == 'C') { rc = cli_cleanup_all(controld_api, host_uname, operation, interval_spec, data_set); if (rc == pcmk_ok) { start_mainloop(controld_api); } } else if ((rsc_cmd == 'R') && rsc) { if (do_force == FALSE) { rsc = uber_parent(rsc); } crm_debug("Re-checking the state of %s (%s requested) on %s", rsc->id, rsc_id, (host_uname? host_uname: "all nodes")); rc = cli_resource_delete(controld_api, host_uname, rsc, NULL, 0, FALSE, data_set); if ((rc == pcmk_ok) && !BE_QUIET) { // Show any reasons why resource might stay stopped cli_resource_check(cib_conn, rsc); } if (rc == pcmk_ok) { start_mainloop(controld_api); } } else if (rsc_cmd == 'R') { const char *router_node = host_uname; int attr_options = pcmk__node_attr_none; if (host_uname) { pe_node_t *node = pe_find_node(data_set->nodes, host_uname); if (pe__is_guest_or_remote_node(node)) { node = pe__current_node(node->details->remote_rsc); if (node == NULL) { CMD_ERR("No cluster connection to Pacemaker Remote node %s detected", host_uname); rc = -ENXIO; goto bail; } router_node = node->details->uname; attr_options |= pcmk__node_attr_remote; } } if (controld_api == NULL) { printf("Dry run: skipping clean-up of %s due to CIB_file\n", host_uname? host_uname : "all nodes"); rc = pcmk_ok; goto bail; } crm_debug("Re-checking the state of all resources on %s", host_uname?host_uname:"all nodes"); rc = pcmk_rc2legacy(pcmk__node_attr_request_clear(NULL, host_uname, NULL, NULL, NULL, NULL, attr_options)); - if (controld_api->reprobe(controld_api, host_uname, - router_node) == pcmk_rc_ok) { + if (pcmk_controld_api_reprobe(controld_api, host_uname, + router_node) == pcmk_rc_ok) { start_mainloop(controld_api); } } else if (rsc_cmd == 'D') { xmlNode *msg_data = NULL; if (rsc_type == NULL) { CMD_ERR("You need to specify a resource type with -t"); rc = -ENXIO; goto bail; } msg_data = create_xml_node(NULL, rsc_type); crm_xml_add(msg_data, XML_ATTR_ID, rsc_id); rc = cib_conn->cmds->remove(cib_conn, XML_CIB_TAG_RESOURCES, msg_data, cib_options); free_xml(msg_data); } else { CMD_ERR("Unknown command: %c", rsc_cmd); } bail: if (is_ocf_rc) { exit_code = rc; } else if (rc != pcmk_ok) { CMD_ERR("Error performing operation: %s", pcmk_strerror(rc)); if (rc == -pcmk_err_no_quorum) { CMD_ERR("To ignore quorum, use the force option"); } if (exit_code == CRM_EX_OK) { exit_code = crm_errno2exit(rc); } } return bye(exit_code); } diff --git a/tools/crm_resource.h b/tools/crm_resource.h index 0bf7bee6b7..cb7f506f3b 100644 --- a/tools/crm_resource.h +++ b/tools/crm_resource.h @@ -1,107 +1,104 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include -#include "crm_resource_controller.h" extern bool print_pending; extern bool scope_master; extern bool do_force; extern bool BE_QUIET; extern int resource_verbose; extern int cib_options; extern char *move_lifetime; extern const char *attr_set_type; -extern pcmk_controld_api_cb_t controld_api_cb; - /* ban */ int cli_resource_prefer(const char *rsc_id, const char *host, cib_t * cib_conn); int cli_resource_ban(const char *rsc_id, const char *host, GListPtr allnodes, cib_t * cib_conn); int cli_resource_clear(const char *rsc_id, const char *host, GListPtr allnodes, cib_t * cib_conn, bool clear_ban_constraints); int cli_resource_clear_all_expired(xmlNode *root, cib_t *cib_conn, const char *rsc, const char *node, bool scope_master); /* print */ void cli_resource_print_cts(pe_resource_t * rsc); void cli_resource_print_raw(pe_resource_t * rsc); void cli_resource_print_cts_constraints(pe_working_set_t * data_set); void cli_resource_print_location(pe_resource_t * rsc, const char *prefix); void cli_resource_print_colocation(pe_resource_t * rsc, bool dependents, bool recursive, int offset); int cli_resource_print(pe_resource_t *rsc, pe_working_set_t *data_set, bool expanded); int cli_resource_print_list(pe_working_set_t * data_set, bool raw); int cli_resource_print_attribute(pe_resource_t *rsc, const char *attr, pe_working_set_t *data_set); int cli_resource_print_property(pe_resource_t *rsc, const char *attr, pe_working_set_t *data_set); int cli_resource_print_operations(const char *rsc_id, const char *host_uname, bool active, pe_working_set_t * data_set); /* runtime */ void cli_resource_check(cib_t * cib, pe_resource_t *rsc); -int cli_resource_fail(pcmk_controld_api_t *controld_api, +int cli_resource_fail(pcmk_ipc_api_t *controld_api, const char *host_uname, const char *rsc_id, pe_working_set_t *data_set); int cli_resource_search(pe_resource_t *rsc, const char *requested_name, pe_working_set_t *data_set); -int cli_resource_delete(pcmk_controld_api_t *controld_api, +int cli_resource_delete(pcmk_ipc_api_t *controld_api, const char *host_uname, pe_resource_t *rsc, const char *operation, const char *interval_spec, bool just_failures, pe_working_set_t *data_set); -int cli_cleanup_all(pcmk_controld_api_t *controld_api, const char *node_name, +int cli_cleanup_all(pcmk_ipc_api_t *controld_api, const char *node_name, const char *operation, const char *interval_spec, pe_working_set_t *data_set); int cli_resource_restart(pe_resource_t *rsc, const char *host, int timeout_ms, cib_t *cib); int cli_resource_move(pe_resource_t *rsc, const char *rsc_id, const char *host_name, cib_t *cib, pe_working_set_t *data_set); int cli_resource_execute_from_params(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 cli_resource_execute(pe_resource_t *rsc, const char *requested_name, const char *rsc_action, GHashTable *override_hash, int timeout_ms, cib_t *cib, pe_working_set_t *data_set); int cli_resource_update_attribute(pe_resource_t *rsc, const char *requested_name, const char *attr_set, const char *attr_id, const char *attr_name, const char *attr_value, bool recursive, cib_t *cib, pe_working_set_t *data_set); int cli_resource_delete_attribute(pe_resource_t *rsc, const char *requested_name, const char *attr_set, const char *attr_id, const char *attr_name, cib_t *cib, pe_working_set_t *data_set); GList* subtract_lists(GList *from, GList *items, GCompareFunc cmp); int update_working_set_xml(pe_working_set_t *data_set, xmlNode **xml); int wait_till_stable(int timeout_ms, cib_t * cib); void cli_resource_why(cib_t *cib_conn, GListPtr resources, pe_resource_t *rsc, pe_node_t *node); diff --git a/tools/crm_resource_controller.c b/tools/crm_resource_controller.c deleted file mode 100644 index 994a7be0ff..0000000000 --- a/tools/crm_resource_controller.c +++ /dev/null @@ -1,425 +0,0 @@ -/* - * Copyright 2020 the Pacemaker project contributors - * - * The version control history for this file may have further details. - * - * This source code is licensed under the GNU General Public License version 2 - * or later (GPLv2+) WITHOUT ANY WARRANTY. - */ - -#include -#include -#include -#include "crm_resource.h" - -// API object's private members -struct controller_private { - char *client_name; // Client name to use with IPC - char *client_uuid; // Client UUID to use with IPC - mainloop_io_t *source; // If main loop used, I/O source for IPC - crm_ipc_t *ipc; // IPC connection to controller - int replies_expected; // How many controller replies are expected - pcmk_controld_api_cb_t dispatch_cb; // Caller's registered dispatch callback - pcmk_controld_api_cb_t destroy_cb; // Caller's registered destroy callback -}; - -static void -call_client_callback(pcmk_controld_api_t *api, pcmk_controld_api_cb_t *cb, - void *api_data) -{ - if ((cb != NULL) && (cb->callback != NULL)) { - cb->callback(api, api_data, cb->user_data); - } -} - -/* - * IPC callbacks when used with main loop - */ - -static void -controller_ipc_destroy(gpointer user_data) -{ - pcmk_controld_api_t *api = user_data; - struct controller_private *private = api->private; - - private->ipc = NULL; - private->source = NULL; - call_client_callback(api, &(private->destroy_cb), NULL); -} - -// \return < 0 if connection is no longer required, >= 0 if it is -static int -controller_ipc_dispatch(const char *buffer, ssize_t length, gpointer user_data) -{ - xmlNode *msg = NULL; - pcmk_controld_api_t *api = user_data; - - CRM_CHECK(buffer && api && api->private, return 0); - - msg = string2xml(buffer); - if (msg == NULL) { - crm_warn("Received malformed controller IPC message"); - } else { - struct controller_private *private = api->private; - - crm_log_xml_trace(msg, "controller-reply"); - private->replies_expected--; - call_client_callback(api, &(private->dispatch_cb), - get_message_xml(msg, F_CRM_DATA)); - free_xml(msg); - } - return 0; -} - -/* - * IPC utilities - */ - -// \return Standard Pacemaker return code -static int -send_hello(crm_ipc_t *ipc, const char *client_name, const char *client_uuid) -{ - xmlNode *hello = create_hello_message(client_uuid, client_name, "0", "1"); - int rc = crm_ipc_send(ipc, hello, 0, 0, NULL); - - free_xml(hello); - if (rc < 0) { - rc = pcmk_legacy2rc(rc); - crm_info("Could not send IPC hello to %s: %s " CRM_XS " rc=%s", - CRM_SYSTEM_CRMD /* ipc->name */, - pcmk_rc_str(rc), rc); - return rc; - } - crm_debug("Sent IPC hello to %s", CRM_SYSTEM_CRMD /* ipc->name */); - return pcmk_rc_ok; -} - -// \return Standard Pacemaker return code -static int -send_controller_request(pcmk_controld_api_t *api, const char *op, - xmlNode *msg_data, const char *node) -{ - int rc; - struct controller_private *private = api->private; - xmlNode *cmd = create_request(op, msg_data, node, CRM_SYSTEM_CRMD, - private->client_name, private->client_uuid); - const char *reference = crm_element_value(cmd, XML_ATTR_REFERENCE); - - if ((cmd == NULL) || (reference == NULL)) { - return EINVAL; - } - - //@TODO pass as args? 0=crm_ipc_flags, 0=timeout_ms (default 5s), NULL=reply - crm_log_xml_trace(cmd, "controller-request"); - rc = crm_ipc_send(private->ipc, cmd, 0, 0, NULL); - free_xml(cmd); - if (rc < 0) { - return pcmk_legacy2rc(rc); - } - private->replies_expected++; - return pcmk_rc_ok; -} - -/* - * pcmk_controld_api_t methods - */ - -static int -controller_connect_mainloop(pcmk_controld_api_t *api) -{ - struct controller_private *private = api->private; - struct ipc_client_callbacks callbacks = { - .dispatch = controller_ipc_dispatch, - .destroy = controller_ipc_destroy, - }; - - private->source = mainloop_add_ipc_client(CRM_SYSTEM_CRMD, - G_PRIORITY_DEFAULT, 0, api, - &callbacks); - if (private->source == NULL) { - return ENOTCONN; - } - - private->ipc = mainloop_get_ipc_client(private->source); - if (private->ipc == NULL) { - (void) api->disconnect(api); - return ENOTCONN; - } - - crm_debug("Connected to %s IPC (attaching to main loop)", CRM_SYSTEM_CRMD); - return pcmk_rc_ok; -} - -static int -controller_connect_no_mainloop(pcmk_controld_api_t *api) -{ - struct controller_private *private = api->private; - - private->ipc = crm_ipc_new(CRM_SYSTEM_CRMD, 0); - if (private->ipc == NULL) { - return ENOTCONN; - } - if (!crm_ipc_connect(private->ipc)) { - crm_ipc_close(private->ipc); - crm_ipc_destroy(private->ipc); - private->ipc = NULL; - return errno; - } - /* @TODO caller needs crm_ipc_get_fd(private->ipc); either add method for - * that, or replace use_mainloop with int *fd - */ - crm_debug("Connected to %s IPC", CRM_SYSTEM_CRMD); - return pcmk_rc_ok; -} - -static void -set_callback(pcmk_controld_api_cb_t *dest, pcmk_controld_api_cb_t *source) -{ - if (source) { - dest->callback = source->callback; - dest->user_data = source->user_data; - } -} - -static int -controller_api_connect(pcmk_controld_api_t *api, bool use_mainloop, - pcmk_controld_api_cb_t *dispatch_cb, - pcmk_controld_api_cb_t *destroy_cb) -{ - int rc = pcmk_rc_ok; - struct controller_private *private; - - if (api == NULL) { - return EINVAL; - } - private = api->private; - - set_callback(&(private->dispatch_cb), dispatch_cb); - set_callback(&(private->destroy_cb), destroy_cb); - - if (private->ipc != NULL) { - return pcmk_rc_ok; // already connected - } - - if (use_mainloop) { - rc = controller_connect_mainloop(api); - } else { - rc = controller_connect_no_mainloop(api); - } - if (rc != pcmk_rc_ok) { - return rc; - } - - rc = send_hello(private->ipc, private->client_name, private->client_uuid); - if (rc != pcmk_rc_ok) { - (void) api->disconnect(api); - } - return rc; -} - -static int -controller_api_disconnect(pcmk_controld_api_t *api) -{ - struct controller_private *private = api->private; - - if (private->source != NULL) { - // Attached to main loop - mainloop_del_ipc_client(private->source); - private->source = NULL; - private->ipc = NULL; - - } else if (private->ipc != NULL) { - // Not attached to main loop - crm_ipc_t *ipc = private->ipc; - - private->ipc = NULL; - crm_ipc_close(ipc); - crm_ipc_destroy(ipc); - } - crm_debug("Disconnected from %s IPC", CRM_SYSTEM_CRMD /* ipc->name */); - return pcmk_rc_ok; -} - -//@TODO dispatch function for non-mainloop a la stonith_dispatch() -//@TODO convenience retry-connect function a la stonith_api_connect_retry() - -static unsigned int -controller_api_replies_expected(pcmk_controld_api_t *api) -{ - if (api != NULL) { - struct controller_private *private = api->private; - - return private->replies_expected; - } - return 0; -} - -static xmlNode * -create_reprobe_message_data(const char *target_node, const char *router_node) -{ - xmlNode *msg_data; - - msg_data = create_xml_node(NULL, "data_for_" CRM_OP_REPROBE); - crm_xml_add(msg_data, XML_LRM_ATTR_TARGET, target_node); - if ((router_node != NULL) && safe_str_neq(router_node, target_node)) { - crm_xml_add(msg_data, XML_LRM_ATTR_ROUTER_NODE, router_node); - } - return msg_data; -} - -static int -controller_api_reprobe(pcmk_controld_api_t *api, const char *target_node, - const char *router_node) -{ - int rc = EINVAL; - - if (api != NULL) { - xmlNode *msg_data; - - crm_debug("Sending %s IPC request to reprobe %s via %s", - CRM_SYSTEM_CRMD, crm_str(target_node), crm_str(router_node)); - msg_data = create_reprobe_message_data(target_node, router_node); - rc = send_controller_request(api, CRM_OP_REPROBE, msg_data, - (router_node? router_node : target_node)); - free_xml(msg_data); - } - return rc; -} - -// \return Standard Pacemaker return code -static int -controller_resource_op(pcmk_controld_api_t *api, const char *op, - const char *target_node, const char *router_node, - bool cib_only, const char *rsc_id, - const char *rsc_long_id, const char *standard, - const char *provider, const char *type) -{ - int rc; - char *key; - xmlNode *msg_data, *xml_rsc, *params; - - if (api == NULL) { - return EINVAL; - } - if (router_node == NULL) { - router_node = target_node; - } - - msg_data = create_xml_node(NULL, XML_GRAPH_TAG_RSC_OP); - - /* The controller logs the transition key from resource op requests, so we - * need to have *something* for it. - */ - key = pcmk__transition_key(0, getpid(), 0, - "xxxxxxxx-xrsc-opxx-xcrm-resourcexxxx"); - crm_xml_add(msg_data, XML_ATTR_TRANSITION_KEY, key); - free(key); - - crm_xml_add(msg_data, XML_LRM_ATTR_TARGET, target_node); - if (safe_str_neq(router_node, target_node)) { - crm_xml_add(msg_data, XML_LRM_ATTR_ROUTER_NODE, router_node); - } - - if (cib_only) { - // Indicate that only the CIB needs to be cleaned - crm_xml_add(msg_data, PCMK__XA_MODE, XML_TAG_CIB); - } - - xml_rsc = create_xml_node(msg_data, XML_CIB_TAG_RESOURCE); - crm_xml_add(xml_rsc, XML_ATTR_ID, rsc_id); - crm_xml_add(xml_rsc, XML_ATTR_ID_LONG, rsc_long_id); - crm_xml_add(xml_rsc, XML_AGENT_ATTR_CLASS, standard); - crm_xml_add(xml_rsc, XML_AGENT_ATTR_PROVIDER, provider); - crm_xml_add(xml_rsc, XML_ATTR_TYPE, type); - - params = create_xml_node(msg_data, XML_TAG_ATTRS); - crm_xml_add(params, XML_ATTR_CRM_VERSION, CRM_FEATURE_SET); - - // The controller parses the timeout from the request - key = crm_meta_name(XML_ATTR_TIMEOUT); - crm_xml_add(params, key, "60000"); /* 1 minute */ //@TODO pass as arg - free(key); - - rc = send_controller_request(api, op, msg_data, router_node); - free_xml(msg_data); - return rc; -} - -static int -controller_api_fail_resource(pcmk_controld_api_t *api, - const char *target_node, const char *router_node, - const char *rsc_id, const char *rsc_long_id, - const char *standard, const char *provider, - const char *type) -{ - crm_debug("Sending %s IPC request to fail %s (a.k.a. %s) on %s via %s", - CRM_SYSTEM_CRMD, crm_str(rsc_id), crm_str(rsc_long_id), - crm_str(target_node), crm_str(router_node)); - return controller_resource_op(api, CRM_OP_LRM_FAIL, target_node, - router_node, false, rsc_id, rsc_long_id, - standard, provider, type); -} - -static int -controller_api_refresh_resource(pcmk_controld_api_t *api, - const char *target_node, - const char *router_node, - const char *rsc_id, const char *rsc_long_id, - const char *standard, const char *provider, - const char *type, bool cib_only) -{ - crm_debug("Sending %s IPC request to refresh %s (a.k.a. %s) on %s via %s", - CRM_SYSTEM_CRMD, crm_str(rsc_id), crm_str(rsc_long_id), - crm_str(target_node), crm_str(router_node)); - return controller_resource_op(api, CRM_OP_LRM_DELETE, target_node, - router_node, cib_only, rsc_id, rsc_long_id, - standard, provider, type); -} - -pcmk_controld_api_t * -pcmk_new_controld_api(const char *client_name, const char *client_uuid) -{ - struct controller_private *private; - pcmk_controld_api_t *api = calloc(1, sizeof(pcmk_controld_api_t)); - - CRM_ASSERT(api != NULL); - - api->private = calloc(1, sizeof(struct controller_private)); - CRM_ASSERT(api->private != NULL); - private = api->private; - - if (client_name == NULL) { - client_name = crm_system_name? crm_system_name : "client"; - } - private->client_name = strdup(client_name); - CRM_ASSERT(private->client_name != NULL); - - if (client_uuid == NULL) { - private->client_uuid = crm_generate_uuid(); - } else { - private->client_uuid = strdup(client_uuid); - } - CRM_ASSERT(private->client_uuid != NULL); - - api->connect = controller_api_connect; - api->disconnect = controller_api_disconnect; - api->replies_expected = controller_api_replies_expected; - api->reprobe = controller_api_reprobe; - api->fail_resource = controller_api_fail_resource; - api->refresh_resource = controller_api_refresh_resource; - return api; -} - -void -pcmk_free_controld_api(pcmk_controld_api_t *api) -{ - if (api != NULL) { - struct controller_private *private = api->private; - - api->disconnect(api); - free(private->client_name); - free(private->client_uuid); - free(api->private); - free(api); - } -} diff --git a/tools/crm_resource_controller.h b/tools/crm_resource_controller.h deleted file mode 100644 index 50e20b43f7..0000000000 --- a/tools/crm_resource_controller.h +++ /dev/null @@ -1,198 +0,0 @@ -/* - * Copyright 2020 the Pacemaker project contributors - * - * The version control history for this file may have further details. - * - * This source code is licensed under the GNU Lesser General Public License - * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. - */ -#ifndef PCMK__CONTROLD_API_H -#define PCMK__CONTROLD_API_H - -#ifdef __cplusplus -extern "C" { -#endif - -#include // bool - -/* This is a demonstration of an abstracted controller IPC API. It is expected - * that this will be improved and moved to libcrmcommon. - * - * @TODO We could consider whether it's reasonable to have a single type for - * all daemons' IPC APIs (e.g. pcmk_ipc_api_t instead of pcmk_*_api_t). They - * could potentially have common connect/disconnect methods and then a void* to - * a group of API-specific methods. - * - * In that case, the callback type would also need to be generic, taking - * (pcmk_ipc_api_t *api, void *api_data, void *user_data), with individual APIs - * having functions for getting useful info from api_data. If all APIs followed - * the call_id model, we could use int call_id instead of api_data. - * - * A major annoyance is that the controller IPC protocol currently does not have - * any way to tie a particular reply to a particular request. The current - * clients (crmadmin, crm_node, and crm_resource) simply know what kind of reply - * to expect for the kind of request they sent. In crm_resource's case, all it - * does is count replies, ignoring their content altogether. - * - * That really forces us to have a single callback for all events rather than a - * per-request callback. That in turn implies that callers can only provide a - * single user data pointer. - * - * @TODO Define protocol version constants to use in hello message. - * @TODO Allow callers to specify timeouts. - * @TODO Define call IDs for controller ops, while somehow maintaining backward - * compatibility, since a client running on a Pacemaker Remote node could - * be older or newer than the controller on the connection's cluster - * node. - * @TODO The controller currently does not respond to hello messages. We should - * establish a common connection handshake protocol for all daemons that - * involves a hello message and acknowledgement. We should support sync - * or async connection (i.e. block until the ack is received, or return - * after the hello is sent and call a connection callback when the hello - * ack is received). - */ - -//! \internal -typedef struct pcmk_controld_api_s pcmk_controld_api_t; - -//! \internal -typedef struct pcmk_controld_api_callback_s { - void (*callback)(pcmk_controld_api_t *api, void *api_data, void *user_data); - void *user_data; -} pcmk_controld_api_cb_t; - -//! \internal -struct pcmk_controld_api_s { - //! \internal - void *private; - - /*! - * \internal - * \brief Connect to the local controller - * - * \param[in] api Controller API instance - * \param[in] use_mainloop If true, attach IPC to main loop - * \param[in] dispatch_cb If not NULL, call this when replies are received - * \param[in] destroy_cb If not NULL, call this if connection drops - * - * \return Standard Pacemaker return code - * \note Only the pointers inside the callback objects need to be - * persistent, not the callback objects themselves. The destroy_cb - * will be called only for unrequested disconnects. - */ - int (*connect)(pcmk_controld_api_t *api, bool use_mainloop, - pcmk_controld_api_cb_t *dispatch_cb, - pcmk_controld_api_cb_t *destroy_cb); - - /*! - * \internal - * \brief Disconnect from the local controller - * - * \param[in] api Controller API instance - * - * \return Standard Pacemaker return code - */ - int (*disconnect)(pcmk_controld_api_t *api); - - /*! - * \internal - * \brief Check number of replies still expected from controller - * - * \param[in] api Controller API instance - * - * \return Number of expected replies - */ - unsigned int (*replies_expected)(pcmk_controld_api_t *api); - - /*! - * \internal - * \brief Send a reprobe controller operation - * - * \param[in] api Controller API instance - * \param[in] target_node Name of node to reprobe - * \param[in] router_node Router node for host - * - * \return Standard Pacemaker return code - */ - int (*reprobe)(pcmk_controld_api_t *api, const char *target_node, - const char *router_node); - - /* @TODO These methods have a lot of arguments. One possibility would be to - * make a struct for agent info (standard/provider/type), which theortically - * could be used throughout pacemaker code. However that would end up being - * really awkward to use generically, since sometimes you need to allocate - * those strings (char *) and other times you only have references into XML - * (const char *). We could make some structs just for this API. - */ - - /*! - * \internal - * \brief Ask the controller to fail a resource - * - * \param[in] api Controller API instance - * \param[in] target_node Name of node resource is on - * \param[in] router_node Router node for target - * \param[in] rsc_id ID of resource to fail - * \param[in] rsc_long_id Long ID of resource (if any) - * \param[in] standard Standard of resource - * \param[in] provider Provider of resource (if any) - * \param[in] type Type of resource to fail - * - * \return Standard Pacemaker return code - */ - int (*fail_resource)(pcmk_controld_api_t *api, const char *target_node, - const char *router_node, const char *rsc_id, - const char *rsc_long_id, const char *standard, - const char *provider, const char *type); - - /*! - * \internal - * \brief Ask the controller to refresh a resource - * - * \param[in] api Controller API instance - * \param[in] target_node Name of node resource is on - * \param[in] router_node Router node for target - * \param[in] rsc_id ID of resource to refresh - * \param[in] rsc_long_id Long ID of resource (if any) - * \param[in] standard Standard of resource - * \param[in] provider Provider of resource (if any) - * \param[in] type Type of resource - * \param[in] cib_only If true, clean resource from CIB only - * - * \return Standard Pacemaker return code - */ - int (*refresh_resource)(pcmk_controld_api_t *api, const char *target_node, - const char *router_node, const char *rsc_id, - const char *rsc_long_id, const char *standard, - const char *provider, const char *type, - bool cib_only); -}; - -/*! - * \internal - * \brief Create new controller IPC API object for clients - * - * \param[in] client_name Client name to use with IPC - * \param[in] client_uuid Client UUID to use with IPC - * - * \return Newly allocated object - * \note This function asserts on errors, so it will never return NULL. - * The caller is responsible for freeing the result with - * pcmk_free_controld_api(). - */ -pcmk_controld_api_t *pcmk_new_controld_api(const char *client_name, - const char *client_uuid); - -/*! - * \internal - * \brief Free a controller IPC API object - * - * \param[in] api Controller IPC API object to free - */ -void pcmk_free_controld_api(pcmk_controld_api_t *api); - -#ifdef __cplusplus -} -#endif - -#endif diff --git a/tools/crm_resource_runtime.c b/tools/crm_resource_runtime.c index 37789d1e17..cc2abeb78c 100644 --- a/tools/crm_resource_runtime.c +++ b/tools/crm_resource_runtime.c @@ -1,2090 +1,2090 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include +#include int resource_verbose = 0; bool do_force = FALSE; const char *attr_set_type = XML_TAG_ATTR_SETS; static int do_find_resource(const char *rsc, pe_resource_t * the_rsc, pe_working_set_t * data_set) { int found = 0; GListPtr lpc = NULL; for (lpc = the_rsc->running_on; lpc != NULL; lpc = lpc->next) { pe_node_t *node = (pe_node_t *) lpc->data; if (BE_QUIET) { fprintf(stdout, "%s\n", node->details->uname); } else { const char *state = ""; if (!pe_rsc_is_clone(the_rsc) && the_rsc->fns->state(the_rsc, TRUE) == RSC_ROLE_MASTER) { state = "Master"; } fprintf(stdout, "resource %s is running on: %s %s\n", rsc, node->details->uname, state); } found++; } if (BE_QUIET == FALSE && found == 0) { fprintf(stderr, "resource %s is NOT running\n", rsc); } return found; } int cli_resource_search(pe_resource_t *rsc, const char *requested_name, pe_working_set_t *data_set) { int found = 0; pe_resource_t *parent = uber_parent(rsc); if (pe_rsc_is_clone(rsc)) { for (GListPtr iter = rsc->children; iter != NULL; iter = iter->next) { found += do_find_resource(requested_name, iter->data, data_set); } /* The anonymous clone children's common ID is supplied */ } else if (pe_rsc_is_clone(parent) && is_not_set(rsc->flags, pe_rsc_unique) && rsc->clone_name && safe_str_eq(requested_name, rsc->clone_name) && safe_str_neq(requested_name, rsc->id)) { for (GListPtr iter = parent->children; iter; iter = iter->next) { found += do_find_resource(requested_name, iter->data, data_set); } } else { found += do_find_resource(requested_name, rsc, data_set); } return found; } #define XPATH_MAX 1024 static int find_resource_attr(cib_t * the_cib, const char *attr, const char *rsc, const char *set_type, const char *set_name, const char *attr_id, const char *attr_name, char **value) { int offset = 0; int rc = pcmk_ok; xmlNode *xml_search = NULL; char *xpath_string = NULL; if(value) { *value = NULL; } if(the_cib == NULL) { return -ENOTCONN; } xpath_string = calloc(1, XPATH_MAX); offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "%s", get_object_path("resources")); offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "//*[@id=\"%s\"]", rsc); if (set_type) { offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "/%s", set_type); if (set_name) { offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "[@id=\"%s\"]", set_name); } } offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "//nvpair["); if (attr_id) { offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "@id=\"%s\"", attr_id); } if (attr_name) { if (attr_id) { offset += snprintf(xpath_string + offset, XPATH_MAX - offset, " and "); } offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "@name=\"%s\"", attr_name); } offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "]"); CRM_LOG_ASSERT(offset > 0); rc = the_cib->cmds->query(the_cib, xpath_string, &xml_search, cib_sync_call | cib_scope_local | cib_xpath); if (rc != pcmk_ok) { goto bail; } crm_log_xml_debug(xml_search, "Match"); if (xml_has_children(xml_search)) { xmlNode *child = NULL; rc = -EINVAL; printf("Multiple attributes match name=%s\n", attr_name); for (child = __xml_first_child(xml_search); child != NULL; child = __xml_next(child)) { printf(" Value: %s \t(id=%s)\n", crm_element_value(child, XML_NVPAIR_ATTR_VALUE), ID(child)); } } else if(value) { const char *tmp = crm_element_value(xml_search, attr); if (tmp) { *value = strdup(tmp); } } bail: free(xpath_string); free_xml(xml_search); return rc; } /* PRIVATE. Use the find_matching_attr_resources instead. */ static void find_matching_attr_resources_recursive(GList/* */ ** result, pe_resource_t * rsc, const char * rsc_id, const char * attr_set, const char * attr_id, const char * attr_name, cib_t * cib, const char * cmd, int depth) { int rc = pcmk_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(result, (pe_resource_t*)gIter->data, rsc_id, attr_set, attr_id, attr_name, cib, cmd, depth+1); /* do it only once for clones */ if(pe_clone == rsc->variant) { break; } } rc = find_resource_attr(cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); /* Post-order traversal. * The root is always on the list and it is the last item. */ if((0 == depth) || (pcmk_ok == rc)) { /* push the head */ *result = g_list_append(*result, rsc); } free(local_attr_id); free(lookup_id); } /* The result is a linearized pre-ordered tree of resources. */ static GList/**/ * find_matching_attr_resources(pe_resource_t * rsc, const char * rsc_id, const char * attr_set, const char * attr_id, const char * attr_name, cib_t * cib, const char * cmd) { int rc = pcmk_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(do_force == TRUE) { return g_list_append(result, rsc); } if(rsc->parent && pe_clone == rsc->parent->variant) { int rc = pcmk_ok; char *local_attr_id = NULL; rc = find_resource_attr(cib, XML_ATTR_ID, rsc_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); free(local_attr_id); if(rc != pcmk_ok) { rsc = rsc->parent; if (BE_QUIET == FALSE) { printf("Performing %s of '%s' on '%s', the parent of '%s'\n", cmd, attr_name, rsc->id, rsc_id); } } return g_list_append(result, rsc); } else if(rsc->parent == NULL && rsc->children && pe_clone == rsc->variant) { pe_resource_t *child = rsc->children->data; if(child->variant == pe_native) { lookup_id = clone_strip(child->id); /* Could be a cloned group! */ rc = find_resource_attr(cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); if(rc == pcmk_ok) { rsc = child; if (BE_QUIET == FALSE) { printf("A value for '%s' already exists in child '%s', performing %s on that instead of '%s'\n", 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(&result, rsc, rsc_id, attr_set, attr_id, attr_name, cib, cmd, 0); return result; } int cli_resource_update_attribute(pe_resource_t *rsc, const char *requested_name, const char *attr_set, const char *attr_id, const char *attr_name, const char *attr_value, bool recursive, cib_t *cib, pe_working_set_t *data_set) { int rc = pcmk_ok; static bool need_init = TRUE; char *local_attr_id = NULL; char *local_attr_set = NULL; GList/**/ *resources = NULL; const char *common_attr_id = attr_id; if(attr_id == NULL && do_force == FALSE && find_resource_attr( cib, XML_ATTR_ID, uber_parent(rsc)->id, NULL, NULL, NULL, attr_name, NULL) == -EINVAL) { printf("\n"); } if (safe_str_eq(attr_set_type, XML_TAG_ATTR_SETS)) { if (do_force == FALSE) { rc = find_resource_attr(cib, XML_ATTR_ID, uber_parent(rsc)->id, XML_TAG_META_SETS, attr_set, attr_id, attr_name, &local_attr_id); if (rc == pcmk_ok && BE_QUIET == FALSE) { printf("WARNING: There is already a meta attribute for '%s' called '%s' (id=%s)\n", uber_parent(rsc)->id, attr_name, local_attr_id); printf(" Delete '%s' first or use the force option to override\n", local_attr_id); } free(local_attr_id); if (rc == pcmk_ok) { return -ENOTUNIQ; } } resources = g_list_append(resources, rsc); } else { resources = find_matching_attr_resources(rsc, requested_name, attr_set, attr_id, attr_name, cib, "update"); } /* If either attr_set or attr_id is specified, * one clearly intends to modify a single resource. * It is the last item on the resource list.*/ for(GList *gIter = (attr_set||attr_id) ? g_list_last(resources) : resources ; gIter; gIter = gIter->next) { char *lookup_id = NULL; xmlNode *xml_top = NULL; xmlNode *xml_obj = NULL; local_attr_id = NULL; local_attr_set = NULL; rsc = (pe_resource_t*)gIter->data; attr_id = common_attr_id; lookup_id = clone_strip(rsc->id); /* Could be a cloned group! */ rc = find_resource_attr(cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); if (rc == pcmk_ok) { crm_debug("Found a match for name=%s: id=%s", attr_name, local_attr_id); attr_id = local_attr_id; } else if (rc != -ENXIO) { free(lookup_id); free(local_attr_id); g_list_free(resources); return rc; } else { const char *tag = crm_element_name(rsc->xml); if (attr_set == NULL) { local_attr_set = crm_strdup_printf("%s-%s", lookup_id, attr_set_type); attr_set = local_attr_set; } if (attr_id == NULL) { local_attr_id = crm_strdup_printf("%s-%s", attr_set, attr_name); attr_id = local_attr_id; } xml_top = create_xml_node(NULL, tag); crm_xml_add(xml_top, XML_ATTR_ID, lookup_id); xml_obj = create_xml_node(xml_top, attr_set_type); crm_xml_add(xml_obj, XML_ATTR_ID, attr_set); } xml_obj = crm_create_nvpair_xml(xml_obj, attr_id, attr_name, attr_value); if (xml_top == NULL) { xml_top = xml_obj; } crm_log_xml_debug(xml_top, "Update"); rc = cib->cmds->modify(cib, XML_CIB_TAG_RESOURCES, xml_top, cib_options); if (rc == pcmk_ok && BE_QUIET == FALSE) { printf("Set '%s' option: id=%s%s%s%s%s value=%s\n", lookup_id, local_attr_id, attr_set ? " set=" : "", attr_set ? attr_set : "", attr_name ? " name=" : "", attr_name ? attr_name : "", attr_value); } free_xml(xml_top); free(lookup_id); free(local_attr_id); free(local_attr_set); if(recursive && safe_str_eq(attr_set_type, XML_TAG_META_SETS)) { GListPtr lpc = NULL; if(need_init) { xmlNode *cib_constraints = get_object_root(XML_CIB_TAG_CONSTRAINTS, data_set->input); need_init = FALSE; unpack_constraints(cib_constraints, data_set); for (lpc = data_set->resources; lpc != NULL; lpc = lpc->next) { pe_resource_t *r = (pe_resource_t *) lpc->data; clear_bit(r->flags, pe_rsc_allocating); } } crm_debug("Looking for dependencies %p", rsc->rsc_cons_lhs); set_bit(rsc->flags, pe_rsc_allocating); for (lpc = rsc->rsc_cons_lhs; lpc != NULL; lpc = lpc->next) { rsc_colocation_t *cons = (rsc_colocation_t *) lpc->data; pe_resource_t *peer = cons->rsc_lh; crm_debug("Checking %s %d", cons->id, cons->score); if (cons->score > 0 && is_not_set(peer->flags, pe_rsc_allocating)) { /* Don't get into colocation loops */ crm_debug("Setting %s=%s for dependent resource %s", attr_name, attr_value, peer->id); cli_resource_update_attribute(peer, peer->id, NULL, NULL, attr_name, attr_value, recursive, cib, data_set); } } } } g_list_free(resources); return rc; } int cli_resource_delete_attribute(pe_resource_t *rsc, const char *requested_name, const char *attr_set, const char *attr_id, const char *attr_name, cib_t *cib, pe_working_set_t *data_set) { int rc = pcmk_ok; GList/**/ *resources = NULL; if(attr_id == NULL && do_force == FALSE && find_resource_attr( cib, XML_ATTR_ID, uber_parent(rsc)->id, NULL, NULL, NULL, attr_name, NULL) == -EINVAL) { printf("\n"); } if(safe_str_eq(attr_set_type, XML_TAG_META_SETS)) { resources = find_matching_attr_resources(rsc, requested_name, attr_set, attr_id, attr_name, cib, "delete"); } else { resources = g_list_append(resources, rsc); } for(GList *gIter = resources; gIter; gIter = gIter->next) { char *lookup_id = NULL; xmlNode *xml_obj = NULL; char *local_attr_id = NULL; rsc = (pe_resource_t*)gIter->data; lookup_id = clone_strip(rsc->id); rc = find_resource_attr(cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); if (rc == -ENXIO) { free(lookup_id); rc = pcmk_ok; continue; } else if (rc != pcmk_ok) { free(lookup_id); g_list_free(resources); return rc; } if (attr_id == NULL) { attr_id = local_attr_id; } xml_obj = crm_create_nvpair_xml(NULL, attr_id, attr_name, NULL); crm_log_xml_debug(xml_obj, "Delete"); CRM_ASSERT(cib); rc = cib->cmds->remove(cib, XML_CIB_TAG_RESOURCES, xml_obj, cib_options); if (rc == pcmk_ok && BE_QUIET == FALSE) { printf("Deleted '%s' option: id=%s%s%s%s%s\n", lookup_id, local_attr_id, attr_set ? " set=" : "", attr_set ? attr_set : "", attr_name ? " name=" : "", attr_name ? attr_name : ""); } free(lookup_id); free_xml(xml_obj); free(local_attr_id); } g_list_free(resources); return rc; } // \return Standard Pacemaker return code static int -send_lrm_rsc_op(pcmk_controld_api_t *controld_api, bool do_fail_resource, +send_lrm_rsc_op(pcmk_ipc_api_t *controld_api, bool do_fail_resource, const char *host_uname, const char *rsc_id, pe_working_set_t *data_set) { const char *router_node = host_uname; const char *rsc_api_id = NULL; const char *rsc_long_id = NULL; const char *rsc_class = NULL; const char *rsc_provider = NULL; const char *rsc_type = NULL; bool cib_only = false; pe_resource_t *rsc = pe_find_resource(data_set->resources, rsc_id); if (rsc == NULL) { CMD_ERR("Resource %s not found", rsc_id); return ENXIO; } else if (rsc->variant != pe_native) { CMD_ERR("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)) { CMD_ERR("Resource %s does not have a class and type", rsc_id); return EINVAL; } if (host_uname == NULL) { CMD_ERR("Please specify a node name"); return EINVAL; } else { pe_node_t *node = pe_find_node(data_set->nodes, host_uname); if (node == NULL) { CMD_ERR("Node %s not found", host_uname); return pcmk_rc_node_unknown; } if (!(node->details->online)) { if (do_fail_resource) { CMD_ERR("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) { CMD_ERR("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 controld_api->fail_resource(controld_api, host_uname, - router_node, rsc_api_id, rsc_long_id, - rsc_class, rsc_provider, rsc_type); + 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 controld_api->refresh_resource(controld_api, host_uname, - router_node, rsc_api_id, - rsc_long_id, rsc_class, - rsc_provider, rsc_type, cib_only); + return pcmk_controld_api_refresh(controld_api, host_uname, router_node, + rsc_api_id, rsc_long_id, rsc_class, + rsc_provider, rsc_type, cib_only); } } /*! * \internal * \brief Get resource name as used in failure-related node attributes * * \param[in] rsc Resource to check * * \return Newly allocated string containing resource's fail name * \note The caller is responsible for freeing the result. */ static inline char * rsc_fail_name(pe_resource_t *rsc) { const char *name = (rsc->clone_name? rsc->clone_name : rsc->id); return is_set(rsc->flags, pe_rsc_unique)? strdup(name) : clone_strip(name); } // \return Standard Pacemaker return code static int -clear_rsc_history(pcmk_controld_api_t *controld_api, const char *host_uname, +clear_rsc_history(pcmk_ipc_api_t *controld_api, const char *host_uname, const char *rsc_id, pe_working_set_t *data_set) { int rc = pcmk_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, data_set); if (rc != pcmk_rc_ok) { return rc; } crm_trace("Processing %d mainloop inputs", - controld_api->replies_expected(controld_api)); + pcmk_controld_api_replies_expected(controld_api)); while (g_main_context_iteration(NULL, FALSE)) { crm_trace("Processed mainloop input, %d still remaining", - controld_api->replies_expected(controld_api)); + pcmk_controld_api_replies_expected(controld_api)); } return rc; } static int -clear_rsc_failures(pcmk_controld_api_t *controld_api, const char *node_name, +clear_rsc_failures(pcmk_ipc_api_t *controld_api, const char *node_name, const char *rsc_id, const char *operation, const char *interval_spec, pe_working_set_t *data_set) { int rc = pcmk_ok; const char *failed_value = NULL; const char *failed_id = NULL; const char *interval_ms_s = NULL; GHashTable *rscs = NULL; GHashTableIter iter; /* Create a hash table to use as a set of resources to clean. This lets us * clean each resource only once (per node) regardless of how many failed * operations it has. */ rscs = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, NULL); // Normalize interval to milliseconds for comparison to history entry if (operation) { interval_ms_s = crm_strdup_printf("%u", crm_parse_interval_spec(interval_spec)); } for (xmlNode *xml_op = __xml_first_child(data_set->failed); xml_op != NULL; xml_op = __xml_next(xml_op)) { failed_id = crm_element_value(xml_op, XML_LRM_ATTR_RSCID); if (failed_id == NULL) { // Malformed history entry, should never happen continue; } // No resource specified means all resources match if (rsc_id) { pe_resource_t *fail_rsc = pe_find_resource_with_flags(data_set->resources, failed_id, pe_find_renamed|pe_find_anon); if (!fail_rsc || safe_str_neq(rsc_id, fail_rsc->id)) { continue; } } // Host name should always have been provided by this point failed_value = crm_element_value(xml_op, XML_ATTR_UNAME); if (safe_str_neq(node_name, failed_value)) { continue; } // No operation specified means all operations match if (operation) { failed_value = crm_element_value(xml_op, XML_LRM_ATTR_TASK); if (safe_str_neq(operation, failed_value)) { continue; } // Interval (if operation was specified) defaults to 0 (not all) failed_value = crm_element_value(xml_op, XML_LRM_ATTR_INTERVAL_MS); if (safe_str_neq(interval_ms_s, failed_value)) { continue; } } /* not available until glib 2.32 g_hash_table_add(rscs, (gpointer) failed_id); */ g_hash_table_insert(rscs, (gpointer) failed_id, (gpointer) failed_id); } g_hash_table_iter_init(&iter, rscs); while (g_hash_table_iter_next(&iter, (gpointer *) &failed_id, NULL)) { crm_debug("Erasing failures of %s on %s", failed_id, node_name); rc = clear_rsc_history(controld_api, node_name, failed_id, data_set); if (rc != pcmk_rc_ok) { return pcmk_rc2legacy(rc); } } g_hash_table_destroy(rscs); return rc; } static int clear_rsc_fail_attrs(pe_resource_t *rsc, const char *operation, const char *interval_spec, pe_node_t *node) { int rc = pcmk_ok; int attr_options = pcmk__node_attr_none; char *rsc_name = rsc_fail_name(rsc); if (pe__is_guest_or_remote_node(node)) { attr_options |= pcmk__node_attr_remote; } rc = pcmk__node_attr_request_clear(NULL, node->details->uname, rsc_name, operation, interval_spec, NULL, attr_options); free(rsc_name); return rc; } int -cli_resource_delete(pcmk_controld_api_t *controld_api, const char *host_uname, +cli_resource_delete(pcmk_ipc_api_t *controld_api, const char *host_uname, pe_resource_t *rsc, const char *operation, const char *interval_spec, bool just_failures, pe_working_set_t *data_set) { int rc = pcmk_ok; pe_node_t *node = NULL; if (rsc == NULL) { return -ENXIO; } else if (rsc->children) { GListPtr lpc = NULL; for (lpc = rsc->children; lpc != NULL; lpc = lpc->next) { pe_resource_t *child = (pe_resource_t *) lpc->data; rc = cli_resource_delete(controld_api, host_uname, child, operation, interval_spec, just_failures, data_set); if (rc != pcmk_ok) { return rc; } } return pcmk_ok; } else if (host_uname == NULL) { GListPtr lpc = NULL; GListPtr nodes = g_hash_table_get_values(rsc->known_on); if(nodes == NULL && do_force) { nodes = pcmk__copy_node_list(data_set->nodes, false); } else if(nodes == NULL && rsc->exclusive_discover) { GHashTableIter iter; pe_node_t *node = NULL; g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void**)&node)) { if(node->weight >= 0) { nodes = g_list_prepend(nodes, node); } } } else if(nodes == NULL) { nodes = g_hash_table_get_values(rsc->allowed_nodes); } for (lpc = nodes; lpc != NULL; lpc = lpc->next) { node = (pe_node_t *) lpc->data; if (node->details->online) { rc = cli_resource_delete(controld_api, node->details->uname, rsc, operation, interval_spec, just_failures, data_set); } if (rc != pcmk_ok) { g_list_free(nodes); return rc; } } g_list_free(nodes); return pcmk_ok; } node = pe_find_node(data_set->nodes, host_uname); if (node == NULL) { printf("Unable to clean up %s because node %s not found\n", rsc->id, host_uname); return -ENODEV; } if (!node->details->rsc_discovery_enabled) { printf("Unable to clean up %s because resource discovery disabled on %s\n", rsc->id, host_uname); return -EOPNOTSUPP; } if (controld_api == NULL) { printf("Dry run: skipping clean-up of %s on %s due to CIB_file\n", rsc->id, host_uname); return pcmk_ok; } rc = clear_rsc_fail_attrs(rsc, operation, interval_spec, node); if (rc != pcmk_rc_ok) { printf("Unable to clean up %s failures on %s: %s\n", rsc->id, host_uname, pcmk_rc_str(rc)); return pcmk_rc2legacy(rc); } if (just_failures) { rc = clear_rsc_failures(controld_api, host_uname, rsc->id, operation, interval_spec, data_set); } else { rc = clear_rsc_history(controld_api, host_uname, rsc->id, data_set); rc = pcmk_rc2legacy(rc); } if (rc != pcmk_ok) { printf("Cleaned %s failures on %s, but unable to clean history: %s\n", rsc->id, host_uname, pcmk_strerror(rc)); } else { printf("Cleaned up %s on %s\n", rsc->id, host_uname); } return rc; } int -cli_cleanup_all(pcmk_controld_api_t *controld_api, const char *node_name, +cli_cleanup_all(pcmk_ipc_api_t *controld_api, const char *node_name, const char *operation, const char *interval_spec, pe_working_set_t *data_set) { int rc = pcmk_ok; int attr_options = pcmk__node_attr_none; const char *display_name = node_name? node_name : "all nodes"; if (controld_api == NULL) { printf("Dry run: skipping clean-up of %s due to CIB_file\n", display_name); return pcmk_ok; } if (node_name) { pe_node_t *node = pe_find_node(data_set->nodes, node_name); if (node == NULL) { CMD_ERR("Unknown node: %s", node_name); return -ENXIO; } if (pe__is_guest_or_remote_node(node)) { attr_options |= pcmk__node_attr_remote; } } rc = pcmk__node_attr_request_clear(NULL, node_name, NULL, operation, interval_spec, NULL, attr_options); if (rc != pcmk_rc_ok) { printf("Unable to clean up all failures on %s: %s\n", display_name, pcmk_rc_str(rc)); return pcmk_rc2legacy(rc); } if (node_name) { rc = clear_rsc_failures(controld_api, node_name, NULL, operation, interval_spec, data_set); if (rc != pcmk_ok) { printf("Cleaned all resource failures on %s, but unable to clean history: %s\n", node_name, pcmk_strerror(rc)); return rc; } } else { for (GList *iter = data_set->nodes; iter; iter = iter->next) { pe_node_t *node = (pe_node_t *) iter->data; rc = clear_rsc_failures(controld_api, node->details->uname, NULL, operation, interval_spec, data_set); if (rc != pcmk_ok) { printf("Cleaned all resource failures on all nodes, but unable to clean history: %s\n", pcmk_strerror(rc)); return rc; } } } printf("Cleaned up all resources on %s\n", display_name); return pcmk_ok; } void cli_resource_check(cib_t * cib_conn, pe_resource_t *rsc) { bool printed = false; char *role_s = NULL; char *managed = NULL; pe_resource_t *parent = uber_parent(rsc); find_resource_attr(cib_conn, XML_NVPAIR_ATTR_VALUE, parent->id, NULL, NULL, NULL, XML_RSC_ATTR_MANAGED, &managed); find_resource_attr(cib_conn, XML_NVPAIR_ATTR_VALUE, parent->id, NULL, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, &role_s); if(role_s) { enum rsc_role_e role = text2role(role_s); free(role_s); if(role == RSC_ROLE_UNKNOWN) { // Treated as if unset } else if(role == RSC_ROLE_STOPPED) { printf("\n * Configuration specifies '%s' should remain stopped\n", parent->id); printed = true; } else if (is_set(parent->flags, pe_rsc_promotable) && (role == RSC_ROLE_SLAVE)) { printf("\n * Configuration specifies '%s' should not be promoted\n", parent->id); printed = true; } } if (managed && !crm_is_true(managed)) { printf("%s * Configuration prevents cluster from stopping or starting unmanaged '%s'\n", (printed? "" : "\n"), parent->id); printed = true; } free(managed); if (rsc->lock_node) { printf("%s * '%s' is locked to node %s due to shutdown\n", (printed? "" : "\n"), parent->id, rsc->lock_node->details->uname); } if (printed) { printf("\n"); } } // \return Standard Pacemaker return code int -cli_resource_fail(pcmk_controld_api_t *controld_api, const char *host_uname, +cli_resource_fail(pcmk_ipc_api_t *controld_api, const char *host_uname, const char *rsc_id, pe_working_set_t *data_set) { crm_notice("Failing %s on %s", rsc_id, host_uname); return send_lrm_rsc_op(controld_api, true, host_uname, rsc_id, data_set); } static GHashTable * generate_resource_params(pe_resource_t * rsc, pe_working_set_t * data_set) { GHashTable *params = NULL; GHashTable *meta = NULL; GHashTable *combined = NULL; GHashTableIter iter; if (!rsc) { crm_err("Resource does not exist in config"); return NULL; } params = crm_str_table_new(); meta = crm_str_table_new(); combined = crm_str_table_new(); get_rsc_attributes(params, rsc, NULL /* TODO: Pass in local node */ , data_set); get_meta_attributes(meta, rsc, NULL /* TODO: Pass in local node */ , data_set); if (params) { char *key = NULL; char *value = NULL; g_hash_table_iter_init(&iter, params); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { g_hash_table_insert(combined, strdup(key), strdup(value)); } g_hash_table_destroy(params); } if (meta) { char *key = NULL; char *value = NULL; g_hash_table_iter_init(&iter, meta); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { char *crm_name = crm_meta_name(key); g_hash_table_insert(combined, crm_name, strdup(value)); } g_hash_table_destroy(meta); } return combined; } static bool resource_is_running_on(pe_resource_t *rsc, const char *host) { bool found = TRUE; GListPtr hIter = NULL; GListPtr hosts = NULL; if(rsc == NULL) { return FALSE; } rsc->fns->location(rsc, &hosts, TRUE); for (hIter = hosts; host != NULL && hIter != NULL; hIter = hIter->next) { pe_node_t *node = (pe_node_t *) hIter->data; if(strcmp(host, node->details->uname) == 0) { crm_trace("Resource %s is running on %s\n", rsc->id, host); goto done; } else if(strcmp(host, node->details->id) == 0) { crm_trace("Resource %s is running on %s\n", rsc->id, host); goto done; } } if(host != NULL) { crm_trace("Resource %s is not running on: %s\n", rsc->id, host); found = FALSE; } else if(host == NULL && hosts == NULL) { crm_trace("Resource %s is not running\n", rsc->id); found = FALSE; } done: g_list_free(hosts); return found; } /*! * \internal * \brief Create a list of all resources active on host from a given list * * \param[in] host Name of host to check whether resources are active * \param[in] rsc_list List of resources to check * * \return New list of resources from list that are active on host */ static GList * get_active_resources(const char *host, GList *rsc_list) { GList *rIter = NULL; GList *active = NULL; for (rIter = rsc_list; rIter != NULL; rIter = rIter->next) { pe_resource_t *rsc = (pe_resource_t *) rIter->data; /* Expand groups to their members, because if we're restarting a member * other than the first, we can't otherwise tell which resources are * stopping and starting. */ if (rsc->variant == pe_group) { active = g_list_concat(active, get_active_resources(host, rsc->children)); } else if (resource_is_running_on(rsc, host)) { active = g_list_append(active, strdup(rsc->id)); } } return active; } GList* subtract_lists(GList *from, GList *items, GCompareFunc cmp) { GList *item = NULL; GList *result = g_list_copy(from); for (item = items; item != NULL; item = item->next) { GList *candidate = NULL; for (candidate = from; candidate != NULL; candidate = candidate->next) { crm_info("Comparing %s with %s", (const char *) candidate->data, (const char *) item->data); if(cmp(candidate->data, item->data) == 0) { result = g_list_remove(result, candidate->data); break; } } } return result; } 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(GList *items, const char *tag) { GList *item = NULL; for (item = items; item != NULL; item = item->next) { fprintf(stdout, "%s%s\n", tag, (const char *)item->data); } } /*! * \internal * \brief Upgrade XML to latest schema version and use it as working set input * * This also updates the working set timestamp to the current time. * * \param[in] data_set Working set instance to update * \param[in] xml XML to use as input * * \return pcmk_ok on success, -ENOKEY if unable to upgrade XML * \note On success, caller is responsible for freeing memory allocated for * data_set->now. * \todo This follows the example of other callers of cli_config_update() * and returns -ENOKEY ("Required key not available") if that fails, * but perhaps -pcmk_err_schema_validation would be better in that case. */ int update_working_set_xml(pe_working_set_t *data_set, xmlNode **xml) { if (cli_config_update(xml, NULL, FALSE) == FALSE) { return -ENOKEY; } data_set->input = *xml; data_set->now = crm_time_new(NULL); return pcmk_ok; } /*! * \internal * \brief Update a working set's XML input based on a CIB query * * \param[in] data_set Data set instance to initialize * \param[in] cib Connection to the CIB manager * * \return pcmk_ok on success, -errno on failure * \note On success, caller is responsible for freeing memory allocated for * data_set->input and data_set->now. */ static int update_working_set_from_cib(pe_working_set_t * data_set, cib_t *cib) { xmlNode *cib_xml_copy = NULL; int rc; rc = cib->cmds->query(cib, NULL, &cib_xml_copy, cib_scope_local | cib_sync_call); if (rc != pcmk_ok) { fprintf(stderr, "Could not obtain the current CIB: %s (%d)\n", pcmk_strerror(rc), rc); return rc; } rc = update_working_set_xml(data_set, &cib_xml_copy); if (rc != pcmk_ok) { fprintf(stderr, "Could not upgrade the current CIB XML\n"); free_xml(cib_xml_copy); return rc; } return pcmk_ok; } static int update_dataset(cib_t *cib, pe_working_set_t * data_set, bool simulate) { char *pid = NULL; char *shadow_file = NULL; cib_t *shadow_cib = NULL; int rc; pe_reset_working_set(data_set); rc = update_working_set_from_cib(data_set, cib); if (rc != pcmk_ok) { return rc; } if(simulate) { pid = pcmk__getpid_s(); shadow_cib = cib_shadow_new(pid); shadow_file = get_shadow_file(pid); if (shadow_cib == NULL) { fprintf(stderr, "Could not create shadow cib: '%s'\n", pid); rc = -ENXIO; goto cleanup; } rc = write_xml_file(data_set->input, shadow_file, FALSE); if (rc < 0) { fprintf(stderr, "Could not populate shadow cib: %s (%d)\n", pcmk_strerror(rc), rc); goto cleanup; } rc = shadow_cib->cmds->signon(shadow_cib, crm_system_name, cib_command); if(rc != pcmk_ok) { fprintf(stderr, "Could not connect to shadow cib: %s (%d)\n", pcmk_strerror(rc), rc); goto cleanup; } pcmk__schedule_actions(data_set, data_set->input, NULL); run_simulation(data_set, shadow_cib, NULL, TRUE); rc = update_dataset(shadow_cib, data_set, FALSE); } else { cluster_status(data_set); } cleanup: /* Do not free data_set->input here, we need rsc->xml to be valid later on */ cib_delete(shadow_cib); free(pid); if(shadow_file) { unlink(shadow_file); free(shadow_file); } return rc; } static int max_delay_for_resource(pe_working_set_t * data_set, pe_resource_t *rsc) { int delay = 0; int max_delay = 0; if(rsc && rsc->children) { GList *iter = NULL; for(iter = rsc->children; iter; iter = iter->next) { pe_resource_t *child = (pe_resource_t *)iter->data; delay = max_delay_for_resource(data_set, child); if(delay > max_delay) { double seconds = delay / 1000.0; crm_trace("Calculated new delay of %.1fs due to %s", seconds, child->id); max_delay = delay; } } } else if(rsc) { char *key = crm_strdup_printf("%s_%s_0", rsc->id, RSC_STOP); pe_action_t *stop = custom_action(rsc, key, RSC_STOP, NULL, TRUE, FALSE, data_set); const char *value = g_hash_table_lookup(stop->meta, XML_ATTR_TIMEOUT); max_delay = value? (int) crm_parse_ll(value, NULL) : -1; pe_free_action(stop); } return max_delay; } static int max_delay_in(pe_working_set_t * data_set, GList *resources) { int max_delay = 0; GList *item = NULL; for (item = resources; item != NULL; item = item->next) { int delay = 0; pe_resource_t *rsc = pe_find_resource(data_set->resources, (const char *)item->data); delay = max_delay_for_resource(data_set, rsc); if(delay > max_delay) { double seconds = delay / 1000.0; crm_trace("Calculated new delay of %.1fs due to %s", seconds, rsc->id); max_delay = delay; } } return 5 + (max_delay / 1000); } #define waiting_for_starts(d, r, h) ((d != NULL) || \ (resource_is_running_on((r), (h)) == FALSE)) /*! * \internal * \brief Restart a resource (on a particular host if requested). * * \param[in] rsc The resource to restart * \param[in] host The host to restart the resource on (or NULL for all) * \param[in] timeout_ms Consider failed if actions do not complete in this time * (specified in milliseconds, but a two-second * granularity is actually used; if 0, a timeout will be * calculated based on the resource timeout) * \param[in] cib Connection to the CIB manager * * \return pcmk_ok on success, -errno on failure (exits on certain failures) */ int cli_resource_restart(pe_resource_t *rsc, const char *host, int timeout_ms, cib_t *cib) { int rc = 0; int lpc = 0; int before = 0; int step_timeout_s = 0; int sleep_interval = 2; int timeout = timeout_ms / 1000; bool stop_via_ban = FALSE; char *rsc_id = NULL; char *orig_target_role = NULL; GList *list_delta = NULL; GList *target_active = NULL; GList *current_active = NULL; GList *restart_target_active = NULL; pe_working_set_t *data_set = NULL; if(resource_is_running_on(rsc, host) == FALSE) { const char *id = rsc->clone_name?rsc->clone_name:rsc->id; if(host) { printf("%s is not running on %s and so cannot be restarted\n", id, host); } else { printf("%s is not running anywhere and so cannot be restarted\n", id); } return -ENXIO; } /* We might set the target-role meta-attribute */ attr_set_type = XML_TAG_META_SETS; rsc_id = strdup(rsc->id); if ((pe_rsc_is_clone(rsc) || pe_bundle_replicas(rsc)) && host) { stop_via_ban = TRUE; } /* grab full cib determine originally active resources disable or ban poll cib and watch for affected resources to get stopped without --timeout, calculate the stop timeout for each step and wait for that if we hit --timeout or the service timeout, re-enable or un-ban, report failure and indicate which resources we couldn't take down if everything stopped, re-enable or un-ban poll cib and watch for affected resources to get started without --timeout, calculate the start timeout for each step and wait for that if we hit --timeout or the service timeout, report (different) failure and indicate which resources we couldn't bring back up report success Optimizations: - use constraints to determine ordered list of affected resources - Allow a --no-deps option (aka. --force-restart) */ data_set = pe_new_working_set(); if (data_set == NULL) { crm_perror(LOG_ERR, "Could not allocate working set"); rc = -ENOMEM; goto done; } set_bit(data_set->flags, pe_flag_no_counts); set_bit(data_set->flags, pe_flag_no_compat); rc = update_dataset(cib, data_set, FALSE); if(rc != pcmk_ok) { fprintf(stdout, "Could not get new resource list: %s (%d)\n", pcmk_strerror(rc), rc); goto done; } restart_target_active = get_active_resources(host, data_set->resources); current_active = get_active_resources(host, data_set->resources); dump_list(current_active, "Origin"); if (stop_via_ban) { /* Stop the clone or bundle instance by banning it from the host */ BE_QUIET = TRUE; rc = cli_resource_ban(rsc_id, host, NULL, cib); } else { /* Stop the resource by setting target-role to Stopped. * Remember any existing target-role so we can restore it later * (though it only makes any difference if it's Slave). */ char *lookup_id = clone_strip(rsc->id); find_resource_attr(cib, XML_NVPAIR_ATTR_VALUE, lookup_id, NULL, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, &orig_target_role); free(lookup_id); rc = cli_resource_update_attribute(rsc, rsc_id, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, RSC_STOPPED, FALSE, cib, data_set); } if(rc != pcmk_ok) { fprintf(stderr, "Could not set target-role for %s: %s (%d)\n", rsc_id, pcmk_strerror(rc), rc); if (current_active) { g_list_free_full(current_active, free); } if (restart_target_active) { g_list_free_full(restart_target_active, free); } goto done; } rc = update_dataset(cib, data_set, TRUE); if(rc != pcmk_ok) { fprintf(stderr, "Could not determine which resources would be stopped\n"); goto failure; } target_active = get_active_resources(host, data_set->resources); dump_list(target_active, "Target"); list_delta = subtract_lists(current_active, target_active, (GCompareFunc) strcmp); fprintf(stdout, "Waiting for %d resources to stop:\n", g_list_length(list_delta)); display_list(list_delta, " * "); step_timeout_s = timeout / sleep_interval; while (list_delta != NULL) { before = g_list_length(list_delta); if(timeout_ms == 0) { step_timeout_s = max_delay_in(data_set, list_delta) / sleep_interval; } /* We probably don't need the entire step timeout */ for(lpc = 0; (lpc < step_timeout_s) && (list_delta != NULL); lpc++) { sleep(sleep_interval); if(timeout) { timeout -= sleep_interval; crm_trace("%ds remaining", timeout); } rc = update_dataset(cib, data_set, FALSE); if(rc != pcmk_ok) { fprintf(stderr, "Could not determine which resources were stopped\n"); goto failure; } if (current_active) { g_list_free_full(current_active, free); } current_active = get_active_resources(host, data_set->resources); g_list_free(list_delta); list_delta = 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 */ fprintf(stderr, "Could not complete shutdown of %s, %d resources remaining\n", rsc_id, g_list_length(list_delta)); display_list(list_delta, " * "); rc = -ETIME; goto failure; } } if (stop_via_ban) { rc = cli_resource_clear(rsc_id, host, NULL, cib, TRUE); } else if (orig_target_role) { rc = cli_resource_update_attribute(rsc, rsc_id, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, orig_target_role, FALSE, cib, data_set); free(orig_target_role); orig_target_role = NULL; } else { rc = cli_resource_delete_attribute(rsc, rsc_id, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, cib, data_set); } if(rc != pcmk_ok) { fprintf(stderr, "Could not unset target-role for %s: %s (%d)\n", rsc_id, pcmk_strerror(rc), rc); goto done; } if (target_active) { g_list_free_full(target_active, free); } target_active = restart_target_active; list_delta = subtract_lists(target_active, current_active, (GCompareFunc) strcmp); fprintf(stdout, "Waiting for %d resources to start again:\n", g_list_length(list_delta)); display_list(list_delta, " * "); step_timeout_s = timeout / sleep_interval; while (waiting_for_starts(list_delta, rsc, host)) { before = g_list_length(list_delta); if(timeout_ms == 0) { step_timeout_s = max_delay_in(data_set, list_delta) / sleep_interval; } /* We probably don't need the entire step timeout */ for (lpc = 0; (lpc < step_timeout_s) && waiting_for_starts(list_delta, rsc, host); lpc++) { sleep(sleep_interval); if(timeout) { timeout -= sleep_interval; crm_trace("%ds remaining", timeout); } rc = update_dataset(cib, data_set, FALSE); if(rc != pcmk_ok) { fprintf(stderr, "Could not determine which resources were started\n"); goto failure; } if (current_active) { g_list_free_full(current_active, free); } /* It's OK if dependent resources moved to a different node, * so we check active resources on all nodes. */ current_active = get_active_resources(NULL, data_set->resources); g_list_free(list_delta); list_delta = 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 */ fprintf(stdout, "Could not complete restart of %s, %d resources remaining\n", rsc_id, g_list_length(list_delta)); display_list(list_delta, " * "); rc = -ETIME; goto failure; } } rc = pcmk_ok; goto done; failure: if (stop_via_ban) { cli_resource_clear(rsc_id, host, NULL, cib, TRUE); } else if (orig_target_role) { cli_resource_update_attribute(rsc, rsc_id, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, orig_target_role, FALSE, cib, data_set); free(orig_target_role); } else { cli_resource_delete_attribute(rsc, rsc_id, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, cib, data_set); } done: if (list_delta) { g_list_free(list_delta); } if (current_active) { g_list_free_full(current_active, free); } if (target_active && (target_active != restart_target_active)) { g_list_free_full(target_active, free); } if (restart_target_active) { g_list_free_full(restart_target_active, free); } free(rsc_id); pe_free_working_set(data_set); return rc; } static inline int action_is_pending(pe_action_t *action) { if(is_set(action->flags, pe_action_optional)) { return FALSE; } else if(is_set(action->flags, pe_action_runnable) == FALSE) { return FALSE; } else if(is_set(action->flags, pe_action_pseudo)) { return FALSE; } else if(safe_str_eq("notify", action->task)) { return FALSE; } return TRUE; } /*! * \internal * \brief Return TRUE if any actions in a list are pending * * \param[in] actions List of actions to check * * \return TRUE if any actions in the list are pending, FALSE otherwise */ static bool actions_are_pending(GListPtr actions) { GListPtr action; for (action = actions; action != NULL; action = action->next) { pe_action_t *a = (pe_action_t *)action->data; if (action_is_pending(a)) { crm_notice("Waiting for %s (flags=0x%.8x)", a->uuid, a->flags); return TRUE; } } return FALSE; } /*! * \internal * \brief Print pending actions to stderr * * \param[in] actions List of actions to check * * \return void */ static void print_pending_actions(GListPtr actions) { GListPtr action; fprintf(stderr, "Pending actions:\n"); for (action = actions; action != NULL; action = action->next) { pe_action_t *a = (pe_action_t *) action->data; if (action_is_pending(a)) { fprintf(stderr, "\tAction %d: %s", a->id, a->uuid); if (a->node) { fprintf(stderr, "\ton %s", a->node->details->uname); } fprintf(stderr, "\n"); } } } /* For --wait, timeout (in seconds) to use if caller doesn't specify one */ #define WAIT_DEFAULT_TIMEOUT_S (60 * 60) /* For --wait, how long to sleep between cluster state checks */ #define WAIT_SLEEP_S (2) /*! * \internal * \brief Wait until all pending cluster actions are complete * * This waits until either the CIB's transition graph is idle or a timeout is * reached. * * \param[in] timeout_ms Consider failed if actions do not complete in this time * (specified in milliseconds, but one-second granularity * is actually used; if 0, a default will be used) * \param[in] cib Connection to the CIB manager * * \return pcmk_ok on success, -errno on failure */ int wait_till_stable(int timeout_ms, cib_t * cib) { pe_working_set_t *data_set = NULL; int rc = -1; 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 = BE_QUIET; // i.e. don't print if quiet data_set = pe_new_working_set(); if (data_set == NULL) { return -ENOMEM; } set_bit(data_set->flags, pe_flag_no_counts); set_bit(data_set->flags, pe_flag_no_compat); do { /* Abort if timeout is reached */ time_diff = expire_time - time(NULL); if (time_diff > 0) { crm_info("Waiting up to %ld seconds for cluster actions to complete", time_diff); } else { print_pending_actions(data_set->actions); pe_free_working_set(data_set); return -ETIME; } if (rc == pcmk_ok) { /* this avoids sleep on first loop iteration */ sleep(WAIT_SLEEP_S); } /* Get latest transition graph */ pe_reset_working_set(data_set); rc = update_working_set_from_cib(data_set, cib); if (rc != pcmk_ok) { pe_free_working_set(data_set); return rc; } pcmk__schedule_actions(data_set, data_set->input, NULL); if (!printed_version_warning) { /* If the DC has a different version than the local node, the two * could come to different conclusions about what actions need to be * done. Warn the user in this case. * * @TODO A possible long-term solution would be to reimplement the * wait as a new controller operation that would be forwarded to the * DC. However, that would have potential problems of its own. */ const char *dc_version = g_hash_table_lookup(data_set->config_hash, "dc-version"); if (safe_str_neq(dc_version, PACEMAKER_VERSION "-" BUILD_VERSION)) { printf("warning: wait option may not work properly in " "mixed-version cluster\n"); printed_version_warning = TRUE; } } } while (actions_are_pending(data_set->actions)); pe_free_working_set(data_set); return pcmk_ok; } int cli_resource_execute_from_params(const char *rsc_name, const char *rsc_class, const char *rsc_prov, const char *rsc_type, const char *action, GHashTable *params, GHashTable *override_hash, int timeout_ms) { GHashTable *params_copy = NULL; int rc = pcmk_ok; svc_action_t *op = NULL; if (safe_str_eq(rsc_class, PCMK_RESOURCE_CLASS_STONITH)) { CMD_ERR("Sorry, the %s option doesn't support %s resources yet", action, rsc_class); crm_exit(CRM_EX_UNIMPLEMENT_FEATURE); } /* If no timeout was provided, grab the default. */ if (timeout_ms == 0) { timeout_ms = crm_get_msec(CRM_DEFAULT_OP_TIMEOUT_S); } /* add meta_timeout env needed by some resource agents */ g_hash_table_insert(params, strdup("CRM_meta_timeout"), crm_strdup_printf("%d", timeout_ms)); /* add crm_feature_set env needed by some resource agents */ g_hash_table_insert(params, strdup(XML_ATTR_CRM_VERSION), strdup(CRM_FEATURE_SET)); /* resources_action_create frees the params hash table it's passed, but we * may need to reuse it in a second call to resources_action_create. Thus * we'll make a copy here so that gets freed and the original remains for * reuse. */ params_copy = crm_str_table_dup(params); op = resources_action_create(rsc_name, rsc_class, rsc_prov, rsc_type, action, 0, timeout_ms, params_copy, 0); if (op == NULL) { /* Re-run with stderr enabled so we can display a sane error message */ crm_enable_stderr(TRUE); params_copy = crm_str_table_dup(params); op = resources_action_create(rsc_name, rsc_class, rsc_prov, rsc_type, action, 0, timeout_ms, params_copy, 0); /* Callers of cli_resource_execute expect that the params hash table will * be freed. That function uses this one, so for that reason and for * making the two act the same, we should free the hash table here too. */ g_hash_table_destroy(params); /* We know op will be NULL, but this makes static analysis happy */ services_action_free(op); crm_exit(CRM_EX_DATAERR); return rc; // Never reached, but helps static analysis } setenv("HA_debug", resource_verbose > 0 ? "1" : "0", 1); if(resource_verbose > 1) { setenv("OCF_TRACE_RA", "1", 1); } /* A resource agent using the standard ocf-shellfuncs library will not print * messages to stderr if it doesn't have a controlling terminal (e.g. if * crm_resource is called via script or ssh). This forces it to do so. */ setenv("OCF_TRACE_FILE", "/dev/stderr", 0); if (override_hash) { GHashTableIter iter; char *name = NULL; char *value = NULL; g_hash_table_iter_init(&iter, override_hash); while (g_hash_table_iter_next(&iter, (gpointer *) & name, (gpointer *) & value)) { printf("Overriding the cluster configuration for '%s' with '%s' = '%s'\n", rsc_name, name, value); g_hash_table_replace(op->params, strdup(name), strdup(value)); } } if (services_action_sync(op)) { int more, lpc, last; char *local_copy = NULL; rc = op->rc; if (op->status == PCMK_LRM_OP_DONE) { printf("Operation %s for %s (%s:%s:%s) returned: '%s' (%d)\n", action, rsc_name, rsc_class, rsc_prov ? rsc_prov : "", rsc_type, services_ocf_exitcode_str(op->rc), op->rc); } else { printf("Operation %s for %s (%s:%s:%s) failed: '%s' (%d)\n", action, rsc_name, rsc_class, rsc_prov ? rsc_prov : "", rsc_type, services_lrm_status_str(op->status), op->status); } /* hide output for validate-all if not in verbose */ if (resource_verbose == 0 && safe_str_eq(action, "validate-all")) goto done; if (op->stdout_data) { local_copy = strdup(op->stdout_data); more = strlen(local_copy); last = 0; for (lpc = 0; lpc < more; lpc++) { if (local_copy[lpc] == '\n' || local_copy[lpc] == 0) { local_copy[lpc] = 0; printf(" > stdout: %s\n", local_copy + last); last = lpc + 1; } } free(local_copy); } if (op->stderr_data) { local_copy = strdup(op->stderr_data); more = strlen(local_copy); last = 0; for (lpc = 0; lpc < more; lpc++) { if (local_copy[lpc] == '\n' || local_copy[lpc] == 0) { local_copy[lpc] = 0; printf(" > stderr: %s\n", local_copy + last); last = lpc + 1; } } free(local_copy); } } else { rc = op->rc == 0 ? pcmk_err_generic : op->rc; } done: services_action_free(op); /* See comment above about why we free params here. */ g_hash_table_destroy(params); return rc; } int cli_resource_execute(pe_resource_t *rsc, const char *requested_name, const char *rsc_action, GHashTable *override_hash, int timeout_ms, cib_t * cib, pe_working_set_t *data_set) { int rc = pcmk_ok; const char *rid = NULL; const char *rtype = NULL; const char *rprov = NULL; const char *rclass = NULL; const char *action = NULL; GHashTable *params = NULL; if (safe_str_eq(rsc_action, "validate")) { action = "validate-all"; } else if (safe_str_eq(rsc_action, "force-check")) { action = "monitor"; } else if (safe_str_eq(rsc_action, "force-stop")) { action = rsc_action+6; } else if (safe_str_eq(rsc_action, "force-start") || safe_str_eq(rsc_action, "force-demote") || safe_str_eq(rsc_action, "force-promote")) { action = rsc_action+6; if(pe_rsc_is_clone(rsc)) { rc = cli_resource_search(rsc, requested_name, data_set); if(rc > 0 && do_force == FALSE) { CMD_ERR("It is not safe to %s %s here: the cluster claims it is already active", action, rsc->id); CMD_ERR("Try setting target-role=Stopped first or specifying " "the force option"); crm_exit(CRM_EX_UNSAFE); } } } else { action = rsc_action; } if(pe_rsc_is_clone(rsc)) { /* Grab the first child resource in the hope it's not a group */ rsc = rsc->children->data; } if(rsc->variant == pe_group) { CMD_ERR("Sorry, the %s option doesn't support group resources", rsc_action); crm_exit(CRM_EX_UNIMPLEMENT_FEATURE); } rclass = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS); rprov = crm_element_value(rsc->xml, XML_AGENT_ATTR_PROVIDER); rtype = crm_element_value(rsc->xml, XML_ATTR_TYPE); params = generate_resource_params(rsc, data_set); if (timeout_ms == 0) { timeout_ms = pe_get_configured_timeout(rsc, action, data_set); } rid = pe_rsc_is_anon_clone(rsc->parent)? requested_name : rsc->id; rc = cli_resource_execute_from_params(rid, rclass, rprov, rtype, action, params, override_hash, timeout_ms); return rc; } int cli_resource_move(pe_resource_t *rsc, const char *rsc_id, const char *host_name, cib_t *cib, pe_working_set_t *data_set) { int rc = pcmk_ok; unsigned int count = 0; pe_node_t *current = NULL; pe_node_t *dest = pe_find_node(data_set->nodes, host_name); bool cur_is_dest = FALSE; if (dest == NULL) { return -pcmk_err_node_unknown; } if (scope_master && is_not_set(rsc->flags, pe_rsc_promotable)) { pe_resource_t *p = uber_parent(rsc); if (is_set(p->flags, pe_rsc_promotable)) { CMD_ERR("Using parent '%s' for move instead of '%s'.", rsc->id, rsc_id); rsc_id = p->id; rsc = p; } else { CMD_ERR("Ignoring master option: %s is not promotable", rsc_id); scope_master = FALSE; } } current = pe__find_active_requires(rsc, &count); if (is_set(rsc->flags, pe_rsc_promotable)) { GListPtr iter = NULL; unsigned int master_count = 0; pe_node_t *master_node = NULL; for(iter = rsc->children; iter; iter = iter->next) { pe_resource_t *child = (pe_resource_t *)iter->data; enum rsc_role_e child_role = child->fns->state(child, TRUE); if(child_role == RSC_ROLE_MASTER) { rsc = child; master_node = pe__current_node(child); master_count++; } } if (scope_master || master_count) { count = master_count; current = master_node; } } if (count > 1) { if (pe_rsc_is_clone(rsc)) { current = NULL; } else { return -pcmk_err_multiple; } } if (current && (current->details == dest->details)) { cur_is_dest = TRUE; if (do_force) { crm_info("%s is already %s on %s, reinforcing placement with location constraint.", rsc_id, scope_master?"promoted":"active", dest->details->uname); } else { return -pcmk_err_already; } } /* Clear any previous prefer constraints across all nodes. */ cli_resource_clear(rsc_id, NULL, data_set->nodes, cib, FALSE); /* Clear any previous ban constraints on 'dest'. */ cli_resource_clear(rsc_id, dest->details->uname, data_set->nodes, cib, TRUE); /* Record an explicit preference for 'dest' */ rc = cli_resource_prefer(rsc_id, dest->details->uname, cib); crm_trace("%s%s now prefers node %s%s", rsc->id, scope_master?" (master)":"", dest->details->uname, do_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(do_force && (cur_is_dest == FALSE)) { /* Ban the original location if possible */ if(current) { (void)cli_resource_ban(rsc_id, current->details->uname, NULL, cib); } else if(count > 1) { CMD_ERR("Resource '%s' is currently %s in %d locations. " "One may now move to %s", rsc_id, (scope_master? "promoted" : "active"), count, dest->details->uname); CMD_ERR("To prevent '%s' from being %s at a specific location, " "specify a node.", rsc_id, (scope_master? "promoted" : "active")); } else { crm_trace("Not banning %s from its current location: not active", rsc_id); } } return rc; } static void cli_resource_why_without_rsc_and_host(cib_t *cib_conn,GListPtr resources) { GListPtr lpc = NULL; GListPtr hosts = NULL; for (lpc = resources; lpc != NULL; lpc = lpc->next) { pe_resource_t *rsc = (pe_resource_t *) lpc->data; rsc->fns->location(rsc, &hosts, TRUE); if (hosts == NULL) { printf("Resource %s is not running\n", rsc->id); } else { printf("Resource %s is running\n", rsc->id); } cli_resource_check(cib_conn, rsc); g_list_free(hosts); hosts = NULL; } } static void cli_resource_why_with_rsc_and_host(cib_t *cib_conn, GListPtr resources, pe_resource_t *rsc, const char *host_uname) { if (resource_is_running_on(rsc, host_uname)) { printf("Resource %s is running on host %s\n",rsc->id,host_uname); } else { printf("Resource %s is not running on host %s\n", rsc->id, host_uname); } cli_resource_check(cib_conn, rsc); } static void cli_resource_why_without_rsc_with_host(cib_t *cib_conn,GListPtr resources,pe_node_t *node) { const char* host_uname = node->details->uname; GListPtr allResources = node->details->allocated_rsc; GListPtr activeResources = node->details->running_rsc; GListPtr unactiveResources = subtract_lists(allResources,activeResources,(GCompareFunc) strcmp); GListPtr lpc = NULL; for (lpc = activeResources; lpc != NULL; lpc = lpc->next) { pe_resource_t *rsc = (pe_resource_t *) lpc->data; printf("Resource %s is running on host %s\n",rsc->id,host_uname); cli_resource_check(cib_conn,rsc); } for(lpc = unactiveResources; lpc != NULL; lpc = lpc->next) { pe_resource_t *rsc = (pe_resource_t *) lpc->data; printf("Resource %s is assigned to host %s but not running\n", rsc->id, host_uname); cli_resource_check(cib_conn,rsc); } g_list_free(allResources); g_list_free(activeResources); g_list_free(unactiveResources); } static void cli_resource_why_with_rsc_without_host(cib_t *cib_conn, GListPtr resources, pe_resource_t *rsc) { GListPtr hosts = NULL; rsc->fns->location(rsc, &hosts, TRUE); printf("Resource %s is %srunning\n", rsc->id, (hosts? "" : "not ")); cli_resource_check(cib_conn, rsc); g_list_free(hosts); } void cli_resource_why(cib_t *cib_conn, GListPtr resources, pe_resource_t *rsc, pe_node_t *node) { const char *host_uname = (node == NULL)? NULL : node->details->uname; if ((rsc == NULL) && (host_uname == NULL)) { cli_resource_why_without_rsc_and_host(cib_conn, resources); } else if ((rsc != NULL) && (host_uname != NULL)) { cli_resource_why_with_rsc_and_host(cib_conn, resources, rsc, host_uname); } else if ((rsc == NULL) && (host_uname != NULL)) { cli_resource_why_without_rsc_with_host(cib_conn, resources, node); } else if ((rsc != NULL) && (host_uname == NULL)) { cli_resource_why_with_rsc_without_host(cib_conn, resources, rsc); } }