gboolean mcp_read_config(void) { int rc = CS_OK; int retries = 0; cmap_handle_t local_handle; uint64_t config = 0; // There can be only one possibility do { rc = cmap_initialize(&local_handle); if (rc != CS_OK) { retries++; printf("cmap connection setup failed: %s. Retrying in %ds\n", cs_strerror(rc), retries); crm_info("cmap connection setup failed: %s. Retrying in %ds", cs_strerror(rc), retries); sleep(retries); } else { break; } } while (retries < 5); if (rc != CS_OK) { printf("Could not connect to Cluster Configuration Database API, error %d\n", rc); crm_warn("Could not connect to Cluster Configuration Database API, error %d", rc); return FALSE; } stack = get_cluster_type(); crm_info("Reading configure for stack: %s", name_for_cluster_type(stack)); /* =::=::= Should we be here =::=::= */ if (stack == pcmk_cluster_corosync) { set_daemon_option("cluster_type", "corosync"); set_daemon_option("quorum_type", "corosync"); } else { crm_err("Unsupported stack type: %s", name_for_cluster_type(stack)); return FALSE; } /* =::=::= Logging =::=::= */ if (daemon_option("debug")) { /* Syslog logging is already setup by crm_log_init() */ } else { /* Check corosync */ char *debug_enabled = NULL; get_config_opt(config, local_handle, "logging.debug", &debug_enabled, "off"); if (crm_is_true(debug_enabled)) { set_daemon_option("debug", "1"); if (get_crm_log_level() < LOG_DEBUG) { set_crm_log_level(LOG_DEBUG); } } else { set_daemon_option("debug", "0"); } free(debug_enabled); } if(local_handle){ gid_t gid = 0; if (crm_user_lookup(CRM_DAEMON_USER, NULL, &gid) < 0) { crm_warn("Could not authorize group with corosync " CRM_XS " No group found for user %s", CRM_DAEMON_USER); } else { char key[PATH_MAX]; snprintf(key, PATH_MAX, "uidgid.gid.%u", gid); rc = cmap_set_uint8(local_handle, key, 1); if (rc != CS_OK) { crm_warn("Could not authorize group with corosync "CRM_XS " group=%u rc=%d (%s)", gid, rc, ais_error2text(rc)); } } } cmap_finalize(local_handle); return TRUE; }
gboolean decode_transition_key( const char *key, char **uuid, int *transition_id, int *action_id, int *target_rc) { int res = 0; gboolean done = FALSE; CRM_CHECK(uuid != NULL, return FALSE); CRM_CHECK(target_rc != NULL, return FALSE); CRM_CHECK(action_id != NULL, return FALSE); CRM_CHECK(transition_id != NULL, return FALSE); crm_malloc0(*uuid, strlen(key)); res = sscanf(key, "%d:%d:%d:%s", action_id, transition_id, target_rc, *uuid); switch(res) { case 4: /* Post Pacemaker 0.6 */ done = TRUE; break; case 3: case 2: /* this can be tricky - the UUID might start with an integer */ /* Until Pacemaker 0.6 */ done = TRUE; *target_rc = -1; res = sscanf(key, "%d:%d:%s", action_id, transition_id, *uuid); if(res == 2) { *action_id = -1; res = sscanf(key, "%d:%s", transition_id, *uuid); CRM_CHECK(res == 2, done = FALSE); } else if(res != 3) { CRM_CHECK(res == 3, done = FALSE); } break; case 1: /* Prior to Heartbeat 2.0.8 */ done = TRUE; *action_id = -1; *target_rc = -1; res = sscanf(key, "%d:%s", transition_id, *uuid); CRM_CHECK(res == 2, done = FALSE); break; default: crm_crit("Unhandled sscanf result (%d) for %s", res, key); } if(strlen(*uuid) != 36) { crm_warn("Bad UUID (%s) in sscanf result (%d) for %s", *uuid, res, key); } if(done == FALSE) { crm_err("Cannot decode '%s' rc=%d", key, res); crm_free(*uuid); *uuid = NULL; *target_rc = -1; *action_id = -1; *transition_id = -1; } return done; }
void do_cib_notify( int options, const char *op, crm_data_t *update, enum cib_errors result, crm_data_t *result_data, const char *msg_type) { HA_Message *update_msg = NULL; const char *type = NULL; const char *id = NULL; update_msg = ha_msg_new(8); if(result_data != NULL) { id = crm_element_value(result_data, XML_ATTR_ID); } ha_msg_add(update_msg, F_TYPE, T_CIB_NOTIFY); ha_msg_add(update_msg, F_SUBTYPE, msg_type); ha_msg_add(update_msg, F_CIB_OPERATION, op); ha_msg_add_int(update_msg, F_CIB_RC, result); if(id != NULL) { ha_msg_add(update_msg, F_CIB_OBJID, id); } if(update != NULL) { crm_debug_4("Setting type to update->name: %s", crm_element_name(update)); ha_msg_add(update_msg, F_CIB_OBJTYPE, crm_element_name(update)); type = crm_element_name(update); } else if(result_data != NULL) { crm_debug_4("Setting type to new_obj->name: %s", crm_element_name(result_data)); ha_msg_add(update_msg, F_CIB_OBJTYPE, crm_element_name(result_data)); type = crm_element_name(result_data); } else { crm_debug_4("Not Setting type"); } attach_cib_generation(update_msg, "cib_generation", the_cib); if(update != NULL) { add_message_xml(update_msg, F_CIB_UPDATE, update); } if(result_data != NULL) { add_message_xml(update_msg, F_CIB_UPDATE_RESULT, result_data); } crm_debug_3("Notifying clients"); g_hash_table_foreach(client_list, cib_notify_client, update_msg); crm_msg_del(update_msg); if(update == NULL) { if(result == cib_ok) { crm_debug_2("Operation %s (on section=%s) completed", op, crm_str(type)); } else { crm_warn("Operation %s (on section=%s) FAILED: (%d) %s", op, crm_str(type), result, cib_error2string(result)); } } else { if(result == cib_ok) { crm_debug_2("Completed %s of <%s %s%s>", op, crm_str(type), id?"id=":"", id?id:""); } else { crm_warn("%s of <%s %s%s> FAILED: %s", op,crm_str(type), id?"id=":"", id?id:"", cib_error2string(result)); } } crm_debug_3("Notify complete"); }
remote_fencing_op_t * initiate_remote_stonith_op(crm_client_t * client, xmlNode * request, gboolean manual_ack) { int query_timeout = 0; xmlNode *query = NULL; const char *client_id = NULL; remote_fencing_op_t *op = NULL; if (client) { client_id = client->id; } else { client_id = crm_element_value(request, F_STONITH_CLIENTID); } CRM_LOG_ASSERT(client_id != NULL); op = create_remote_stonith_op(client_id, request, FALSE); op->owner = TRUE; if (manual_ack) { crm_notice("Initiating manual confirmation for %s: %s", op->target, op->id); return op; } CRM_CHECK(op->action, return NULL); if (stonith_topology_next(op) != pcmk_ok) { op->state = st_failed; } switch (op->state) { case st_failed: crm_warn("Initiation of remote operation %s for %s: failed (%s)", op->action, op->target, op->id); remote_op_done(op, NULL, -EINVAL, FALSE); return op; case st_duplicate: crm_info("Initiating remote operation %s for %s: %s (duplicate)", op->action, op->target, op->id); return op; default: crm_notice("Initiating remote operation %s for %s: %s (%d)", op->action, op->target, op->id, op->state); } query = stonith_create_op(op->client_callid, op->id, STONITH_OP_QUERY, NULL, 0); crm_xml_add(query, F_STONITH_REMOTE_OP_ID, op->id); crm_xml_add(query, F_STONITH_TARGET, op->target); crm_xml_add(query, F_STONITH_ACTION, op->action); crm_xml_add(query, F_STONITH_ORIGIN, op->originator); crm_xml_add(query, F_STONITH_CLIENTID, op->client_id); crm_xml_add(query, F_STONITH_CLIENTNAME, op->client_name); crm_xml_add_int(query, F_STONITH_TIMEOUT, op->base_timeout); send_cluster_message(NULL, crm_msg_stonith_ng, query, FALSE); free_xml(query); query_timeout = op->base_timeout * TIMEOUT_MULTIPLY_FACTOR; op->query_timer = g_timeout_add((1000 * query_timeout), remote_op_query_timeout, op); return op; }
gboolean cib_action_update(crm_action_t * action, int status, int op_rc) { lrmd_event_data_t *op = NULL; xmlNode *state = NULL; xmlNode *rsc = NULL; xmlNode *xml_op = NULL; xmlNode *action_rsc = NULL; int rc = pcmk_ok; const char *name = NULL; const char *value = NULL; const char *rsc_id = NULL; const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); const char *task_uuid = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY); const char *target_uuid = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID); int call_options = cib_quorum_override | cib_scope_local; int target_rc = get_target_rc(action); if (status == PCMK_LRM_OP_PENDING) { crm_debug("%s %d: Recording pending operation %s on %s", crm_element_name(action->xml), action->id, task_uuid, target); } else { crm_warn("%s %d: %s on %s timed out", crm_element_name(action->xml), action->id, task_uuid, target); } action_rsc = find_xml_node(action->xml, XML_CIB_TAG_RESOURCE, TRUE); if (action_rsc == NULL) { return FALSE; } rsc_id = ID(action_rsc); CRM_CHECK(rsc_id != NULL, crm_log_xml_err(action->xml, "Bad:action"); return FALSE); /* update the CIB <node_state id="hadev"> <lrm> <lrm_resources> <lrm_resource id="rsc2" last_op="start" op_code="0" target="hadev"/> */ state = create_xml_node(NULL, XML_CIB_TAG_STATE); crm_xml_add(state, XML_ATTR_UUID, target_uuid); crm_xml_add(state, XML_ATTR_UNAME, target); rsc = create_xml_node(state, XML_CIB_TAG_LRM); crm_xml_add(rsc, XML_ATTR_ID, target_uuid); rsc = create_xml_node(rsc, XML_LRM_TAG_RESOURCES); rsc = create_xml_node(rsc, XML_LRM_TAG_RESOURCE); crm_xml_add(rsc, XML_ATTR_ID, rsc_id); name = XML_ATTR_TYPE; value = crm_element_value(action_rsc, name); crm_xml_add(rsc, name, value); name = XML_AGENT_ATTR_CLASS; value = crm_element_value(action_rsc, name); crm_xml_add(rsc, name, value); name = XML_AGENT_ATTR_PROVIDER; value = crm_element_value(action_rsc, name); crm_xml_add(rsc, name, value); op = convert_graph_action(NULL, action, status, op_rc); op->call_id = -1; op->user_data = generate_transition_key(transition_graph->id, action->id, target_rc, te_uuid); xml_op = create_operation_update(rsc, op, CRM_FEATURE_SET, target_rc, __FUNCTION__, LOG_INFO); lrmd_free_event(op); crm_trace("Updating CIB with \"%s\" (%s): %s %s on %s", status < 0 ? "new action" : XML_ATTR_TIMEOUT, crm_element_name(action->xml), crm_str(task), rsc_id, target); crm_log_xml_trace(xml_op, "Op"); rc = fsa_cib_conn->cmds->update(fsa_cib_conn, XML_CIB_TAG_STATUS, state, call_options); crm_trace("Updating CIB with %s action %d: %s on %s (call_id=%d)", services_lrm_status_str(status), action->id, task_uuid, target, rc); fsa_register_cib_callback(rc, FALSE, NULL, cib_action_updated); free_xml(state); action->sent_update = TRUE; if (rc < pcmk_ok) { return FALSE; } return TRUE; }
/* Returns FALSE if 'op' should be free'd by the caller */ gboolean services_os_action_execute(svc_action_t * op, gboolean synchronous) { int stdout_fd[2]; int stderr_fd[2]; sigset_t mask; sigset_t old_mask; struct stat st; if (pipe(stdout_fd) < 0) { crm_err("pipe() failed"); } if (pipe(stderr_fd) < 0) { crm_err("pipe() failed"); } /* Fail fast */ if(stat(op->opaque->exec, &st) != 0) { int rc = errno; crm_warn("Cannot execute '%s': %s (%d)", op->opaque->exec, pcmk_strerror(rc), rc); services_handle_exec_error(op, rc); if (!synchronous) { return operation_finalize(op); } return FALSE; } if (synchronous) { sigemptyset(&mask); sigaddset(&mask, SIGCHLD); sigemptyset(&old_mask); if (sigprocmask(SIG_BLOCK, &mask, &old_mask) < 0) { crm_perror(LOG_ERR, "sigprocmask() failed"); } } op->pid = fork(); switch (op->pid) { case -1: { int rc = errno; close(stdout_fd[0]); close(stdout_fd[1]); close(stderr_fd[0]); close(stderr_fd[1]); crm_err("Could not execute '%s': %s (%d)", op->opaque->exec, pcmk_strerror(rc), rc); services_handle_exec_error(op, rc); if (!synchronous) { return operation_finalize(op); } return FALSE; } case 0: /* Child */ close(stdout_fd[0]); close(stderr_fd[0]); if (STDOUT_FILENO != stdout_fd[1]) { if (dup2(stdout_fd[1], STDOUT_FILENO) != STDOUT_FILENO) { crm_err("dup2() failed (stdout)"); } close(stdout_fd[1]); } if (STDERR_FILENO != stderr_fd[1]) { if (dup2(stderr_fd[1], STDERR_FILENO) != STDERR_FILENO) { crm_err("dup2() failed (stderr)"); } close(stderr_fd[1]); } action_launch_child(op); } /* Only the parent reaches here */ close(stdout_fd[1]); close(stderr_fd[1]); op->opaque->stdout_fd = stdout_fd[0]; set_fd_opts(op->opaque->stdout_fd, O_NONBLOCK); op->opaque->stderr_fd = stderr_fd[0]; set_fd_opts(op->opaque->stderr_fd, O_NONBLOCK); if (synchronous) { action_synced_wait(op, mask); if (sigismember(&old_mask, SIGCHLD) == 0) { if (sigprocmask(SIG_UNBLOCK, &mask, NULL) < 0) { crm_perror(LOG_ERR, "sigprocmask() to unblocked failed"); } } } else { crm_trace("Async waiting for %d - %s", op->pid, op->opaque->exec); mainloop_child_add(op->pid, op->timeout, op->id, op, operation_finished); op->opaque->stdout_gsource = mainloop_add_fd(op->id, G_PRIORITY_LOW, op->opaque->stdout_fd, op, &stdout_callbacks); op->opaque->stderr_gsource = mainloop_add_fd(op->id, G_PRIORITY_LOW, op->opaque->stderr_fd, op, &stderr_callbacks); } return TRUE; }
int cib_process_xpath(const char *op, int options, const char *section, xmlNode * req, xmlNode * input, xmlNode * existing_cib, xmlNode ** result_cib, xmlNode ** answer) { int lpc = 0; int max = 0; int rc = pcmk_ok; gboolean is_query = safe_str_eq(op, CIB_OP_QUERY); xmlXPathObjectPtr xpathObj = NULL; crm_trace("Processing \"%s\" event", op); if (is_query) { xpathObj = xpath_search(existing_cib, section); } else { xpathObj = xpath_search(*result_cib, section); } max = numXpathResults(xpathObj); if (max < 1 && safe_str_eq(op, CIB_OP_DELETE)) { crm_debug("%s was already removed", section); } else if (max < 1) { crm_debug("%s: %s does not exist", op, section); rc = -ENXIO; } else if (is_query) { if (max > 1) { *answer = create_xml_node(NULL, "xpath-query"); } } if (safe_str_eq(op, CIB_OP_DELETE) && (options & cib_multiple)) { dedupXpathResults(xpathObj); } for (lpc = 0; lpc < max; lpc++) { xmlChar *path = NULL; xmlNode *match = getXpathResult(xpathObj, lpc); if (match == NULL) { continue; } path = xmlGetNodePath(match); crm_debug("Processing %s op for %s (%s)", op, section, path); free(path); if (safe_str_eq(op, CIB_OP_DELETE)) { if (match == *result_cib) { /* Attempting to delete the whole "/cib" */ crm_warn("Cannot perform %s for %s: The xpath is addressing the whole /cib", op, section); rc = -EINVAL; break; } free_xml(match); if ((options & cib_multiple) == 0) { break; } } else if (safe_str_eq(op, CIB_OP_MODIFY)) { if (update_xml_child(match, input) == FALSE) { rc = -ENXIO; } else if ((options & cib_multiple) == 0) { break; } } else if (safe_str_eq(op, CIB_OP_CREATE)) { add_node_copy(match, input); break; } else if (safe_str_eq(op, CIB_OP_QUERY)) { if (options & cib_no_children) { const char *tag = TYPE(match); xmlNode *shallow = create_xml_node(*answer, tag); copy_in_properties(shallow, match); if (*answer == NULL) { *answer = shallow; } } else if (options & cib_xpath_address) { int path_len = 0; char *path = NULL; xmlNode *parent = match; while (parent && parent->type == XML_ELEMENT_NODE) { int extra = 1; char *new_path = NULL; const char *id = crm_element_value(parent, XML_ATTR_ID); extra += strlen((const char *)parent->name); if (id) { extra += 8; /* [@id=""] */ extra += strlen(id); } path_len += extra; new_path = malloc(path_len + 1); if(new_path == NULL) { break; } else if (id) { snprintf(new_path, path_len + 1, "/%s[@id='%s']%s", parent->name, id, path ? path : ""); } else { snprintf(new_path, path_len + 1, "/%s%s", parent->name, path ? path : ""); } free(path); path = new_path; parent = parent->parent; } crm_trace("Got: %s\n", path); if (*answer == NULL) { *answer = create_xml_node(NULL, "xpath-query"); } parent = create_xml_node(*answer, "xpath-query-path"); crm_xml_add(parent, XML_ATTR_ID, path); free(path); } else if (*answer) { add_node_copy(*answer, match); } else { *answer = match; } } else if (safe_str_eq(op, CIB_OP_REPLACE)) { xmlNode *parent = match->parent; free_xml(match); if (input != NULL) { add_node_copy(parent, input); } if ((options & cib_multiple) == 0) { break; } } } freeXpathObject(xpathObj); return rc; }
gboolean subsystem_msg_dispatch(IPC_Channel *sender, void *user_data) { int lpc = 0; xmlNode *msg = NULL; xmlNode *data = NULL; gboolean all_is_well = TRUE; const char *sys_to; const char *task; gboolean (*process_function) (xmlNode *msg, xmlNode *data, IPC_Channel *sender) = NULL; while(IPC_ISRCONN(sender)) { gboolean process = FALSE; if(sender->ops->is_message_pending(sender) == 0) { break; } msg = xmlfromIPC(sender, MAX_IPC_DELAY); if (msg == NULL) { break; } lpc++; crm_log_xml(LOG_MSG, __FUNCTION__, msg); sys_to = crm_element_value(msg, F_CRM_SYS_TO); task = crm_element_value(msg, F_CRM_TASK); if(safe_str_eq(task, CRM_OP_HELLO)) { process = TRUE; } else if(sys_to == NULL) { crm_err("Value of %s was NULL!!", F_CRM_SYS_TO); } else if(task == NULL) { crm_err("Value of %s was NULL!!", F_CRM_TASK); } else { process = TRUE; } if(process == FALSE) { free_xml(msg); msg = NULL; continue; } data = get_message_xml(msg, F_CRM_DATA); process_function = user_data; if(FALSE == process_function(msg, data, sender)) { crm_warn("Received a message destined for %s" " by mistake", sys_to); } free_xml(msg); msg = NULL; if(sender->ch_status == IPC_CONNECT) { break; } } crm_debug_2("Processed %d messages", lpc); if (sender->ch_status != IPC_CONNECT) { crm_err("The server %d has left us: Shutting down...NOW", sender->farside_pid); exit(1); /* shutdown properly later */ return !all_is_well; } return all_is_well; }
/* A_ELECTION_VOTE */ void do_election_vote(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { struct timeval age; xmlNode *vote = NULL; gboolean not_voting = FALSE; /* don't vote if we're in one of these states or wanting to shut down */ switch (cur_state) { case S_STARTING: case S_RECOVERY: case S_STOPPING: case S_TERMINATE: crm_warn("Not voting in election, we're in state %s", fsa_state2string(cur_state)); not_voting = TRUE; break; default: break; } if (not_voting == FALSE) { if (is_set(fsa_input_register, R_STARTING)) { not_voting = TRUE; } } if (not_voting) { if (AM_I_DC) { register_fsa_input(C_FSA_INTERNAL, I_RELEASE_DC, NULL); } else { register_fsa_input(C_FSA_INTERNAL, I_PENDING, NULL); } return; } vote = create_request(CRM_OP_VOTE, NULL, NULL, CRM_SYSTEM_CRMD, CRM_SYSTEM_CRMD, NULL); current_election_id++; crm_xml_add(vote, F_CRM_ELECTION_OWNER, fsa_our_uuid); crm_xml_add_int(vote, F_CRM_ELECTION_ID, current_election_id); crm_uptime(&age); crm_xml_add_int(vote, F_CRM_ELECTION_AGE_S, age.tv_sec); crm_xml_add_int(vote, F_CRM_ELECTION_AGE_US, age.tv_usec); send_cluster_message(NULL, crm_msg_crmd, vote, TRUE); free_xml(vote); crm_debug("Started election %d", current_election_id); if (voted) { g_hash_table_destroy(voted); } voted = NULL; if (cur_state == S_ELECTION || cur_state == S_RELEASE_DC) { crm_timer_start(election_timeout); } else if (cur_state != S_INTEGRATION) { crm_err("Broken? Voting in state %s", fsa_state2string(cur_state)); } return; }
int register_fsa_input_adv( enum crmd_fsa_cause cause, enum crmd_fsa_input input, void *data, long long with_actions, gboolean prepend, const char *raised_from) { unsigned old_len = g_list_length(fsa_message_queue); fsa_data_t *fsa_data = NULL; last_data_id++; CRM_CHECK(raised_from != NULL, raised_from = "<unknown>"); crm_debug("%s %s FSA input %d (%s) (cause=%s) %s data", raised_from, prepend?"prepended":"appended",last_data_id, fsa_input2string(input), fsa_cause2string(cause), data?"with":"without"); if(input == I_WAIT_FOR_EVENT) { do_fsa_stall = TRUE; crm_debug("Stalling the FSA pending further input: cause=%s", fsa_cause2string(cause)); if(old_len > 0) { crm_warn("%s stalled the FSA with pending inputs", raised_from); fsa_dump_queue(LOG_DEBUG); } if(data == NULL) { set_bit_inplace(fsa_actions, with_actions); with_actions = A_NOTHING; return 0; } crm_err("%s stalled the FSA with data - this may be broken", raised_from); } if(old_len == 0) { last_was_vote = FALSE; } if(input == I_NULL && with_actions == A_NOTHING /* && data == NULL */) { /* no point doing anything */ crm_err("Cannot add entry to queue: no input and no action"); return 0; } else if(data == NULL) { last_was_vote = FALSE; #if 0 } else if(last_was_vote && cause == C_HA_MESSAGE && input == I_ROUTER) { const char *op = cl_get_string( ((ha_msg_input_t*)data)->msg, F_CRM_TASK); if(safe_str_eq(op, CRM_OP_VOTE)) { /* It is always safe to treat N successive votes as * a single one * * If all the discarded votes are more "loosing" than * the first then the result is accurate * (win or loose). * * If any of the discarded votes are less "loosing" * than the first then we will cast our vote and the * eventual winner will vote us down again (which * even in the case that N=2, is no worse than if we * had not disarded the vote). */ crm_debug_2("Vote compression: %d", old_len); return 0; } #endif } else if (cause == C_HA_MESSAGE && input == I_ROUTER) { const char *op = cl_get_string( ((ha_msg_input_t*)data)->msg, F_CRM_TASK); if(safe_str_eq(op, CRM_OP_VOTE)) { last_was_vote = TRUE; crm_debug_3("Added vote: %d", old_len); } } else { last_was_vote = FALSE; } crm_malloc0(fsa_data, sizeof(fsa_data_t)); fsa_data->id = last_data_id; fsa_data->fsa_input = input; fsa_data->fsa_cause = cause; fsa_data->origin = raised_from; fsa_data->data = NULL; fsa_data->data_type = fsa_dt_none; fsa_data->actions = with_actions; if(with_actions != A_NOTHING) { crm_debug_3("Adding actions %.16llx to input", with_actions); } if(data != NULL) { switch(cause) { case C_FSA_INTERNAL: case C_CRMD_STATUS_CALLBACK: case C_IPC_MESSAGE: case C_HA_MESSAGE: crm_debug_3("Copying %s data from %s as a HA msg", fsa_cause2string(cause), raised_from); fsa_data->data = copy_ha_msg_input(data); fsa_data->data_type = fsa_dt_ha_msg; break; case C_LRM_OP_CALLBACK: crm_debug_3("Copying %s data from %s as lrm_op_t", fsa_cause2string(cause), raised_from); fsa_data->data = copy_lrm_op((lrm_op_t*)data); fsa_data->data_type = fsa_dt_lrm; break; case C_CCM_CALLBACK: crm_debug_3("Copying %s data from %s as CCM data", fsa_cause2string(cause), raised_from); fsa_data->data = copy_ccm_data(data); fsa_data->data_type = fsa_dt_ccm; break; case C_SUBSYSTEM_CONNECT: case C_LRM_MONITOR_CALLBACK: case C_TIMER_POPPED: case C_SHUTDOWN: case C_HEARTBEAT_FAILED: case C_HA_DISCONNECT: case C_ILLEGAL: case C_UNKNOWN: case C_STARTUP: crm_err("Copying %s data (from %s)" " not yet implemented", fsa_cause2string(cause), raised_from); exit(1); break; } crm_debug_4("%s data copied", fsa_cause2string(fsa_data->fsa_cause)); } /* make sure to free it properly later */ if(prepend) { crm_debug_2("Prepending input"); fsa_message_queue = g_list_prepend(fsa_message_queue, fsa_data); } else { fsa_message_queue = g_list_append(fsa_message_queue, fsa_data); } crm_debug_2("Queue len: %d", g_list_length(fsa_message_queue)); fsa_dump_queue(LOG_DEBUG_2); if(old_len == g_list_length(fsa_message_queue)) { crm_err("Couldnt add message to the queue"); } if(fsa_source) { crm_debug_3("Triggering FSA: %s", __FUNCTION__); G_main_set_trigger(fsa_source); } return last_data_id; }
/* A_ELECTION_COUNT */ enum election_result election_count_vote(election_t *e, xmlNode *vote, bool can_win) { int age = 0; int election_id = -1; int log_level = LOG_INFO; gboolean use_born_on = FALSE; gboolean done = FALSE; gboolean we_loose = FALSE; const char *op = NULL; const char *from = NULL; const char *reason = "unknown"; const char *election_owner = NULL; crm_node_t *our_node = NULL, *your_node = NULL; static int election_wins = 0; xmlNode *novote = NULL; time_t tm_now = time(NULL); static time_t expires = 0; static time_t last_election_loss = 0; /* if the membership copy is NULL we REALLY shouldn't be voting * the question is how we managed to get here. */ CRM_CHECK(vote != NULL, return election_error); if(e == NULL) { crm_info("Not voting in election: not initialized"); return election_lost; } else if(crm_peer_cache == NULL) { crm_info("Not voting in election: no peer cache"); return election_lost; } op = crm_element_value(vote, F_CRM_TASK); from = crm_element_value(vote, F_CRM_HOST_FROM); election_owner = crm_element_value(vote, F_CRM_ELECTION_OWNER); crm_element_value_int(vote, F_CRM_ELECTION_ID, &election_id); your_node = crm_get_peer(0, from); our_node = crm_get_peer(0, e->uname); if (e->voted == NULL) { crm_debug("Created voted hash"); e->voted = g_hash_table_new_full(crm_str_hash, g_str_equal, g_hash_destroy_str, g_hash_destroy_str); } if (is_heartbeat_cluster()) { use_born_on = TRUE; } else if (is_classic_ais_cluster()) { use_born_on = TRUE; } if(can_win == FALSE) { reason = "Not eligible"; we_loose = TRUE; } else if (our_node == NULL || crm_is_peer_active(our_node) == FALSE) { reason = "We are not part of the cluster"; log_level = LOG_ERR; we_loose = TRUE; } else if (election_id != e->count && crm_str_eq(our_node->uuid, election_owner, TRUE)) { log_level = LOG_TRACE; reason = "Superseded"; done = TRUE; } else if (your_node == NULL || crm_is_peer_active(your_node) == FALSE) { /* Possibly we cached the message in the FSA queue at a point that it wasn't */ reason = "Peer is not part of our cluster"; log_level = LOG_WARNING; done = TRUE; } else if (crm_str_eq(op, CRM_OP_NOVOTE, TRUE)) { char *op_copy = strdup(op); char *uname_copy = strdup(from); CRM_ASSERT(crm_str_eq(our_node->uuid, election_owner, TRUE)); /* update the list of nodes that have voted */ g_hash_table_replace(e->voted, uname_copy, op_copy); reason = "Recorded"; done = TRUE; } else { struct timeval your_age; const char *your_version = crm_element_value(vote, F_CRM_VERSION); int tv_sec = 0; int tv_usec = 0; crm_element_value_int(vote, F_CRM_ELECTION_AGE_S, &tv_sec); crm_element_value_int(vote, F_CRM_ELECTION_AGE_US, &tv_usec); your_age.tv_sec = tv_sec; your_age.tv_usec = tv_usec; age = crm_compare_age(your_age); if (crm_str_eq(from, e->uname, TRUE)) { char *op_copy = strdup(op); char *uname_copy = strdup(from); CRM_ASSERT(crm_str_eq(our_node->uuid, election_owner, TRUE)); /* update ourselves in the list of nodes that have voted */ g_hash_table_replace(e->voted, uname_copy, op_copy); reason = "Recorded"; done = TRUE; } else if (compare_version(your_version, CRM_FEATURE_SET) < 0) { reason = "Version"; we_loose = TRUE; } else if (compare_version(your_version, CRM_FEATURE_SET) > 0) { reason = "Version"; } else if (age < 0) { reason = "Uptime"; we_loose = TRUE; } else if (age > 0) { reason = "Uptime"; /* TODO: Check for y(our) born < 0 */ } else if (use_born_on && your_node->born < our_node->born) { reason = "Born"; we_loose = TRUE; } else if (use_born_on && your_node->born > our_node->born) { reason = "Born"; } else if (e->uname == NULL) { reason = "Unknown host name"; we_loose = TRUE; } else if (strcasecmp(e->uname, from) > 0) { reason = "Host name"; we_loose = TRUE; } else { reason = "Host name"; CRM_ASSERT(strcasecmp(e->uname, from) < 0); /* can't happen... * } else if(strcasecmp(e->uname, from) == 0) { * */ } } if (expires < tm_now) { election_wins = 0; expires = tm_now + STORM_INTERVAL; } else if (done == FALSE && we_loose == FALSE) { int peers = 1 + g_hash_table_size(crm_peer_cache); /* If every node has to vote down every other node, thats N*(N-1) total elections * Allow some leway before _really_ complaining */ election_wins++; if (election_wins > (peers * peers)) { crm_warn("Election storm detected: %d elections in %d seconds", election_wins, STORM_INTERVAL); election_wins = 0; expires = tm_now + STORM_INTERVAL; crm_write_blackbox(0, NULL); } } if (done) { do_crm_log(log_level + 1, "Election %d (current: %d, owner: %s): Processed %s from %s (%s)", election_id, e->count, election_owner, op, from, reason); return e->state; } else if(we_loose == FALSE) { do_crm_log(log_level, "Election %d (owner: %s) pass: %s from %s (%s)", election_id, election_owner, op, from, reason); if (last_election_loss == 0 || tm_now - last_election_loss > (time_t) loss_dampen) { last_election_loss = 0; election_timeout_stop(e); /* Start a new election by voting down this, and other, peers */ e->state = election_start; return e->state; } crm_info("Election %d ignore: We already lost an election less than %ds ago (%s)", election_id, loss_dampen, ctime(&last_election_loss)); } novote = create_request(CRM_OP_NOVOTE, NULL, from, CRM_SYSTEM_CRMD, CRM_SYSTEM_CRMD, NULL); do_crm_log(log_level, "Election %d (owner: %s) lost: %s from %s (%s)", election_id, election_owner, op, from, reason); election_timeout_stop(e); crm_xml_add(novote, F_CRM_ELECTION_OWNER, election_owner); crm_xml_add_int(novote, F_CRM_ELECTION_ID, election_id); send_cluster_message(your_node, crm_msg_crmd, novote, TRUE); free_xml(novote); last_election_loss = tm_now; e->state = election_lost; return e->state; }
/* * CFG functionality stolen from node_name() in corosync-quorumtool.c * This resolves the first address assigned to a node and returns the name or IP address. */ char * corosync_node_name(uint64_t /*cmap_handle_t */ cmap_handle, uint32_t nodeid) { int lpc = 0; int rc = CS_OK; int retries = 0; char *name = NULL; cmap_handle_t local_handle = 0; /* nodeid == 0 == CMAN_NODEID_US */ if (nodeid == 0) { nodeid = get_local_nodeid(0); } if (cmap_handle == 0 && local_handle == 0) { retries = 0; crm_trace("Initializing CMAP connection"); do { rc = cmap_initialize(&local_handle); if (rc != CS_OK) { retries++; crm_debug("API connection setup failed: %s. Retrying in %ds", cs_strerror(rc), retries); sleep(retries); } } while (retries < 5 && rc != CS_OK); if (rc != CS_OK) { crm_warn("Could not connect to Cluster Configuration Database API, error %s", cs_strerror(rc)); local_handle = 0; } } if (cmap_handle == 0) { cmap_handle = local_handle; } while (name == NULL && cmap_handle != 0) { uint32_t id = 0; char *key = NULL; key = g_strdup_printf("nodelist.node.%d.nodeid", lpc); rc = cmap_get_uint32(cmap_handle, key, &id); crm_trace("Checking %u vs %u from %s", nodeid, id, key); g_free(key); if (rc != CS_OK) { break; } if (nodeid == id) { crm_trace("Searching for node name for %u in nodelist.node.%d %s", nodeid, lpc, name); if (name == NULL) { key = g_strdup_printf("nodelist.node.%d.ring0_addr", lpc); rc = cmap_get_string(cmap_handle, key, &name); crm_trace("%s = %s", key, name); if (node_name_is_valid(key, name) == FALSE) { free(name); name = NULL; } g_free(key); } if (name == NULL) { key = g_strdup_printf("nodelist.node.%d.name", lpc); rc = cmap_get_string(cmap_handle, key, &name); crm_trace("%s = %s %d", key, name, rc); g_free(key); } break; } lpc++; } if(local_handle) { cmap_finalize(local_handle); } if (name == NULL) { crm_notice("Unable to get node name for nodeid %u", nodeid); } return name; }
gboolean corosync_initialize_nodelist(void *cluster, gboolean force_member, xmlNode * xml_parent) { int lpc = 0; int rc = CS_OK; int retries = 0; gboolean any = FALSE; cmap_handle_t cmap_handle; do { rc = cmap_initialize(&cmap_handle); if (rc != CS_OK) { retries++; crm_debug("API connection setup failed: %s. Retrying in %ds", cs_strerror(rc), retries); sleep(retries); } } while (retries < 5 && rc != CS_OK); if (rc != CS_OK) { crm_warn("Could not connect to Cluster Configuration Database API, error %d", rc); return FALSE; } crm_peer_init(); crm_trace("Initializing corosync nodelist"); for (lpc = 0;; lpc++) { uint32_t nodeid = 0; char *name = NULL; char *key = NULL; key = g_strdup_printf("nodelist.node.%d.nodeid", lpc); rc = cmap_get_uint32(cmap_handle, key, &nodeid); g_free(key); if (rc != CS_OK) { break; } name = corosync_node_name(cmap_handle, nodeid); if (name != NULL) { GHashTableIter iter; crm_node_t *node = NULL; g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { if(node && node->uname && strcasecmp(node->uname, name) == 0) { if (node->id && node->id != nodeid) { crm_crit("Nodes %u and %u share the same name '%s': shutting down", node->id, nodeid, name); crm_exit(DAEMON_RESPAWN_STOP); } } } } if (nodeid > 0 || name != NULL) { crm_trace("Initializing node[%d] %u = %s", lpc, nodeid, name); crm_get_peer(nodeid, name); } if (nodeid > 0 && name != NULL) { any = TRUE; if (xml_parent) { xmlNode *node = create_xml_node(xml_parent, XML_CIB_TAG_NODE); crm_xml_add_int(node, XML_ATTR_ID, nodeid); crm_xml_add(node, XML_ATTR_UNAME, name); if (force_member) { crm_xml_add(node, XML_ATTR_TYPE, CRM_NODE_MEMBER); } } } free(name); } cmap_finalize(cmap_handle); return any; }
gboolean check_message_sanity(const AIS_Message * msg, const char *data) { gboolean sane = TRUE; int dest = msg->host.type; int tmp_size = msg->header.size - sizeof(AIS_Message); if (sane && msg->header.size == 0) { crm_warn("Message with no size"); sane = FALSE; } if (sane && msg->header.error != CS_OK) { crm_warn("Message header contains an error: %d", msg->header.error); sane = FALSE; } if (sane && ais_data_len(msg) != tmp_size) { crm_warn("Message payload size is incorrect: expected %d, got %d", ais_data_len(msg), tmp_size); sane = TRUE; } if (sane && ais_data_len(msg) == 0) { crm_warn("Message with no payload"); sane = FALSE; } if (sane && data && msg->is_compressed == FALSE) { int str_size = strlen(data) + 1; if (ais_data_len(msg) != str_size) { int lpc = 0; crm_warn("Message payload is corrupted: expected %d bytes, got %d", ais_data_len(msg), str_size); sane = FALSE; for (lpc = (str_size - 10); lpc < msg->size; lpc++) { if (lpc < 0) { lpc = 0; } crm_debug("bad_data[%d]: %d / '%c'", lpc, data[lpc], data[lpc]); } } } if (sane == FALSE) { crm_err("Invalid message %d: (dest=%s:%s, from=%s:%s.%u, compressed=%d, size=%d, total=%d)", msg->id, ais_dest(&(msg->host)), msg_type2text(dest), ais_dest(&(msg->sender)), msg_type2text(msg->sender.type), msg->sender.pid, msg->is_compressed, ais_data_len(msg), msg->header.size); } else { crm_trace ("Verfied message %d: (dest=%s:%s, from=%s:%s.%u, compressed=%d, size=%d, total=%d)", msg->id, ais_dest(&(msg->host)), msg_type2text(dest), ais_dest(&(msg->sender)), msg_type2text(msg->sender.type), msg->sender.pid, msg->is_compressed, ais_data_len(msg), msg->header.size); } return sane; }
static void operation_finished(mainloop_child_t * p, pid_t pid, int core, int signo, int exitcode) { svc_action_t *op = mainloop_child_userdata(p); char *prefix = crm_strdup_printf("%s:%d", op->id, op->pid); mainloop_clear_child_userdata(p); op->status = PCMK_LRM_OP_DONE; CRM_ASSERT(op->pid == pid); crm_trace("%s %p %p", prefix, op->opaque->stderr_gsource, op->opaque->stdout_gsource); if (op->opaque->stderr_gsource) { /* Make sure we have read everything from the buffer. * Depending on the priority mainloop gives the fd, operation_finished * could occur before all the reads are done. Force the read now.*/ crm_trace("%s dispatching stderr", prefix); dispatch_stderr(op); crm_trace("%s: %p", op->id, op->stderr_data); mainloop_del_fd(op->opaque->stderr_gsource); op->opaque->stderr_gsource = NULL; } if (op->opaque->stdout_gsource) { /* Make sure we have read everything from the buffer. * Depending on the priority mainloop gives the fd, operation_finished * could occur before all the reads are done. Force the read now.*/ crm_trace("%s dispatching stdout", prefix); dispatch_stdout(op); crm_trace("%s: %p", op->id, op->stdout_data); mainloop_del_fd(op->opaque->stdout_gsource); op->opaque->stdout_gsource = NULL; } if (signo) { if (mainloop_child_timeout(p)) { crm_warn("%s - timed out after %dms", prefix, op->timeout); op->status = PCMK_LRM_OP_TIMEOUT; op->rc = PCMK_OCF_TIMEOUT; } else { do_crm_log_unlikely((op->cancel) ? LOG_INFO : LOG_WARNING, "%s - terminated with signal %d", prefix, signo); op->status = PCMK_LRM_OP_ERROR; op->rc = PCMK_OCF_SIGNAL; } } else { op->rc = exitcode; crm_debug("%s - exited with rc=%d", prefix, exitcode); } free(prefix); prefix = crm_strdup_printf("%s:%d:stderr", op->id, op->pid); crm_log_output(LOG_NOTICE, prefix, op->stderr_data); free(prefix); prefix = crm_strdup_printf("%s:%d:stdout", op->id, op->pid); crm_log_output(LOG_DEBUG, prefix, op->stdout_data); free(prefix); operation_finalize(op); }
gboolean cib_action_update(crm_action_t *action, int status, int op_rc) { char *op_id = NULL; char *code = NULL; char *digest = NULL; xmlNode *tmp = NULL; xmlNode *params = NULL; xmlNode *state = NULL; xmlNode *rsc = NULL; xmlNode *xml_op = NULL; xmlNode *action_rsc = NULL; enum cib_errors rc = cib_ok; const char *name = NULL; const char *value = NULL; const char *rsc_id = NULL; const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); const char *task_uuid = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY); const char *target_uuid = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID); int call_options = cib_quorum_override|cib_scope_local; if(status == LRM_OP_PENDING) { crm_debug("%s %d: Recording pending operation %s on %s", crm_element_name(action->xml), action->id, task_uuid, target); } else { crm_warn("%s %d: %s on %s timed out", crm_element_name(action->xml), action->id, task_uuid, target); } action_rsc = find_xml_node(action->xml, XML_CIB_TAG_RESOURCE, TRUE); if(action_rsc == NULL) { return FALSE; } rsc_id = ID(action_rsc); CRM_CHECK(rsc_id != NULL, crm_log_xml_err(action->xml, "Bad:action"); return FALSE); /* update the CIB <node_state id="hadev"> <lrm> <lrm_resources> <lrm_resource id="rsc2" last_op="start" op_code="0" target="hadev"/> */ state = create_xml_node(NULL, XML_CIB_TAG_STATE); crm_xml_add(state, XML_ATTR_UUID, target_uuid); crm_xml_add(state, XML_ATTR_UNAME, target); rsc = create_xml_node(state, XML_CIB_TAG_LRM); crm_xml_add(rsc, XML_ATTR_ID, target_uuid); rsc = create_xml_node(rsc, XML_LRM_TAG_RESOURCES); rsc = create_xml_node(rsc, XML_LRM_TAG_RESOURCE); crm_xml_add(rsc, XML_ATTR_ID, rsc_id); name = XML_ATTR_TYPE; value = crm_element_value(action_rsc, name); crm_xml_add(rsc, name, value); name = XML_AGENT_ATTR_CLASS; value = crm_element_value(action_rsc, name); crm_xml_add(rsc, name, value); name = XML_AGENT_ATTR_PROVIDER; value = crm_element_value(action_rsc, name); crm_xml_add(rsc, name, value); xml_op = create_xml_node(rsc, XML_LRM_TAG_RSC_OP); crm_xml_add(xml_op, XML_ATTR_ID, task); op_id = generate_op_key(rsc_id, task, action->interval); crm_xml_add(xml_op, XML_ATTR_ID, op_id); crm_free(op_id); crm_xml_add_int(xml_op, XML_LRM_ATTR_CALLID, -1); crm_xml_add(xml_op, XML_LRM_ATTR_TASK, task); crm_xml_add(xml_op, XML_ATTR_CRM_VERSION, CRM_FEATURE_SET); crm_xml_add_int(xml_op, XML_LRM_ATTR_OPSTATUS, status); crm_xml_add_int(xml_op, XML_LRM_ATTR_INTERVAL, action->interval); crm_xml_add_int(xml_op, XML_LRM_ATTR_RC, op_rc); crm_xml_add(xml_op, XML_ATTR_ORIGIN, __FUNCTION__); if(crm_str_eq(task, CRMD_ACTION_MIGRATED, TRUE)) { char *key = crm_meta_name("migrate_source_uuid"); xmlNode *attrs = first_named_child(action->xml, XML_TAG_ATTRS); const char *host = crm_element_value(attrs, key); CRM_CHECK(host != NULL, crm_log_xml_err(action->xml, "Bad Op")); crm_xml_add(xml_op, CRMD_ACTION_MIGRATED, host); crm_free(key); } code = generate_transition_key( transition_graph->id, action->id, get_target_rc(action), te_uuid); crm_xml_add(xml_op, XML_ATTR_TRANSITION_KEY, code); crm_free(code); code = generate_transition_magic( crm_element_value(xml_op, XML_ATTR_TRANSITION_KEY), status, op_rc); crm_xml_add(xml_op, XML_ATTR_TRANSITION_MAGIC, code); crm_free(code); tmp = find_xml_node(action->xml, "attributes", TRUE); params = create_xml_node(NULL, XML_TAG_PARAMS); copy_in_properties(params, tmp); filter_action_parameters(params, CRM_FEATURE_SET); digest = calculate_xml_digest(params, TRUE, FALSE); /* info for now as this area has been problematic to debug */ crm_debug("Calculated digest %s for %s (%s)\n", digest, ID(xml_op), crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC)); crm_log_xml(LOG_DEBUG, "digest:source", params); crm_xml_add(xml_op, XML_LRM_ATTR_OP_DIGEST, digest); crm_free(digest); free_xml(params); crm_debug_3("Updating CIB with \"%s\" (%s): %s %s on %s", status<0?"new action":XML_ATTR_TIMEOUT, crm_element_name(action->xml), crm_str(task), rsc_id, target); rc = fsa_cib_conn->cmds->update( fsa_cib_conn, XML_CIB_TAG_STATUS, state, call_options); crm_debug_2("Updating CIB with %s action %d: %s on %s (call_id=%d)", op_status2text(status), action->id, task_uuid, target, rc); add_cib_op_callback(fsa_cib_conn, rc, FALSE, NULL, cib_action_updated); free_xml(state); action->sent_update = TRUE; if(rc < cib_ok) { return FALSE; } return TRUE; }
static void action_synced_wait(svc_action_t * op, sigset_t mask) { #ifndef HAVE_SYS_SIGNALFD_H CRM_ASSERT(FALSE); #else int status = 0; int timeout = op->timeout; int sfd = -1; time_t start = -1; struct pollfd fds[3]; int wait_rc = 0; sfd = signalfd(-1, &mask, SFD_NONBLOCK); if (sfd < 0) { crm_perror(LOG_ERR, "signalfd() failed"); } fds[0].fd = op->opaque->stdout_fd; fds[0].events = POLLIN; fds[0].revents = 0; fds[1].fd = op->opaque->stderr_fd; fds[1].events = POLLIN; fds[1].revents = 0; fds[2].fd = sfd; fds[2].events = POLLIN; fds[2].revents = 0; crm_trace("Waiting for %d", op->pid); start = time(NULL); do { int poll_rc = poll(fds, 3, timeout); if (poll_rc > 0) { if (fds[0].revents & POLLIN) { svc_read_output(op->opaque->stdout_fd, op, FALSE); } if (fds[1].revents & POLLIN) { svc_read_output(op->opaque->stderr_fd, op, TRUE); } if (fds[2].revents & POLLIN) { struct signalfd_siginfo fdsi; ssize_t s; s = read(sfd, &fdsi, sizeof(struct signalfd_siginfo)); if (s != sizeof(struct signalfd_siginfo)) { crm_perror(LOG_ERR, "Read from signal fd %d failed", sfd); } else if (fdsi.ssi_signo == SIGCHLD) { wait_rc = waitpid(op->pid, &status, WNOHANG); if (wait_rc < 0){ crm_perror(LOG_ERR, "waitpid() for %d failed", op->pid); } else if (wait_rc > 0) { break; } } } } else if (poll_rc == 0) { timeout = 0; break; } else if (poll_rc < 0) { if (errno != EINTR) { crm_perror(LOG_ERR, "poll() failed"); break; } } timeout = op->timeout - (time(NULL) - start) * 1000; } while ((op->timeout < 0 || timeout > 0)); crm_trace("Child done: %d", op->pid); if (wait_rc <= 0) { int killrc = kill(op->pid, SIGKILL); op->rc = PCMK_OCF_UNKNOWN_ERROR; if (op->timeout > 0 && timeout <= 0) { op->status = PCMK_LRM_OP_TIMEOUT; crm_warn("%s:%d - timed out after %dms", op->id, op->pid, op->timeout); } else { op->status = PCMK_LRM_OP_ERROR; } if (killrc && errno != ESRCH) { crm_err("kill(%d, KILL) failed: %d", op->pid, errno); } /* * From sigprocmask(2): * It is not possible to block SIGKILL or SIGSTOP. Attempts to do so are silently ignored. * * This makes it safe to skip WNOHANG here */ waitpid(op->pid, &status, 0); } else if (WIFEXITED(status)) { op->status = PCMK_LRM_OP_DONE; op->rc = WEXITSTATUS(status); crm_info("Managed %s process %d exited with rc=%d", op->id, op->pid, op->rc); } else if (WIFSIGNALED(status)) { int signo = WTERMSIG(status); op->status = PCMK_LRM_OP_ERROR; crm_err("Managed %s process %d exited with signal=%d", op->id, op->pid, signo); } #ifdef WCOREDUMP if (WCOREDUMP(status)) { crm_err("Managed %s process %d dumped core", op->id, op->pid); } #endif svc_read_output(op->opaque->stdout_fd, op, FALSE); svc_read_output(op->opaque->stderr_fd, op, TRUE); close(op->opaque->stdout_fd); close(op->opaque->stderr_fd); close(sfd); #endif }
/* A_CIB_STOP, A_CIB_START, A_CIB_RESTART, */ void do_cib_control(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t *msg_data) { struct crm_subsystem_s *this_subsys = cib_subsystem; long long stop_actions = A_CIB_STOP; long long start_actions = A_CIB_START; if(action & stop_actions) { if (fsa_cib_conn->state != cib_disconnected && last_resource_update != 0) { crm_info("Waiting for resource update %d to complete", last_resource_update); crmd_fsa_stall(NULL); return; } crm_info("Disconnecting CIB"); clear_bit_inplace(fsa_input_register, R_CIB_CONNECTED); CRM_ASSERT(fsa_cib_conn != NULL); fsa_cib_conn->cmds->del_notify_callback( fsa_cib_conn, T_CIB_DIFF_NOTIFY, do_cib_updated); if(fsa_cib_conn->state != cib_disconnected) { fsa_cib_conn->cmds->set_slave( fsa_cib_conn, cib_scope_local); fsa_cib_conn->cmds->signoff(fsa_cib_conn); } } if(action & start_actions) { int rc = cib_ok; CRM_ASSERT(fsa_cib_conn != NULL); if(cur_state == S_STOPPING) { crm_err("Ignoring request to start %s after shutdown", this_subsys->name); return; } rc = fsa_cib_conn->cmds->signon( fsa_cib_conn, CRM_SYSTEM_CRMD, cib_command); if(rc != cib_ok) { /* a short wait that usually avoids stalling the FSA */ sleep(1); rc = fsa_cib_conn->cmds->signon( fsa_cib_conn, CRM_SYSTEM_CRMD, cib_command); } if(rc != cib_ok){ crm_info("Could not connect to the CIB service: %s", cib_error2string(rc)); } else if(cib_ok != fsa_cib_conn->cmds->set_connection_dnotify( fsa_cib_conn, crmd_cib_connection_destroy)) { crm_err("Could not set dnotify callback"); } else if(cib_ok != fsa_cib_conn->cmds->add_notify_callback( fsa_cib_conn, T_CIB_REPLACE_NOTIFY, do_cib_replaced)) { crm_err("Could not set CIB notification callback"); } else if(cib_ok != fsa_cib_conn->cmds->add_notify_callback( fsa_cib_conn, T_CIB_DIFF_NOTIFY, do_cib_updated)) { crm_err("Could not set CIB notification callback"); } else { set_bit_inplace( fsa_input_register, R_CIB_CONNECTED); } if(is_set(fsa_input_register, R_CIB_CONNECTED) == FALSE) { cib_retries++; crm_warn("Couldn't complete CIB registration %d" " times... pause and retry", cib_retries); if(cib_retries < 30) { crm_timer_start(wait_timer); crmd_fsa_stall(NULL); } else { crm_err("Could not complete CIB" " registration %d times..." " hard error", cib_retries); register_fsa_error( C_FSA_INTERNAL, I_ERROR, NULL); } } else { int call_id = 0; crm_info("CIB connection established"); call_id = fsa_cib_conn->cmds->query( fsa_cib_conn, NULL, NULL, cib_scope_local); add_cib_op_callback(fsa_cib_conn, call_id, FALSE, NULL, revision_check_callback); cib_retries = 0; } } }
enum cib_errors cib_process_replace(const char *op, int options, const char *section, xmlNode * req, xmlNode * input, xmlNode * existing_cib, xmlNode ** result_cib, xmlNode ** answer) { const char *tag = NULL; gboolean verbose = FALSE; enum cib_errors result = cib_ok; crm_debug_2("Processing \"%s\" event for section=%s", op, crm_str(section)); if (options & cib_xpath) { return cib_process_xpath(op, options, section, req, input, existing_cib, result_cib, answer); } *answer = NULL; if (input == NULL) { return cib_NOOBJECT; } tag = crm_element_name(input); if (options & cib_verbose) { verbose = TRUE; } if (safe_str_eq(XML_CIB_TAG_SECTION_ALL, section)) { section = NULL; } else if (safe_str_eq(tag, section)) { section = NULL; } if (safe_str_eq(tag, XML_TAG_CIB)) { int updates = 0; int epoch = 0; int admin_epoch = 0; int replace_updates = 0; int replace_epoch = 0; int replace_admin_epoch = 0; const char *reason = NULL; cib_version_details(existing_cib, &admin_epoch, &epoch, &updates); cib_version_details(input, &replace_admin_epoch, &replace_epoch, &replace_updates); if (replace_admin_epoch < admin_epoch) { reason = XML_ATTR_GENERATION_ADMIN; } else if (replace_admin_epoch > admin_epoch) { /* no more checks */ } else if (replace_epoch < epoch) { reason = XML_ATTR_GENERATION; } else if (replace_epoch > epoch) { /* no more checks */ } else if (replace_updates < updates) { reason = XML_ATTR_NUMUPDATES; } if (reason != NULL) { crm_warn("Replacement %d.%d.%d not applied to %d.%d.%d:" " current %s is greater than the replacement", replace_admin_epoch, replace_epoch, replace_updates, admin_epoch, epoch, updates, reason); result = cib_old_data; } free_xml(*result_cib); *result_cib = copy_xml(input); } else { xmlNode *obj_root = NULL; gboolean ok = TRUE; obj_root = get_object_root(section, *result_cib); ok = replace_xml_child(NULL, obj_root, input, FALSE); if (ok == FALSE) { crm_debug_2("No matching object to replace"); result = cib_NOTEXISTS; } } return result; }
/*! * \brief Find a transition event that would have made a specified node down * * \param[in] id If nonzero, also consider this action ID a match * \param[in] target UUID of node to match * \param[in] filter If not NULL, only match CRM actions of this type * \param[in] quiet If FALSE, log a warning if no match found * * \return Matching event if found, NULL otherwise * * \note "Down" events are CRM_OP_FENCE and CRM_OP_SHUTDOWN. * \todo This should detect normal pacemaker_remote node stop events, * where action->type is action_type_rsc, * XML_LRM_ATTR_TASK is CRMD_ACTION_STOP, * and the affected resource creates a remote node that matches target. * Then, peer_update_callback() could ignore these. */ crm_action_t * match_down_event(int id, const char *target, const char *filter, bool quiet) { const char *this_action = NULL; const char *this_node = NULL; crm_action_t *match = NULL; GListPtr gIter = NULL; GListPtr gIter2 = NULL; gIter = transition_graph->synapses; for (; gIter != NULL; gIter = gIter->next) { synapse_t *synapse = (synapse_t *) gIter->data; /* lookup event */ gIter2 = synapse->actions; for (; gIter2 != NULL; gIter2 = gIter2->next) { crm_action_t *action = (crm_action_t *) gIter2->data; if (id > 0 && action->id == id) { match = action; break; } this_action = crm_element_value(action->xml, XML_LRM_ATTR_TASK); if (action->type != action_type_crm) { continue; } else if (filter != NULL && safe_str_neq(this_action, filter)) { continue; } else if (safe_str_neq(this_action, CRM_OP_FENCE) && safe_str_neq(this_action, CRM_OP_SHUTDOWN)) { continue; } this_node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID); if (this_node == NULL) { crm_log_xml_err(action->xml, "No node uuid"); } if (safe_str_neq(this_node, target)) { crm_trace("Action %d node %s is not a match for %s", action->id, this_node, target); continue; } match = action; id = action->id; break; } if (match != NULL) { break; } } if (match != NULL) { crm_debug("Match found for action %d: %s on %s", id, crm_element_value(match->xml, XML_LRM_ATTR_TASK_KEY), target); } else if (id > 0) { crm_err("No match for action %d", id); } else if(quiet == FALSE) { crm_warn("No match for shutdown action on %s", target); } return match; }
/*! * \internal * \brief Create a new remote stonith op * \param client, he local stonith client id that initaited the operation * \param request, The request from the client that started the operation * \param peer, Is this operation owned by another stonith peer? Operations * owned by other peers are stored on all the stonith nodes, but only the * owner executes the operation. All the nodes get the results to the operation * once the owner finishes executing it. */ void * create_remote_stonith_op(const char *client, xmlNode * request, gboolean peer) { remote_fencing_op_t *op = NULL; xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, request, LOG_TRACE); int call_options = 0; if (remote_op_list == NULL) { remote_op_list = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, free_remote_op); } /* If this operation is owned by another node, check to make * sure we haven't already created this operation. */ if (peer && dev) { const char *op_id = crm_element_value(dev, F_STONITH_REMOTE_OP_ID); CRM_CHECK(op_id != NULL, return NULL); op = g_hash_table_lookup(remote_op_list, op_id); if (op) { crm_debug("%s already exists", op_id); return op; } } op = calloc(1, sizeof(remote_fencing_op_t)); crm_element_value_int(request, F_STONITH_TIMEOUT, (int *)&(op->base_timeout)); if (peer && dev) { op->id = crm_element_value_copy(dev, F_STONITH_REMOTE_OP_ID); } else { op->id = crm_generate_uuid(); } g_hash_table_replace(remote_op_list, op->id, op); CRM_LOG_ASSERT(g_hash_table_lookup(remote_op_list, op->id) != NULL); crm_trace("Created %s", op->id); op->state = st_query; op->replies_expected = fencing_active_peers(); op->action = crm_element_value_copy(dev, F_STONITH_ACTION); op->originator = crm_element_value_copy(dev, F_STONITH_ORIGIN); op->delegate = crm_element_value_copy(dev, F_STONITH_DELEGATE); /* May not be set */ op->created = time(NULL); if (op->originator == NULL) { /* Local or relayed request */ op->originator = strdup(stonith_our_uname); } CRM_LOG_ASSERT(client != NULL); if (client) { op->client_id = strdup(client); } op->client_name = crm_element_value_copy(request, F_STONITH_CLIENTNAME); op->target = crm_element_value_copy(dev, F_STONITH_TARGET); op->request = copy_xml(request); /* TODO: Figure out how to avoid this */ crm_element_value_int(request, F_STONITH_CALLOPTS, &call_options); op->call_options = call_options; crm_element_value_int(request, F_STONITH_CALLID, (int *)&(op->client_callid)); crm_trace("%s new stonith op: %s - %s of %s for %s", (peer && dev) ? "Recorded" : "Generated", op->id, op->action, op->target, op->client_name); if (op->call_options & st_opt_cs_nodeid) { int nodeid = crm_atoi(op->target, NULL); crm_node_t *node = crm_get_peer(nodeid, NULL); /* Ensure the conversion only happens once */ op->call_options &= ~st_opt_cs_nodeid; if (node && node->uname) { free(op->target); op->target = strdup(node->uname); } else { crm_warn("Could not expand nodeid '%s' into a host name (%p)", op->target, node); } } /* check to see if this is a duplicate operation of another in-flight operation */ merge_duplicates(op); return op; }
gboolean read_config(void) { int rc = CS_OK; int retries = 0; gboolean have_log = FALSE; char *logging_debug = NULL; char *logging_logfile = NULL; char *logging_to_logfile = NULL; char *logging_to_syslog = NULL; char *logging_syslog_facility = NULL; enum cluster_type_e stack = pcmk_cluster_unknown; #if HAVE_CONFDB char *value = NULL; confdb_handle_t config; confdb_handle_t top_handle = 0; hdb_handle_t local_handle; static confdb_callbacks_t callbacks = { }; do { rc = confdb_initialize(&config, &callbacks); if(rc != CS_OK) { retries++; printf("Connection setup failed: %d. Retrying in %ds\n", rc, retries); sleep(retries); } else { break; } } while(retries < 5); #elif HAVE_CMAP cmap_handle_t local_handle; /* There can be only one (possibility if confdb isn't around) */ do { rc = cmap_initialize(&local_handle); if(rc != CS_OK) { retries++; printf("API connection setup failed: %s. Retrying in %ds\n", cs_strerror(rc), retries); crm_info("API connection setup failed: %s. Retrying in %ds", cs_strerror(rc), retries); sleep(retries); } else { break; } } while(retries < 5); #endif if (rc != CS_OK) { printf("Could not connect to Cluster Configuration Database API, error %d\n", rc); crm_warn("Could not connect to Cluster Configuration Database API, error %d", rc); return FALSE; } stack = get_cluster_type(); crm_info("Reading configure for stack: %s", name_for_cluster_type(stack)); /* =::=::= Should we be here =::=::= */ if (stack == pcmk_cluster_corosync) { setenv("HA_cluster_type", "corosync", 1); setenv("HA_quorum_type", "corosync", 1); #if HAVE_CONFDB } else if (stack == pcmk_cluster_cman) { setenv("HA_cluster_type", "cman", 1); setenv("HA_quorum_type", "cman", 1); enable_crmd_as_root(TRUE); use_cman = TRUE; } else if (stack == pcmk_cluster_classic_ais) { setenv("HA_cluster_type", "openais", 1); setenv("HA_quorum_type", "pcmk", 1); /* Look for a service block to indicate our plugin is loaded */ top_handle = config_find_init(config); local_handle = config_find_next(config, "service", top_handle); while (local_handle) { get_config_opt(config, local_handle, "name", &value, NULL); if (safe_str_eq("pacemaker", value)) { get_config_opt(config, local_handle, "ver", &value, "0"); if (safe_str_eq(value, "1")) { get_config_opt(config, local_handle, "use_logd", &value, "no"); setenv("HA_use_logd", value, 1); setenv("HA_LOGD", value, 1); get_config_opt(config, local_handle, "use_mgmtd", &value, "no"); enable_mgmtd(crm_is_true(value)); } else { crm_err("We can only start Pacemaker from init if using version 1" " of the Pacemaker plugin for Corosync. Terminating."); exit(100); } break; } local_handle = config_find_next(config, "service", top_handle); } free(value); #endif } else { crm_err("Unsupported stack type: %s", name_for_cluster_type(stack)); return FALSE; } #if HAVE_CONFDB top_handle = config_find_init(config); local_handle = config_find_next(config, "logging", top_handle); get_config_opt(config, local_handle, "debug", &logging_debug, "off"); get_config_opt(config, local_handle, "logfile", &logging_logfile, "/var/log/pacemaker"); get_config_opt(config, local_handle, "to_logfile", &logging_to_logfile, "off"); get_config_opt(config, local_handle, "to_syslog", &logging_to_syslog, "on"); get_config_opt(config, local_handle, "syslog_facility", &logging_syslog_facility, "daemon"); confdb_finalize(config); #elif HAVE_CMAP /* =::=::= Logging =::=::= */ get_config_opt(local_handle, "logging.debug", &logging_debug, "off"); get_config_opt(local_handle, "logging.logfile", &logging_logfile, "/var/log/pacemaker"); get_config_opt(local_handle, "logging.to_logfile", &logging_to_logfile, "off"); get_config_opt(local_handle, "logging.to_syslog", &logging_to_syslog, "on"); get_config_opt(local_handle, "logging.syslog_facility", &logging_syslog_facility, "daemon"); cmap_finalize(local_handle); #endif if (crm_is_true(logging_debug)) { setenv("HA_debug", "1", 1); if(get_crm_log_level() < LOG_DEBUG) { set_crm_log_level(LOG_DEBUG); } } else { setenv("HA_debug", "0", 1); } if (crm_is_true(logging_to_logfile)) { if(crm_add_logfile(logging_logfile)) { setenv("HA_debugfile", logging_logfile, 1); setenv("HA_DEBUGLOG", logging_logfile, 1); setenv("HA_LOGFILE", logging_logfile, 1); have_log = TRUE; } else { crm_err("Couldn't create logfile: %s", logging_logfile); } } if (have_log && crm_is_true(logging_to_syslog) == FALSE) { crm_info("User configured file based logging and explicitly disabled syslog."); free(logging_syslog_facility); logging_syslog_facility = NULL; } else { if (crm_is_true(logging_to_syslog) == FALSE) { crm_err ("Please enable some sort of logging, either 'to_logfile: on' or 'to_syslog: on'."); crm_err("If you use file logging, be sure to also define a value for 'logfile'"); } } if(logging_syslog_facility) { setenv("HA_logfacility", logging_syslog_facility, 1); setenv("HA_LOGFACILITY", logging_syslog_facility, 1); } else { unsetenv("HA_logfacility"); unsetenv("HA_LOGFACILITY"); } free(logging_debug); free(logging_logfile); free(logging_to_logfile); free(logging_to_syslog); free(logging_syslog_facility); return TRUE; }
/* coverity[-alloc] Memory is referenced in one or both hashtables */ crm_node_t * crm_get_peer(unsigned int id, const char *uname) { crm_node_t *node = NULL; CRM_ASSERT(id > 0 || uname != NULL); crm_peer_init(); if (node == NULL && uname != NULL) { node = g_hash_table_lookup(crm_peer_cache, uname); } if (node == NULL && id > 0) { node = g_hash_table_lookup(crm_peer_id_cache, GUINT_TO_POINTER(id)); if (node && node->uname && uname) { crm_crit("Node %s and %s share the same cluster node id '%u'!", node->uname, uname, id); /* NOTE: Calling crm_new_peer() means the entry in * crm_peer_id_cache will point to the new entity * * TO-DO: Replace the old uname instead? */ node = NULL; } } if (node == NULL) { crm_debug("Creating entry for node %s/%u", uname, id); node = calloc(1, sizeof(crm_node_t)); CRM_ASSERT(node); } if (id > 0 && node->id != id) { node->id = id; crm_info("Node %s now has id: %u", crm_str(uname), id); g_hash_table_replace(crm_peer_id_cache, GUINT_TO_POINTER(node->id), node); } if (uname && node->uname == NULL) { node->uname = strdup(uname); crm_info("Node %u is now known as %s", id, uname); g_hash_table_replace(crm_peer_cache, node->uname, node); if (crm_status_callback) { crm_status_callback(crm_status_uname, node, NULL); } } if (node && node->uname && node->uuid == NULL) { const char *uuid = get_node_uuid(id, node->uname); if(uuid) { node->uuid = strdup(uuid); crm_info("Node %u has uuid %s", id, node->uuid); } else { crm_warn("Cannot obtain a UUID for node %d/%s", id, node->uname); } } return node; }
gboolean upstart_job_exec(svc_action_t * op, gboolean synchronous) { char *job = NULL; int arg_wait = TRUE; const char *arg_env = "pacemaker=1"; const char *action = op->action; DBusError error; DBusMessage *msg = NULL; DBusMessage *reply = NULL; DBusMessageIter iter, array_iter; op->rc = PCMK_OCF_UNKNOWN_ERROR; CRM_ASSERT(upstart_init()); if (safe_str_eq(op->action, "meta-data")) { op->stdout_data = upstart_job_metadata(op->agent); op->rc = PCMK_OCF_OK; goto cleanup; } if(!upstart_job_by_name(op->agent, &job, op->timeout)) { crm_debug("Could not obtain job named '%s' to %s", op->agent, action); if (!g_strcmp0(action, "stop")) { op->rc = PCMK_OCF_OK; } else { op->rc = PCMK_OCF_NOT_INSTALLED; op->status = PCMK_LRM_OP_NOT_INSTALLED; } goto cleanup; } if (safe_str_eq(op->action, "monitor") || safe_str_eq(action, "status")) { char *path = get_first_instance(job, op->timeout); op->rc = PCMK_OCF_NOT_RUNNING; if(path) { DBusPendingCall *pending = NULL; char *state = pcmk_dbus_get_property( upstart_proxy, BUS_NAME, path, UPSTART_06_API ".Instance", "state", op->synchronous?NULL:upstart_job_check, op, op->synchronous?NULL:&pending, op->timeout); free(job); free(path); if(op->synchronous) { upstart_job_check("state", state, op); free(state); return op->rc == PCMK_OCF_OK; } else if (pending) { services_set_op_pending(op, pending); return TRUE; } return FALSE; } goto cleanup; } else if (!g_strcmp0(action, "start")) { action = "Start"; } else if (!g_strcmp0(action, "stop")) { action = "Stop"; } else if (!g_strcmp0(action, "restart")) { action = "Restart"; } else { op->rc = PCMK_OCF_UNIMPLEMENT_FEATURE; goto cleanup; } crm_debug("Calling %s for %s on %s", action, op->rsc, job); msg = dbus_message_new_method_call(BUS_NAME, // target for the method call job, // object to call on UPSTART_JOB_IFACE, // interface to call on action); // method name CRM_ASSERT(msg != NULL); dbus_message_iter_init_append (msg, &iter); CRM_LOG_ASSERT(dbus_message_iter_open_container (&iter, DBUS_TYPE_ARRAY, DBUS_TYPE_STRING_AS_STRING, &array_iter)); CRM_LOG_ASSERT(dbus_message_iter_append_basic (&array_iter, DBUS_TYPE_STRING, &arg_env)); CRM_LOG_ASSERT(dbus_message_iter_close_container (&iter, &array_iter)); CRM_LOG_ASSERT(dbus_message_append_args(msg, DBUS_TYPE_BOOLEAN, &arg_wait, DBUS_TYPE_INVALID)); if (op->synchronous == FALSE) { DBusPendingCall* pending = pcmk_dbus_send(msg, upstart_proxy, upstart_async_dispatch, op, op->timeout); free(job); if(pending) { services_set_op_pending(op, pending); return TRUE; } return FALSE; } dbus_error_init(&error); reply = pcmk_dbus_send_recv(msg, upstart_proxy, &error, op->timeout); if(error.name) { if(!upstart_mask_error(op, error.name)) { crm_err("Could not issue %s for %s: %s (%s)", action, op->rsc, error.name, job); } } else if (!g_strcmp0(op->action, "stop")) { /* No return vaue */ op->rc = PCMK_OCF_OK; } else if(!pcmk_dbus_type_check(reply, NULL, DBUS_TYPE_OBJECT_PATH, __FUNCTION__, __LINE__)) { crm_warn("Call to %s passed but return type was unexpected", op->action); op->rc = PCMK_OCF_OK; } else { const char *path = NULL; dbus_message_get_args (reply, NULL, DBUS_TYPE_OBJECT_PATH, &path, DBUS_TYPE_INVALID); crm_info("Call to %s passed: %s", op->action, path); op->rc = PCMK_OCF_OK; } cleanup: free(job); if(msg) { dbus_message_unref(msg); } if(reply) { dbus_message_unref(reply); } if (op->synchronous == FALSE) { operation_finalize(op); return TRUE; } return op->rc == PCMK_OCF_OK; }
void attrd_peer_update(crm_node_t *peer, xmlNode *xml, const char *host, bool filter) { bool changed = FALSE; attribute_t *a; attribute_value_t *v = NULL; int dampen = 0; const char *op = crm_element_value(xml, F_ATTRD_TASK); const char *attr = crm_element_value(xml, F_ATTRD_ATTRIBUTE); const char *value = crm_element_value(xml, F_ATTRD_VALUE); const char *dvalue = crm_element_value(xml, F_ATTRD_DAMPEN); if (attr == NULL) { crm_warn("Peer update did not specify attribute"); return; } a = g_hash_table_lookup(attributes, attr); if(a == NULL) { if (op == NULL /* The xml children from an ATTRD_OP_SYNC_RESPONSE have no F_ATTRD_TASK */ || safe_str_eq(op, ATTRD_OP_UPDATE) || safe_str_eq(op, ATTRD_OP_UPDATE_BOTH)) { a = create_attribute(xml); } else { crm_warn("Update error (attribute %s not found)", attr); return; } } if (op == NULL /* The xml children from an ATTRD_OP_SYNC_RESPONSE have no F_ATTRD_TASK */ || safe_str_eq(op, ATTRD_OP_UPDATE_BOTH) || safe_str_eq(op, ATTRD_OP_UPDATE_DELAY)) { if (dvalue) { dampen = crm_get_msec(dvalue); if (dampen >= 0) { if (a->timeout_ms != dampen) { mainloop_timer_stop(a->timer); mainloop_timer_del(a->timer); a->timeout_ms = dampen; if (dampen > 0) { a->timer = mainloop_timer_add(a->id, a->timeout_ms, FALSE, attribute_timer_cb, a); crm_info("Update attribute %s with delay %dms (%s)", a->id, dampen, dvalue); } else { a->timer = NULL; crm_info("Update attribute %s with not delay", a->id); } //if dampen is changed, attrd writes in a current value immediately. write_or_elect_attribute(a); if (safe_str_eq(op, ATTRD_OP_UPDATE_DELAY)) { return; } } else { if (safe_str_eq(op, ATTRD_OP_UPDATE_DELAY)) { crm_trace("Unchanged attribute %s with delay %dms (%s).(ATTRD_OP_UPDATE_DELAY)", a->id, dampen, dvalue); return; } } } else { crm_warn("Update error (A positive number is necessary for delay parameter. attribute %s : %dms (%s))", a->id, dampen, dvalue); return; } } else { crm_warn("Update error (delay parameter is necessary for the update of the attribute %s)", a->id); return; } } if(host == NULL) { GHashTableIter vIter; g_hash_table_iter_init(&vIter, a->values); crm_debug("Setting %s for all hosts to %s", attr, value); xml_remove_prop(xml, F_ATTRD_HOST_ID); while (g_hash_table_iter_next(&vIter, (gpointer *) & host, NULL)) { attrd_peer_update(peer, xml, host, filter); } return; } v = attrd_lookup_or_create_value(a->values, host, xml); if(filter && safe_str_neq(v->current, value) && safe_str_eq(host, attrd_cluster->uname)) { xmlNode *sync = create_xml_node(NULL, __FUNCTION__); crm_notice("%s[%s]: local value '%s' takes priority over '%s' from %s", a->id, host, v->current, value, peer->uname); crm_xml_add(sync, F_ATTRD_TASK, ATTRD_OP_SYNC_RESPONSE); v = g_hash_table_lookup(a->values, host); build_attribute_xml(sync, a->id, a->set, a->uuid, a->timeout_ms, a->user, a->is_private, v->nodename, v->nodeid, v->current); crm_xml_add_int(sync, F_ATTRD_WRITER, election_state(writer)); send_attrd_message(peer, sync); free_xml(sync); } else if(safe_str_neq(v->current, value)) { crm_info("Setting %s[%s]: %s -> %s from %s", attr, host, v->current, value, peer->uname); free(v->current); if(value) { v->current = strdup(value); } else { v->current = NULL; } changed = TRUE; } else { crm_trace("Unchanged %s[%s] from %s is %s", attr, host, peer->uname, value); } a->changed |= changed; if(changed) { if(a->timer) { crm_trace("Delayed write out (%dms) for %s", a->timeout_ms, a->id); mainloop_timer_start(a->timer); } else { write_or_elect_attribute(a); } } /* this only involves cluster nodes. */ if(v->nodeid == 0 && (v->is_remote == FALSE)) { if(crm_element_value_int(xml, F_ATTRD_HOST_ID, (int*)&v->nodeid) == 0) { /* Create the name/id association */ crm_node_t *peer = crm_get_peer(v->nodeid, host); crm_trace("We know %s's node id now: %s", peer->uname, peer->uuid); if(election_state(writer) == election_won) { write_attributes(FALSE, TRUE); return; } } } }
static ssize_t crm_cs_flush(gpointer data) { int sent = 0; ssize_t rc = 0; int queue_len = 0; static unsigned int last_sent = 0; cpg_handle_t *handle = (cpg_handle_t *)data; if (*handle == 0) { crm_trace("Connection is dead"); return pcmk_ok; } queue_len = g_list_length(cs_message_queue); if ((queue_len % 1000) == 0 && queue_len > 1) { crm_err("CPG queue has grown to %d", queue_len); } else if (queue_len == CS_SEND_MAX) { crm_warn("CPG queue has grown to %d", queue_len); } if (cs_message_timer) { /* There is already a timer, wait until it goes off */ crm_trace("Timer active %d", cs_message_timer); return pcmk_ok; } while (cs_message_queue && sent < CS_SEND_MAX) { struct iovec *iov = cs_message_queue->data; errno = 0; rc = cpg_mcast_joined(*handle, CPG_TYPE_AGREED, iov, 1); if (rc != CS_OK) { break; } sent++; last_sent++; crm_trace("CPG message sent, size=%d", iov->iov_len); cs_message_queue = g_list_remove(cs_message_queue, iov); free(iov->iov_base); free(iov); } queue_len -= sent; if (sent > 1 || cs_message_queue) { crm_info("Sent %d CPG messages (%d remaining, last=%u): %s (%d)", sent, queue_len, last_sent, ais_error2text(rc), rc); } else { crm_trace("Sent %d CPG messages (%d remaining, last=%u): %s (%d)", sent, queue_len, last_sent, ais_error2text(rc), rc); } if (cs_message_queue) { uint32_t delay_ms = 100; if(rc != CS_OK) { /* Proportionally more if sending failed but cap at 1s */ delay_ms = QB_MIN(1000, CS_SEND_MAX + (10 * queue_len)); } cs_message_timer = g_timeout_add(delay_ms, crm_cs_flush_cb, data); } return rc; }
int init_remote_listener(int port, gboolean encrypted) { int ssock; struct sockaddr_in saddr; int optval; if (port <= 0) { /* dont start it */ return 0; } if (encrypted) { #ifndef HAVE_GNUTLS_GNUTLS_H crm_warn("TLS support is not available"); return 0; #else crm_notice("Starting a tls listener on port %d.", port); gnutls_global_init(); /* gnutls_global_set_log_level (10); */ gnutls_global_set_log_function(debug_log); gnutls_dh_params_init(&dh_params); gnutls_dh_params_generate2(dh_params, DH_BITS); gnutls_anon_allocate_server_credentials(&anon_cred_s); gnutls_anon_set_server_dh_params(anon_cred_s, dh_params); #endif } else { crm_warn("Starting a plain_text listener on port %d.", port); } #ifndef HAVE_PAM crm_warn("PAM is _not_ enabled!"); #endif /* create server socket */ ssock = socket(AF_INET, SOCK_STREAM, 0); if (ssock == -1) { crm_perror(LOG_ERR, "Can not create server socket." ERROR_SUFFIX); return -1; } /* reuse address */ optval = 1; setsockopt(ssock, SOL_SOCKET, SO_REUSEADDR, &optval, sizeof(optval)); /* bind server socket */ memset(&saddr, '\0', sizeof(saddr)); saddr.sin_family = AF_INET; saddr.sin_addr.s_addr = INADDR_ANY; saddr.sin_port = htons(port); if (bind(ssock, (struct sockaddr *)&saddr, sizeof(saddr)) == -1) { crm_perror(LOG_ERR, "Can not bind server socket." ERROR_SUFFIX); return -2; } if (listen(ssock, 10) == -1) { crm_perror(LOG_ERR, "Can not start listen." ERROR_SUFFIX); return -3; } G_main_add_fd(G_PRIORITY_HIGH, ssock, FALSE, cib_remote_listen, NULL, default_ipc_connection_destroy); return ssock; }
/* A_DC_JOIN_PROCESS_ACK */ void do_dc_join_ack(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { int join_id = -1; int call_id = 0; ha_msg_input_t *join_ack = fsa_typed_data(fsa_dt_ha_msg); const char *op = crm_element_value(join_ack->msg, F_CRM_TASK); const char *join_from = crm_element_value(join_ack->msg, F_CRM_HOST_FROM); crm_node_t *peer = crm_get_peer(0, join_from); if (safe_str_neq(op, CRM_OP_JOIN_CONFIRM) || peer == NULL) { crm_debug("Ignoring op=%s message from %s", op, join_from); return; } crm_trace("Processing ack from %s", join_from); crm_element_value_int(join_ack->msg, F_CRM_JOIN_ID, &join_id); if (peer->join != crm_join_finalized) { crm_info("Join not in progress: ignoring join-%d from %s (phase = %d)", join_id, join_from, peer->join); return; } else if (join_id != current_join_id) { crm_err("Invalid response from %s: join-%d vs. join-%d", join_from, join_id, current_join_id); crm_update_peer_join(__FUNCTION__, peer, crm_join_nack); return; } crm_update_peer_join(__FUNCTION__, peer, crm_join_confirmed); crm_info("join-%d: Updating node state to %s for %s", join_id, CRMD_JOINSTATE_MEMBER, join_from); /* update CIB with the current LRM status from the node * We don't need to notify the TE of these updates, a transition will * be started in due time */ erase_status_tag(join_from, XML_CIB_TAG_LRM, cib_scope_local); if (safe_str_eq(join_from, fsa_our_uname)) { xmlNode *now_dc_lrmd_state = do_lrm_query(TRUE, fsa_our_uname); if (now_dc_lrmd_state != NULL) { crm_debug("LRM state is updated from do_lrm_query.(%s)", join_from); fsa_cib_update(XML_CIB_TAG_STATUS, now_dc_lrmd_state, cib_scope_local | cib_quorum_override | cib_can_create, call_id, NULL); free(now_dc_lrmd_state); } else { crm_warn("Could not get our LRM state. LRM state is updated from join_ack->xml.(%s)", join_from); fsa_cib_update(XML_CIB_TAG_STATUS, join_ack->xml, cib_scope_local | cib_quorum_override | cib_can_create, call_id, NULL); } } else { crm_debug("LRM state is updated from join_ack->xml.(%s)", join_from); fsa_cib_update(XML_CIB_TAG_STATUS, join_ack->xml, cib_scope_local | cib_quorum_override | cib_can_create, call_id, NULL); } fsa_register_cib_callback(call_id, FALSE, NULL, join_update_complete_callback); crm_debug("join-%d: Registered callback for LRM update %d", join_id, call_id); }
void cib_notify_client(gpointer key, gpointer value, gpointer user_data) { IPC_Channel *ipc_client = NULL; HA_Message *update_msg = user_data; cib_client_t *client = value; const char *type = NULL; gboolean is_pre = FALSE; gboolean is_post = FALSE; gboolean is_confirm = FALSE; gboolean is_replace = FALSE; gboolean is_diff = FALSE; gboolean do_send = FALSE; int qlen = 0; int max_qlen = 0; CRM_DEV_ASSERT(client != NULL); CRM_DEV_ASSERT(update_msg != NULL); type = cl_get_string(update_msg, F_SUBTYPE); CRM_DEV_ASSERT(type != NULL); if(client == NULL) { crm_warn("Skipping NULL client"); return; } else if(client->channel == NULL) { crm_warn("Skipping client with NULL channel"); return; } else if(client->name == NULL) { crm_debug_2("Skipping unnammed client / comamnd channel"); return; } if(safe_str_eq(type, T_CIB_PRE_NOTIFY)) { is_pre = TRUE; } else if(safe_str_eq(type, T_CIB_POST_NOTIFY)) { is_post = TRUE; } else if(safe_str_eq(type, T_CIB_UPDATE_CONFIRM)) { is_confirm = TRUE; } else if(safe_str_eq(type, T_CIB_DIFF_NOTIFY)) { is_diff = TRUE; } else if(safe_str_eq(type, T_CIB_REPLACE_NOTIFY)) { is_replace = TRUE; } ipc_client = client->channel; qlen = ipc_client->send_queue->current_qlen; max_qlen = ipc_client->send_queue->max_qlen; #if 1 /* get_chan_status() causes memory to be allocated that isnt free'd * until the message is read (which messes up the memory stats) */ if(ipc_client->ops->get_chan_status(ipc_client) != IPC_CONNECT) { crm_debug_2("Skipping notification to disconnected" " client %s/%s", client->name, client->id); } else if(client->pre_notify && is_pre) { if(qlen < (int)(0.4 * max_qlen)) { do_send = TRUE; } else { crm_warn("Throttling pre-notifications due to" " high load: queue=%d (max=%d)", qlen, max_qlen); } } else if(client->post_notify && is_post) { if(qlen < (int)(0.7 * max_qlen)) { do_send = TRUE; } else { crm_warn("Throttling post-notifications due to" " extreme load: queue=%d (max=%d)", qlen, max_qlen); } /* these are critical */ } else #endif if(client->diffs && is_diff) { do_send = TRUE; } else if(client->confirmations && is_confirm) { do_send = TRUE; } else if(client->replace && is_replace) { do_send = TRUE; } if(do_send) { crm_debug_2("Notifying client %s/%s of %s update (queue=%d)", client->name, client->channel_name, type, qlen); if(ipc_client->send_queue->current_qlen >= ipc_client->send_queue->max_qlen) { /* We never want the CIB to exit because our client is slow */ crm_crit("%s-notification of client %s/%s failed - queue saturated", is_confirm?"Confirmation":is_post?"Post":"Pre", client->name, client->id); } else if(send_ipc_message(ipc_client, update_msg) == FALSE) { crm_warn("Notification of client %s/%s failed", client->name, client->id); } } else { crm_debug_3("Client %s/%s not interested in %s notifications", client->name, client->channel_name, type); } }
gboolean read_config(void) { int rc = CS_OK; int retries = 0; gboolean have_log = FALSE; char *logging_debug = NULL; char *logging_logfile = NULL; char *logging_to_logfile = NULL; char *logging_to_syslog = NULL; char *logging_syslog_facility = NULL; enum cluster_type_e stack = pcmk_cluster_unknown; #if HAVE_CONFDB char *value = NULL; confdb_handle_t config; confdb_handle_t top_handle = 0; hdb_handle_t local_handle; static confdb_callbacks_t callbacks = { }; do { rc = confdb_initialize(&config, &callbacks); if(rc < 0) { retries++; printf("Connection setup failed: %d. Retrying in %ds\n", rc, retries); sleep(retries); } else { break; } } while(retries < 5); #elif HAVE_CMAP cmap_handle_t local_handle; /* There can be only one (possibility if confdb isn't around) */ do { rc = cmap_initialize(&local_handle); if(rc < 0) { retries++; printf("Connection setup failed: %s. Retrying in %ds\n", cs_strerror(rc), retries); sleep(retries); } else { break; } } while(retries < 5); #endif if (rc != CS_OK) { printf("Could not initialize Cluster Configuration Database API instance, error %d\n", rc); return FALSE; } stack = get_cluster_type(); crm_info("Reading configure for stack: %s", name_for_cluster_type(stack)); /* =::=::= Should we be here =::=::= */ if (stack == pcmk_cluster_corosync) { setenv("HA_cluster_type", "corosync", 1); setenv("HA_quorum_type", "corosync", 1); #if HAVE_CONFDB } else if (stack == pcmk_cluster_cman) { setenv("HA_cluster_type", "cman", 1); setenv("HA_quorum_type", "cman", 1); enable_crmd_as_root(TRUE); use_cman = TRUE; } else if (stack == pcmk_cluster_classic_ais) { setenv("HA_cluster_type", "openais", 1); setenv("HA_quorum_type", "pcmk", 1); /* Look for a service block to indicate our plugin is loaded */ top_handle = config_find_init(config); local_handle = config_find_next(config, "service", top_handle); while (local_handle) { crm_free(value); get_config_opt(config, local_handle, "name", &value, NULL); if (safe_str_eq("pacemaker", value)) { crm_free(value); get_config_opt(config, local_handle, "ver", &value, "0"); if (safe_str_eq(value, "1")) { crm_free(value); get_config_opt(config, local_handle, "use_logd", &value, "no"); setenv("HA_use_logd", value, 1); setenv("HA_LOGD", value, 1); crm_free(value); get_config_opt(config, local_handle, "use_mgmtd", &value, "no"); enable_mgmtd(crm_is_true(value)); } else { crm_err("We can only start Pacemaker from init if using version 1" " of the Pacemaker plugin for Corosync. Terminating."); exit(100); } break; } local_handle = config_find_next(config, "service", top_handle); } crm_free(value); #endif } else { crm_err("Unsupported stack type: %s", name_for_cluster_type(stack)); return FALSE; } #if HAVE_CONFDB top_handle = config_find_init(config); local_handle = config_find_next(config, "logging", top_handle); get_config_opt(config, local_handle, "debug", &logging_debug, "off"); get_config_opt(config, local_handle, "logfile", &logging_logfile, "/var/log/pacemaker"); get_config_opt(config, local_handle, "to_logfile", &logging_to_logfile, "off"); get_config_opt(config, local_handle, "to_syslog", &logging_to_syslog, "on"); get_config_opt(config, local_handle, "syslog_facility", &logging_syslog_facility, "daemon"); confdb_finalize(config); #elif HAVE_CMAP /* =::=::= Logging =::=::= */ get_config_opt(local_handle, "logging.debug", &logging_debug, "off"); get_config_opt(local_handle, "logging.logfile", &logging_logfile, "/var/log/pacemaker"); get_config_opt(local_handle, "logging.to_logfile", &logging_to_logfile, "off"); get_config_opt(local_handle, "logging.to_syslog", &logging_to_syslog, "on"); get_config_opt(local_handle, "logging.syslog_facility", &logging_syslog_facility, "daemon"); cmap_finalize(local_handle); #endif if (crm_is_true(logging_debug)) { setenv("HA_debug", "1", 1); if(get_crm_log_level() < LOG_DEBUG) { set_crm_log_level(LOG_DEBUG); } } else { setenv("HA_debug", "0", 1); } if (crm_is_true(logging_to_logfile)) { struct stat parent; uid_t pcmk_uid = 0; uid_t pcmk_gid = getegid(); FILE *logfile = NULL; char *parent_dir = NULL; struct passwd *pcmk_user = getpwnam(CRM_DAEMON_USER); if(pcmk_user) { pcmk_uid = pcmk_user->pw_uid; } else { crm_err("User %s does not exist. Terminating", CRM_DAEMON_USER); exit(100); } parent_dir = dirname(strdup(logging_logfile)); rc = stat(parent_dir, &parent); if (rc != 0) { crm_err("Directory '%s' does not exist for logfile '%s'", parent_dir, logging_logfile); } else if (parent.st_uid == pcmk_uid && (parent.st_mode & (S_IRUSR | S_IWUSR))) { /* all good - user */ logfile = fopen(logging_logfile, "a"); } else if (parent.st_gid == pcmk_gid && (parent.st_mode & S_IXGRP)) { /* all good - group */ logfile = fopen(logging_logfile, "a"); } else { crm_err ("Daemons running as %s do not have permission to access '%s'. Logging to '%s' is disabled", CRM_DAEMON_USER, parent_dir, logging_logfile); crm_err ("Either %s must be owned by %s with rwx permissions, or have rwx permissions for group %d", parent_dir, CRM_DAEMON_USER, pcmk_gid); } if (logfile) { int logfd = fileno(logfile); setenv("HA_debugfile", logging_logfile, 1); setenv("HA_DEBUGLOG", logging_logfile, 1); setenv("HA_LOGFILE", logging_logfile, 1); /* Ensure the file has the correct permissions */ rc = fchown(logfd, pcmk_uid, pcmk_gid); if(rc < 0) { crm_warn("Cannot change the ownership of %s to user %s and gid %d", logging_logfile, CRM_DAEMON_USER, pcmk_gid); } rc = fchmod(logfd, S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP); if(rc < 0) { crm_warn("Cannot change the mode of %s to rw-rw----", logging_logfile); } have_log = TRUE; fprintf(logfile, "Set r/w permissions for uid=%d, gid=%d on %s\n", pcmk_uid, pcmk_gid, logging_logfile); if(fflush(logfile) < 0 || fsync(logfd) < 0) { crm_err("Couldn't write out logfile: %s", logging_logfile); have_log = FALSE; } fclose(logfile); } else { crm_err("Couldn't create logfile: %s", logging_logfile); } crm_free(parent_dir); } if (have_log && crm_is_true(logging_to_syslog) == FALSE) { crm_info("User configured file based logging and explicitly disabled syslog."); free(logging_syslog_facility); logging_syslog_facility = NULL; } else { if (crm_is_true(logging_to_syslog) == FALSE) { crm_err ("Please enable some sort of logging, either 'to_logfile: on' or 'to_syslog: on'."); crm_err("If you use file logging, be sure to also define a value for 'logfile'"); } } if(logging_syslog_facility) { setenv("HA_logfacility", logging_syslog_facility, 1); setenv("HA_LOGFACILITY", logging_syslog_facility, 1); } else { unsetenv("HA_logfacility"); unsetenv("HA_LOGFACILITY"); } free(logging_debug); free(logging_logfile); free(logging_to_logfile); free(logging_to_syslog); free(logging_syslog_facility); return TRUE; }