int main(int argc, char **argv) { int rc = 0; int flag = 0; int index = 0; #ifdef ENABLE_PCMK_REMOTE crm_log_init("pacemaker_remoted", LOG_INFO, TRUE, FALSE, argc, argv, FALSE); crm_set_options(NULL, "[options]", long_options, "Daemon for controlling services confirming to different standards"); #else crm_log_init("lrmd", LOG_INFO, TRUE, FALSE, argc, argv, FALSE); crm_set_options(NULL, "[options]", long_options, "Pacemaker Remote daemon for extending pacemaker functionality to remote nodes."); #endif while (1) { flag = crm_get_option(argc, argv, &index); if (flag == -1) { break; } switch (flag) { case 'r': break; case 'l': crm_add_logfile(optarg); break; case 'V': crm_bump_log_level(argc, argv); break; case '?': case '$': crm_help(flag, EX_OK); break; default: crm_help('?', EX_USAGE); break; } } /* Used by RAs - Leave owned by root */ crm_build_path(CRM_RSCTMP_DIR, 0755); /* Legacy: Used by RAs - Leave owned by root */ crm_build_path(HA_STATE_DIR"/heartbeat/rsctmp", 0755); rsc_list = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, free_rsc); ipcs = mainloop_add_ipc_server(CRM_SYSTEM_LRMD, QB_IPC_SHM, &lrmd_ipc_callbacks); if (ipcs == NULL) { crm_err("Failed to create IPC server: shutting down and inhibiting respawn"); crm_exit(DAEMON_RESPAWN_STOP); } #ifdef ENABLE_PCMK_REMOTE { const char *remote_port_str = getenv("PCMK_remote_port"); int remote_port = remote_port_str ? atoi(remote_port_str) : DEFAULT_REMOTE_PORT; if (lrmd_init_remote_tls_server(remote_port) < 0) { crm_err("Failed to create TLS server on port %d: shutting down and inhibiting respawn", remote_port); crm_exit(DAEMON_RESPAWN_STOP); } ipc_proxy_init(); } #endif mainloop_add_signal(SIGTERM, lrmd_shutdown); mainloop = g_main_new(FALSE); crm_info("Starting"); g_main_run(mainloop); mainloop_del_ipc_server(ipcs); #ifdef ENABLE_PCMK_REMOTE lrmd_tls_server_destroy(); ipc_proxy_cleanup(); #endif crm_client_cleanup(); g_hash_table_destroy(rsc_list); if (stonith_api) { stonith_api->cmds->disconnect(stonith_api); stonith_api_delete(stonith_api); } return rc; }
static void resource_ipc_connection_destroy(gpointer user_data) { crm_info("Connection to CRMd was terminated"); exit(1); }
/* CRM_OP_JOIN_ACKNACKを送信後、ノードからCRM_OP_JOIN_CONFIRMメッセージが送られて来たときの処理 */ 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 *join_id_s = NULL; const char *join_state = NULL; 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); if(safe_str_neq(op, CRM_OP_JOIN_CONFIRM)) { /* CRM_OP_JOIN_CONFIRMメッセージ以外のメッセージは無視してログを出力 */ crm_debug("Ignoring op=%s message from %s", op, join_from); return; } crm_element_value_int(join_ack->msg, F_CRM_JOIN_ID, &join_id); join_id_s = crm_element_value(join_ack->msg, F_CRM_JOIN_ID); /* now update them to "member" */ crm_debug_2("Processing ack from %s", join_from); /* CRM_OP_JOIN_CONFIRMを送ってきたノードがfinalized_nodesハッシュテーブルに含まれるかサーチする */ join_state = (const char *) g_hash_table_lookup(finalized_nodes, join_from); if(join_state == NULL) { /* 含まれていない場合は、無視 */ crm_err("Join not in progress: ignoring join-%d from %s", join_id, join_from); return; } else if(safe_str_neq(join_state, CRMD_JOINSTATE_MEMBER)) { crm_err("Node %s wasnt invited to join the cluster",join_from); /* 含まれていて、CRMD_JOINSTATE_MEMBER状態の場合は、finalized_nodesハッシュテーブルからCRM_OP_JOIN_CONFIRMを送ってきたノードを削除 */ g_hash_table_remove(finalized_nodes, join_from); 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); /* JOIN_IDが異なる場合も、finalized_nodesハッシュテーブルからCRM_OP_JOIN_CONFIRMを送ってきたノードを削除 */ g_hash_table_remove(finalized_nodes, join_from); return; } /* 上記以外の場合も、finalized_nodesハッシュテーブルからCRM_OP_JOIN_CONFIRMを送ってきたノードを削除 */ g_hash_table_remove(finalized_nodes, join_from); /* confirmed_nodesハッシュテーブルにRM_OP_JOIN_CONFIRMを送ってきたノードでサーチ */ if(g_hash_table_lookup(confirmed_nodes, join_from) != NULL) { /* 既に登録済みの場合は、エラーメッセージを表示 */ crm_err("join-%d: hash already contains confirmation from %s", join_id, join_from); } /* confirmed_nodesハッシュテーブルにCRM_OP_JOIN_CONFIRMを送ってきたノードを追加 */ g_hash_table_insert( confirmed_nodes, crm_strdup(join_from), crm_strdup(join_id_s)); /* join完了ログを出す */ 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 dont need to notify the TE of these updates, a transition will * be started in due time */ /* CIBからCRM_OP_JOIN_CONFIRMを送ってきたノードのXML_CIB_TAG_LRM情報を消去 */ erase_status_tag(join_from, XML_CIB_TAG_LRM, cib_scope_local); /* CIBを更新 */ /* CRM_OP_JOIN_CONFIRMを送ってきたノードのXML_CIB_TAG_LRM情報も追加される */ /* クラスタ構成ノードのLRMD情報が最新で更新 */ fsa_cib_update(XML_CIB_TAG_STATUS, join_ack->xml, cib_scope_local|cib_quorum_override|cib_can_create, call_id); /* CIBの更新コールバックをセット */ add_cib_op_callback( fsa_cib_conn, call_id, FALSE, NULL, join_update_complete_callback); crm_debug("join-%d: Registered callback for LRM update %d", join_id, call_id); }
crm_node_t * crm_update_peer(const char *source, unsigned int id, uint64_t born, uint64_t seen, int32_t votes, uint32_t children, const char *uuid, const char *uname, const char *addr, const char *state) { #if SUPPORT_PLUGIN gboolean addr_changed = FALSE; gboolean votes_changed = FALSE; #endif crm_node_t *node = NULL; id = get_corosync_id(id, uuid); node = crm_get_peer(id, uname); CRM_ASSERT(node != NULL); if (node->uuid == NULL) { if (is_openais_cluster()) { /* Yes, overrule whatever was passed in */ crm_peer_uuid(node); } else if (uuid != NULL) { node->uuid = strdup(uuid); } } if (children > 0) { crm_update_peer_proc(source, node, children, state); } if (state != NULL) { crm_update_peer_state(source, node, state, seen); } #if SUPPORT_HEARTBEAT if (born != 0) { node->born = born; } #endif #if SUPPORT_PLUGIN /* These were only used by the plugin */ if (born != 0) { node->born = born; } if (votes > 0 && node->votes != votes) { votes_changed = TRUE; node->votes = votes; } if (addr != NULL) { if (node->addr == NULL || crm_str_eq(node->addr, addr, FALSE) == FALSE) { addr_changed = TRUE; free(node->addr); node->addr = strdup(addr); } } if (addr_changed || votes_changed) { crm_info("%s: Node %s: id=%u state=%s addr=%s%s votes=%d%s born=" U64T " seen=" U64T " proc=%.32x", source, node->uname, node->id, node->state, node->addr, addr_changed ? " (new)" : "", node->votes, votes_changed ? " (new)" : "", node->born, node->last_seen, node->processes); } #endif return node; }
gboolean process_graph_event(xmlNode * event, const char *event_node) { int rc = -1; int status = -1; int callid = -1; int action = -1; int target_rc = -1; int transition_num = -1; char *update_te_uuid = NULL; gboolean stop_early = FALSE; gboolean passed = FALSE; const char *id = NULL; const char *desc = NULL; const char *magic = NULL; CRM_ASSERT(event != NULL); /* <lrm_rsc_op id="rsc_east-05_last_0" operation_key="rsc_east-05_monitor_0" operation="monitor" crm-debug-origin="do_update_resource" crm_feature_set="3.0.6" transition-key="9:2:7:be2e97d9-05e2-439d-863e-48f7aecab2aa" transition-magic="0:7;9:2:7:be2e97d9-05e2-439d-863e-48f7aecab2aa" call-id="17" rc-code="7" op-status="0" interval="0" last-run="1355361636" last-rc-change="1355361636" exec-time="128" queue-time="0" op-digest="c81f5f40b1c9e859c992e800b1aa6972"/> */ id = crm_element_value(event, XML_LRM_ATTR_TASK_KEY); crm_element_value_int(event, XML_LRM_ATTR_RC, &rc); crm_element_value_int(event, XML_LRM_ATTR_OPSTATUS, &status); crm_element_value_int(event, XML_LRM_ATTR_CALLID, &callid); magic = crm_element_value(event, XML_ATTR_TRANSITION_KEY); if (magic == NULL) { /* non-change */ return FALSE; } if (decode_transition_key(magic, &update_te_uuid, &transition_num, &action, &target_rc) == FALSE) { crm_err("Invalid event %s.%d detected: %s", id, callid, magic); abort_transition(INFINITY, tg_restart, "Bad event", event); return FALSE; } if (status == PCMK_LRM_OP_PENDING) { goto bail; } if (transition_num == -1) { desc = "initiated outside of the cluster"; abort_transition(INFINITY, tg_restart, "Unexpected event", event); } else if (action < 0 || crm_str_eq(update_te_uuid, te_uuid, TRUE) == FALSE) { desc = "initiated by a different node"; abort_transition(INFINITY, tg_restart, "Foreign event", event); stop_early = TRUE; /* This could be an lrm status refresh */ } else if (transition_graph->id != transition_num) { desc = "arrived really late"; abort_transition(INFINITY, tg_restart, "Old event", event); stop_early = TRUE; /* This could be an lrm status refresh */ } else if (transition_graph->complete) { desc = "arrived late"; abort_transition(INFINITY, tg_restart, "Inactive graph", event); } else if (match_graph_event(action, event, event_node, status, rc, target_rc) < 0) { desc = "unknown"; abort_transition(INFINITY, tg_restart, "Unknown event", event); } else if (rc == target_rc) { passed = TRUE; crm_trace("Processed update to %s: %s", id, magic); } if (passed == FALSE) { if (update_failcount(event, event_node, rc, target_rc, transition_num == -1)) { /* Turns out this wasn't an lrm status refresh update aferall */ stop_early = FALSE; desc = "failed"; } crm_info("Detected action (%d.%d) %s.%d=%s: %s", transition_num, action, id, callid, services_ocf_exitcode_str(rc), desc); } bail: free(update_te_uuid); return stop_early; }
void write_attribute(attribute_t *a) { int private_updates = 0, cib_updates = 0; xmlNode *xml_top = NULL; attribute_value_t *v = NULL; GHashTableIter iter; enum cib_call_options flags = cib_quorum_override; if (a == NULL) { return; } /* If this attribute will be written to the CIB ... */ if (!a->is_private) { /* Defer the write if now's not a good time */ if (the_cib == NULL) { crm_info("Write out of '%s' delayed: cib not connected", a->id); return; } else if (a->update && (a->update < last_cib_op_done)) { crm_info("Write out of '%s' continuing: update %d considered lost", a->id, a->update); } else if (a->update) { crm_info("Write out of '%s' delayed: update %d in progress", a->id, a->update); return; } else if (mainloop_timer_running(a->timer)) { crm_info("Write out of '%s' delayed: timer is running", a->id); return; } /* Initialize the status update XML */ xml_top = create_xml_node(NULL, XML_CIB_TAG_STATUS); } /* Attribute will be written shortly, so clear changed flag */ a->changed = FALSE; /* We will check all peers' uuids shortly, so initialize this to false */ a->unknown_peer_uuids = FALSE; /* Iterate over each peer value of this attribute */ g_hash_table_iter_init(&iter, a->values); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) & v)) { crm_node_t *peer = crm_get_peer_full(v->nodeid, v->nodename, CRM_GET_PEER_ANY); /* If the value's peer info does not correspond to a peer, ignore it */ if (peer == NULL) { crm_notice("Update error (peer not found): %s[%s]=%s failed (host=%p)", v->nodename, a->id, v->current, peer); continue; } /* If we're just learning the peer's node id, remember it */ if (peer->id && (v->nodeid == 0)) { crm_trace("Updating value's nodeid"); v->nodeid = peer->id; } /* If this is a private attribute, no update needs to be sent */ if (a->is_private) { private_updates++; continue; } /* If the peer is found, but its uuid is unknown, defer write */ if (peer->uuid == NULL) { a->unknown_peer_uuids = TRUE; crm_notice("Update error (unknown peer uuid, retry will be attempted once uuid is discovered): %s[%s]=%s failed (host=%p)", v->nodename, a->id, v->current, peer); continue; } /* Add this value to status update XML */ crm_debug("Update: %s[%s]=%s (%s %u %u %s)", v->nodename, a->id, v->current, peer->uuid, peer->id, v->nodeid, peer->uname); build_update_element(xml_top, a, peer->uuid, v->current); cib_updates++; free(v->requested); v->requested = NULL; if (v->current) { v->requested = strdup(v->current); } else { /* Older attrd versions don't know about the cib_mixed_update * flag so make sure it goes to the local cib which does */ flags |= cib_mixed_update|cib_scope_local; } } if (private_updates) { crm_info("Processed %d private change%s for %s, id=%s, set=%s", private_updates, ((private_updates == 1)? "" : "s"), a->id, (a->uuid? a->uuid : "<n/a>"), a->set); } if (cib_updates) { crm_log_xml_trace(xml_top, __FUNCTION__); a->update = cib_internal_op(the_cib, CIB_OP_MODIFY, NULL, XML_CIB_TAG_STATUS, xml_top, NULL, flags, a->user); crm_info("Sent update %d with %d changes for %s, id=%s, set=%s", a->update, cib_updates, a->id, (a->uuid? a->uuid : "<n/a>"), a->set); the_cib->cmds->register_callback_full(the_cib, a->update, 120, FALSE, strdup(a->id), "attrd_cib_callback", attrd_cib_callback, free); } free_xml(xml_top); }
enum crmd_fsa_input handle_request(xmlNode * stored_msg) { xmlNode *msg = NULL; const char *op = crm_element_value(stored_msg, F_CRM_TASK); /* Optimize this for the DC - it has the most to do */ if (op == NULL) { crm_log_xml_err(stored_msg, "Bad message"); return I_NULL; } /*========== DC-Only Actions ==========*/ if (AM_I_DC) { if (strcmp(op, CRM_OP_JOIN_ANNOUNCE) == 0) { return I_NODE_JOIN; } else if (strcmp(op, CRM_OP_JOIN_REQUEST) == 0) { return I_JOIN_REQUEST; } else if (strcmp(op, CRM_OP_JOIN_CONFIRM) == 0) { return I_JOIN_RESULT; } else if (strcmp(op, CRM_OP_SHUTDOWN) == 0) { const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM); gboolean dc_match = safe_str_eq(host_from, fsa_our_dc); if (is_set(fsa_input_register, R_SHUTDOWN)) { crm_info("Shutting ourselves down (DC)"); return I_STOP; } else if (dc_match) { crm_err("We didnt ask to be shut down, yet our" " TE is telling us too." " Better get out now!"); return I_TERMINATE; } else if (fsa_state != S_STOPPING) { crm_err("Another node is asking us to shutdown" " but we think we're ok."); return I_ELECTION; } } else if (strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0) { /* a slave wants to shut down */ /* create cib fragment and add to message */ return handle_shutdown_request(stored_msg); } } /*========== common actions ==========*/ if (strcmp(op, CRM_OP_NOVOTE) == 0) { ha_msg_input_t fsa_input; fsa_input.msg = stored_msg; register_fsa_input_adv(C_HA_MESSAGE, I_NULL, &fsa_input, A_ELECTION_COUNT | A_ELECTION_CHECK, FALSE, __FUNCTION__); } else if (strcmp(op, CRM_OP_CLEAR_FAILCOUNT) == 0) { return handle_failcount_op(stored_msg); } else if (strcmp(op, CRM_OP_VOTE) == 0) { /* count the vote and decide what to do after that */ ha_msg_input_t fsa_input; fsa_input.msg = stored_msg; register_fsa_input_adv(C_HA_MESSAGE, I_NULL, &fsa_input, A_ELECTION_COUNT | A_ELECTION_CHECK, FALSE, __FUNCTION__); /* Sometimes we _must_ go into S_ELECTION */ if (fsa_state == S_HALT) { crm_debug("Forcing an election from S_HALT"); return I_ELECTION; #if 0 } else if (AM_I_DC) { /* This is the old way of doing things but what is gained? */ return I_ELECTION; #endif } } else if (strcmp(op, CRM_OP_JOIN_OFFER) == 0) { crm_debug("Raising I_JOIN_OFFER: join-%s", crm_element_value(stored_msg, F_CRM_JOIN_ID)); return I_JOIN_OFFER; } else if (strcmp(op, CRM_OP_JOIN_ACKNAK) == 0) { crm_debug("Raising I_JOIN_RESULT: join-%s", crm_element_value(stored_msg, F_CRM_JOIN_ID)); return I_JOIN_RESULT; } else if (strcmp(op, CRM_OP_LRM_DELETE) == 0 || strcmp(op, CRM_OP_LRM_FAIL) == 0 || strcmp(op, CRM_OP_LRM_REFRESH) == 0 || strcmp(op, CRM_OP_REPROBE) == 0) { crm_xml_add(stored_msg, F_CRM_SYS_TO, CRM_SYSTEM_LRMD); return I_ROUTER; } else if (strcmp(op, CRM_OP_NOOP) == 0) { return I_NULL; } else if (strcmp(op, CRM_OP_LOCAL_SHUTDOWN) == 0) { crm_shutdown(SIGTERM); /*return I_SHUTDOWN; */ return I_NULL; /*========== (NOT_DC)-Only Actions ==========*/ } else if (AM_I_DC == FALSE && strcmp(op, CRM_OP_SHUTDOWN) == 0) { const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM); gboolean dc_match = safe_str_eq(host_from, fsa_our_dc); if (dc_match || fsa_our_dc == NULL) { if (is_set(fsa_input_register, R_SHUTDOWN) == FALSE) { crm_err("We didn't ask to be shut down, yet our" " DC is telling us too."); set_bit(fsa_input_register, R_STAYDOWN); return I_STOP; } crm_info("Shutting down"); return I_STOP; } else { crm_warn("Discarding %s op from %s", op, host_from); } } else if (strcmp(op, CRM_OP_PING) == 0) { /* eventually do some stuff to figure out * if we /are/ ok */ const char *sys_to = crm_element_value(stored_msg, F_CRM_SYS_TO); xmlNode *ping = create_xml_node(NULL, XML_CRM_TAG_PING); crm_xml_add(ping, XML_PING_ATTR_STATUS, "ok"); crm_xml_add(ping, XML_PING_ATTR_SYSFROM, sys_to); crm_xml_add(ping, "crmd_state", fsa_state2string(fsa_state)); /* Ok, so technically not so interesting, but CTS needs to see this */ crm_notice("Current ping state: %s", fsa_state2string(fsa_state)); msg = create_reply(stored_msg, ping); relay_message(msg, TRUE); free_xml(ping); free_xml(msg); } else if (strcmp(op, CRM_OP_RM_NODE_CACHE) == 0) { int id = 0; const char *name = NULL; xmlNode *options = get_xpath_object("//" XML_TAG_OPTIONS, stored_msg, LOG_ERR); if (options) { crm_element_value_int(options, XML_ATTR_ID, &id); name = crm_element_value(options, XML_ATTR_UNAME); } reap_crm_member(id, name); } else { crm_err("Unexpected request (%s) sent to %s", op, AM_I_DC ? "the DC" : "non-DC node"); crm_log_xml_err(stored_msg, "Unexpected"); } return I_NULL; }
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; }
void call_remote_stonith(remote_fencing_op_t * op, st_query_result_t * peer) { const char *device = NULL; int timeout = op->base_timeout; crm_trace("State for %s.%.8s: %s %d", op->target, op->client_name, op->id, op->state); if (peer == NULL && !is_set(op->call_options, st_opt_topology)) { peer = stonith_choose_peer(op); } if (!op->op_timer_total) { int total_timeout = get_op_total_timeout(op, peer, op->base_timeout); op->total_timeout = TIMEOUT_MULTIPLY_FACTOR * total_timeout; op->op_timer_total = g_timeout_add(1000 * op->total_timeout, remote_op_timeout, op); report_timeout_period(op, op->total_timeout); crm_info("Total remote op timeout set to %d for fencing of node %s for %s.%.8s", total_timeout, op->target, op->client_name, op->id); } if (is_set(op->call_options, st_opt_topology) && op->devices) { /* Ignore any preference, they might not have the device we need */ /* When using topology, the stonith_choose_peer function pops off * the peer from the op's query results. Make sure to calculate * the op_timeout before calling this function when topology is in use */ peer = stonith_choose_peer(op); device = op->devices->data; timeout = get_device_timeout(peer, device, op->base_timeout); } if (peer) { int timeout_one = 0; xmlNode *query = stonith_create_op(op->client_callid, op->id, STONITH_OP_FENCE, 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, timeout); if (device) { timeout_one = TIMEOUT_MULTIPLY_FACTOR * get_device_timeout(peer, device, op->base_timeout); crm_info("Requesting that %s perform op %s %s with %s for %s (%ds)", peer->host, op->action, op->target, device, op->client_name, timeout_one); crm_xml_add(query, F_STONITH_DEVICE, device); crm_xml_add(query, F_STONITH_MODE, "slave"); } else { timeout_one = TIMEOUT_MULTIPLY_FACTOR * get_peer_timeout(peer, op->base_timeout); crm_info("Requesting that %s perform op %s %s for %s (%ds)", peer->host, op->action, op->target, op->client_name, timeout_one); crm_xml_add(query, F_STONITH_MODE, "smart"); } op->state = st_exec; if (op->op_timer_one) { g_source_remove(op->op_timer_one); } op->op_timer_one = g_timeout_add((1000 * timeout_one), remote_op_timeout_one, op); send_cluster_message(crm_get_peer(0, peer->host), crm_msg_stonith_ng, query, FALSE); peer->tried = TRUE; free_xml(query); return; } else if (op->owner == FALSE) { crm_err("The termination of %s for %s is not ours to control", op->target, op->client_name); } else if (op->query_timer == 0) { /* We've exhausted all available peers */ crm_info("No remaining peers capable of terminating %s for %s (%d)", op->target, op->client_name, op->state); CRM_LOG_ASSERT(op->state < st_done); remote_op_timeout(op); } else if(op->replies >= op->replies_expected || op->replies >= fencing_active_peers()) { int rc = -EHOSTUNREACH; /* if the operation never left the query state, * but we have all the expected replies, then no devices * are available to execute the fencing operation. */ if (op->state == st_query) { crm_info("None of the %d peers have devices capable of terminating %s for %s (%d)", op->replies, op->target, op->client_name, op->state); rc = -ENODEV; } else { crm_info("None of the %d peers are capable of terminating %s for %s (%d)", op->replies, op->target, op->client_name, op->state); } op->state = st_failed; remote_op_done(op, NULL, rc, FALSE); } else if (device) { crm_info("Waiting for additional peers capable of terminating %s with %s for %s.%.8s", op->target, device, op->client_name, op->id); } else { crm_info("Waiting for additional peers capable of terminating %s for %s%.8s", op->target, op->client_name, op->id); } }
int process_remote_stonith_query(xmlNode * msg) { int devices = 0; gboolean host_is_target = FALSE; const char *id = NULL; const char *host = NULL; remote_fencing_op_t *op = NULL; st_query_result_t *result = NULL; uint32_t active = fencing_active_peers(); xmlNode *dev = get_xpath_object("//@" F_STONITH_REMOTE_OP_ID, msg, LOG_ERR); xmlNode *child = NULL; CRM_CHECK(dev != NULL, return -EPROTO); id = crm_element_value(dev, F_STONITH_REMOTE_OP_ID); CRM_CHECK(id != NULL, return -EPROTO); dev = get_xpath_object("//@st-available-devices", msg, LOG_ERR); CRM_CHECK(dev != NULL, return -EPROTO); crm_element_value_int(dev, "st-available-devices", &devices); op = g_hash_table_lookup(remote_op_list, id); if (op == NULL) { crm_debug("Unknown or expired remote op: %s", id); return -EOPNOTSUPP; } op->replies++; host = crm_element_value(msg, F_ORIG); host_is_target = safe_str_eq(host, op->target); if (devices <= 0) { /* If we're doing 'known' then we might need to fire anyway */ crm_trace("Query result from %s (%d devices)", host, devices); if(op->state == st_query && (op->replies >= op->replies_expected || op->replies >= active)) { crm_info("All queries have arrived, continuing (%d, %d, %d) ", op->replies_expected, active, op->replies); call_remote_stonith(op, NULL); } return pcmk_ok; } else if (host_is_target) { if (op->call_options & st_opt_allow_suicide) { crm_trace("Allowing %s to potentialy fence itself", op->target); } else { crm_info("Ignoring reply from %s, hosts are not permitted to commit suicide", op->target); return pcmk_ok; } } crm_info("Query result %d of %d from %s (%d devices)", op->replies, op->replies_expected, host, devices); result = calloc(1, sizeof(st_query_result_t)); result->host = strdup(host); result->devices = devices; result->custom_action_timeouts = g_hash_table_new_full(crm_str_hash, g_str_equal, free, NULL); result->verified_devices = g_hash_table_new_full(crm_str_hash, g_str_equal, free, NULL); for (child = __xml_first_child(dev); child != NULL; child = __xml_next(child)) { const char *device = ID(child); int action_timeout = 0; int verified = 0; if (device) { result->device_list = g_list_prepend(result->device_list, strdup(device)); crm_element_value_int(child, F_STONITH_ACTION_TIMEOUT, &action_timeout); crm_element_value_int(child, F_STONITH_DEVICE_VERIFIED, &verified); if (action_timeout) { crm_trace("Peer %s with device %s returned action timeout %d", result->host, device, action_timeout); g_hash_table_insert(result->custom_action_timeouts, strdup(device), GINT_TO_POINTER(action_timeout)); } if (verified) { crm_trace("Peer %s has confirmed a verified device %s", result->host, device); g_hash_table_insert(result->verified_devices, strdup(device), GINT_TO_POINTER(verified)); } } } CRM_CHECK(devices == g_list_length(result->device_list), crm_err("Mis-match: Query claimed to have %d devices but %d found", devices, g_list_length(result->device_list))); op->query_results = g_list_insert_sorted(op->query_results, result, sort_peers); if (is_set(op->call_options, st_opt_topology)) { /* If we start the fencing before all the topology results are in, * it is possible fencing levels will be skipped because of the missing * query results. */ if (op->state == st_query && all_topology_devices_found(op)) { /* All the query results are in for the topology, start the fencing ops. */ crm_trace("All topology devices found"); call_remote_stonith(op, result); } else if(op->state == st_query && (op->replies >= op->replies_expected || op->replies >= active)) { crm_info("All topology queries have arrived, continuing (%d, %d, %d) ", op->replies_expected, active, op->replies); call_remote_stonith(op, NULL); } } else if (op->state == st_query) { /* We have a result for a non-topology fencing op that looks promising, * go ahead and start fencing before query timeout */ if (host_is_target == FALSE && g_hash_table_size(result->verified_devices)) { /* we have a verified device living on a peer that is not the target */ crm_trace("Found %d verified devices", g_hash_table_size(result->verified_devices)); call_remote_stonith(op, result); } else if (safe_str_eq(op->action, "on")) { crm_trace("Unfencing %s", op->target); call_remote_stonith(op, result); } else if(op->replies >= op->replies_expected || op->replies >= active) { crm_info("All queries have arrived, continuing (%d, %d, %d) ", op->replies_expected, active, op->replies); call_remote_stonith(op, NULL); } else { crm_trace("Waiting for more peer results before launching fencing operation"); } } else if (op->state == st_done) { crm_info("Discarding query result from %s (%d devices): Operation is in state %d", result->host, result->devices, op->state); } return pcmk_ok; }
int process_remote_stonith_exec(xmlNode * msg) { int rc = 0; const char *id = NULL; const char *device = NULL; remote_fencing_op_t *op = NULL; xmlNode *dev = get_xpath_object("//@" F_STONITH_REMOTE_OP_ID, msg, LOG_ERR); CRM_CHECK(dev != NULL, return -EPROTO); id = crm_element_value(dev, F_STONITH_REMOTE_OP_ID); CRM_CHECK(id != NULL, return -EPROTO); dev = get_xpath_object("//@" F_STONITH_RC, msg, LOG_ERR); CRM_CHECK(dev != NULL, return -EPROTO); crm_element_value_int(dev, F_STONITH_RC, &rc); device = crm_element_value(dev, F_STONITH_DEVICE); if (remote_op_list) { op = g_hash_table_lookup(remote_op_list, id); } if (op == NULL && rc == pcmk_ok) { /* Record successful fencing operations */ const char *client_id = crm_element_value(dev, F_STONITH_CLIENTID); op = create_remote_stonith_op(client_id, dev, TRUE); } if (op == NULL) { /* Could be for an event that began before we started */ /* TODO: Record the op for later querying */ crm_info("Unknown or expired remote op: %s", id); return -EOPNOTSUPP; } if (op->devices && device && safe_str_neq(op->devices->data, device)) { crm_err ("Received outdated reply for device %s (instead of %s) to %s node %s. Operation already timed out at remote level.", device, op->devices->data, op->action, op->target); return rc; } if (safe_str_eq(crm_element_value(msg, F_SUBTYPE), "broadcast")) { crm_debug("Marking call to %s for %s on behalf of %s@%s.%.8s: %s (%d)", op->action, op->target, op->client_name, op->id, op->originator, pcmk_strerror(rc), rc); if (rc == pcmk_ok) { op->state = st_done; } else { op->state = st_failed; } remote_op_done(op, msg, rc, FALSE); return pcmk_ok; } else if (safe_str_neq(op->originator, stonith_our_uname)) { /* If this isn't a remote level broadcast, and we are not the * originator of the operation, we should not be receiving this msg. */ crm_err ("%s received non-broadcast fencing result for operation it does not own (device %s targeting %s)", stonith_our_uname, device, op->target); return rc; } if (is_set(op->call_options, st_opt_topology)) { const char *device = crm_element_value(msg, F_STONITH_DEVICE); crm_notice("Call to %s for %s on behalf of %s@%s: %s (%d)", device, op->target, op->client_name, op->originator, pcmk_strerror(rc), rc); /* We own the op, and it is complete. broadcast the result to all nodes * and notify our local clients. */ if (op->state == st_done) { remote_op_done(op, msg, rc, FALSE); return rc; } /* An operation completed succesfully but has not yet been marked as done. * Continue the topology if more devices exist at the current level, otherwise * mark as done. */ if (rc == pcmk_ok) { if (op->devices) { /* Success, are there any more? */ op->devices = op->devices->next; } /* if no more devices at this fencing level, we are done, * else we need to contine with executing the next device in the list */ if (op->devices == NULL) { crm_trace("Marking complex fencing op for %s as complete", op->target); op->state = st_done; remote_op_done(op, msg, rc, FALSE); return rc; } } else { /* This device failed, time to try another topology level. If no other * levels are available, mark this operation as failed and report results. */ if (stonith_topology_next(op) != pcmk_ok) { op->state = st_failed; remote_op_done(op, msg, rc, FALSE); return rc; } } } else if (rc == pcmk_ok && op->devices == NULL) { crm_trace("All done for %s", op->target); op->state = st_done; remote_op_done(op, msg, rc, FALSE); return rc; } /* Retry on failure or execute the rest of the topology */ crm_trace("Next for %s on behalf of %s@%s (rc was %d)", op->target, op->originator, op->client_name, rc); call_remote_stonith(op, NULL); return rc; }
int 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; int result = pcmk_ok; crm_trace("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 -EINVAL; } tag = crm_element_name(input); 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; const char *peer = crm_element_value(req, F_ORIG); const char *digest = crm_element_value(req, XML_ATTR_DIGEST); if (digest) { const char *version = crm_element_value(req, XML_ATTR_CRM_VERSION); char *digest_verify = calculate_xml_versioned_digest(input, FALSE, TRUE, version ? version : CRM_FEATURE_SET); if (safe_str_neq(digest_verify, digest)) { crm_err("Digest mis-match on replace from %s: %s vs. %s (expected)", peer, digest_verify, digest); reason = "digest mismatch"; } else { crm_info("Digest matched on replace from %s: %s", peer, digest); } free(digest_verify); } else { crm_trace("No digest to verify"); } 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_info("Replacement %d.%d.%d from %s not applied to %d.%d.%d:" " current %s is greater than the replacement", replace_admin_epoch, replace_epoch, replace_updates, peer, admin_epoch, epoch, updates, reason); result = -pcmk_err_old_data; } else { crm_info("Replaced %d.%d.%d with %d.%d.%d from %s", admin_epoch, epoch, updates, replace_admin_epoch, replace_epoch, replace_updates, peer); } 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_trace("No matching object to replace"); result = -ENXIO; } } return result; }
int main(int argc, char **argv) { GListPtr lpc = NULL; gboolean process = TRUE; gboolean all_good = TRUE; enum transition_status graph_rc = -1; crm_graph_t *transition = NULL; ha_time_t *a_date = NULL; cib_t *cib_conn = NULL; xmlNode *cib_object = NULL; int argerr = 0; int flag; char *msg_buffer = NULL; gboolean optional = FALSE; pe_working_set_t data_set; const char *source = NULL; const char *xml_file = NULL; const char *dot_file = NULL; const char *graph_file = NULL; const char *input_file = NULL; const char *input_xml = NULL; /* disable glib's fancy allocators that can't be free'd */ GMemVTable vtable; vtable.malloc = malloc; vtable.realloc = realloc; vtable.free = free; vtable.calloc = calloc; vtable.try_malloc = malloc; vtable.try_realloc = realloc; g_mem_set_vtable(&vtable); crm_log_init_quiet(NULL, LOG_CRIT, FALSE, FALSE, argc, argv); crm_set_options(NULL, "[-?Vv] -[Xxp] {other options}", long_options, "Calculate the cluster's response to the supplied cluster state\n" "\nSuperceeded by crm_simulate and likely to be removed in a future release\n\n"); while (1) { int option_index = 0; flag = crm_get_option(argc, argv, &option_index); if (flag == -1) break; switch (flag) { case 'S': do_simulation = TRUE; break; case 'a': all_actions = TRUE; break; case 'w': inhibit_exit = TRUE; break; case 'X': /*use_stdin = TRUE;*/ input_xml = optarg; break; case 's': show_scores = TRUE; break; case 'U': show_utilization = TRUE; break; case 'x': xml_file = optarg; break; case 'd': use_date = optarg; break; case 'D': dot_file = optarg; break; case 'G': graph_file = optarg; break; case 'I': input_file = optarg; break; case 'V': crm_bump_log_level(); break; case 'L': USE_LIVE_CIB = TRUE; break; case '$': case '?': crm_help(flag, 0); break; default: fprintf(stderr, "Option -%c is not yet supported\n", flag); ++argerr; break; } } if (optind < argc) { printf("non-option ARGV-elements: "); while (optind < argc) { printf("%s ", argv[optind++]); } printf("\n"); } if (optind > argc) { ++argerr; } if (argerr) { crm_err("%d errors in option parsing", argerr); crm_help('?', 1); } update_all_trace_data(); /* again, so we see which trace points got updated */ if (USE_LIVE_CIB) { int rc = cib_ok; source = "live cib"; cib_conn = cib_new(); rc = cib_conn->cmds->signon(cib_conn, "ptest", cib_command); if (rc == cib_ok) { crm_info("Reading XML from: live cluster"); cib_object = get_cib_copy(cib_conn); } else { fprintf(stderr, "Live CIB query failed: %s\n", cib_error2string(rc)); return 3; } if (cib_object == NULL) { fprintf(stderr, "Live CIB query failed: empty result\n"); return 3; } } else if (xml_file != NULL) { source = xml_file; cib_object = filename2xml(xml_file); } else if (use_stdin) { source = "stdin"; cib_object = filename2xml(NULL); } else if (input_xml) { source = "input string"; cib_object = string2xml(input_xml); } if (cib_object == NULL && source) { fprintf(stderr, "Could not parse configuration input from: %s\n", source); return 4; } else if (cib_object == NULL) { fprintf(stderr, "No configuration specified\n"); crm_help('?', 1); } if (get_object_root(XML_CIB_TAG_STATUS, cib_object) == NULL) { create_xml_node(cib_object, XML_CIB_TAG_STATUS); } if (cli_config_update(&cib_object, NULL, FALSE) == FALSE) { free_xml(cib_object); return cib_STALE; } if (validate_xml(cib_object, NULL, FALSE) != TRUE) { free_xml(cib_object); return cib_dtd_validation; } if (input_file != NULL) { FILE *input_strm = fopen(input_file, "w"); if (input_strm == NULL) { crm_perror(LOG_ERR, "Could not open %s for writing", input_file); } else { msg_buffer = dump_xml_formatted(cib_object); if (fprintf(input_strm, "%s\n", msg_buffer) < 0) { crm_perror(LOG_ERR, "Write to %s failed", input_file); } fflush(input_strm); fclose(input_strm); crm_free(msg_buffer); } } if (use_date != NULL) { a_date = parse_date(&use_date); log_date(LOG_WARNING, "Set fake 'now' to", a_date, ha_log_date | ha_log_time); log_date(LOG_WARNING, "Set fake 'now' to (localtime)", a_date, ha_log_date | ha_log_time | ha_log_local); } set_working_set_defaults(&data_set); if (process) { if (show_scores && show_utilization) { fprintf(stdout, "Allocation scores and utilization information:\n"); } else if (show_scores) { fprintf(stdout, "Allocation scores:\n"); } else if (show_utilization) { fprintf(stdout, "Utilization information:\n"); } do_calculations(&data_set, cib_object, a_date); } msg_buffer = dump_xml_formatted(data_set.graph); if (safe_str_eq(graph_file, "-")) { fprintf(stdout, "%s\n", msg_buffer); fflush(stdout); } else if (graph_file != NULL) { FILE *graph_strm = fopen(graph_file, "w"); if (graph_strm == NULL) { crm_perror(LOG_ERR, "Could not open %s for writing", graph_file); } else { if (fprintf(graph_strm, "%s\n\n", msg_buffer) < 0) { crm_perror(LOG_ERR, "Write to %s failed", graph_file); } fflush(graph_strm); fclose(graph_strm); } } crm_free(msg_buffer); if (dot_file != NULL) { dot_strm = fopen(dot_file, "w"); if (dot_strm == NULL) { crm_perror(LOG_ERR, "Could not open %s for writing", dot_file); } } if (dot_strm == NULL) { goto simulate; } init_dotfile(); for (lpc = data_set.actions; lpc != NULL; lpc = lpc->next) { action_t *action = (action_t *) lpc->data; const char *style = "filled"; const char *font = "black"; const char *color = "black"; const char *fill = NULL; char *action_name = create_action_name(action); crm_trace("Action %d: %p", action->id, action); if (is_set(action->flags, pe_action_pseudo)) { font = "orange"; } style = "dashed"; if (is_set(action->flags, pe_action_dumped)) { style = "bold"; color = "green"; } else if (action->rsc != NULL && is_not_set(action->rsc->flags, pe_rsc_managed)) { color = "purple"; if (all_actions == FALSE) { goto dont_write; } } else if (is_set(action->flags, pe_action_optional)) { color = "blue"; if (all_actions == FALSE) { goto dont_write; } } else { color = "red"; CRM_CHECK(is_set(action->flags, pe_action_runnable) == FALSE,; ); } set_bit_inplace(action->flags, pe_action_dumped); dot_write("\"%s\" [ style=%s color=\"%s\" fontcolor=\"%s\" %s%s]", action_name, style, color, font, fill ? "fillcolor=" : "", fill ? fill : ""); dont_write: crm_free(action_name); }
int main(int argc, char ** argv) { int flag; int rc = 0; int quiet = 0; int verbose = 0; int argerr = 0; int timeout = 120; int option_index = 0; int fence_level = 0; char name[512]; char value[512]; const char *agent = NULL; const char *device = NULL; const char *target = NULL; char action = 0; stonith_t *st = NULL; stonith_key_value_t *params = NULL; stonith_key_value_t *devices = NULL; stonith_key_value_t *dIter = NULL; crm_log_init(NULL, LOG_NOTICE, FALSE, FALSE, argc, argv, FALSE); crm_set_options(NULL, "mode [options]", long_options, "Provides access to the stonith-ng API.\n" "\nAllows the administrator to add/remove/list devices, check device and host status and fence hosts\n"); while (1) { flag = crm_get_option(argc, argv, &option_index); if (flag == -1) break; switch(flag) { case 'V': verbose = 1; crm_bump_log_level(); break; case '$': case '?': crm_help(flag, EX_OK); break; case 'L': case 'I': action = flag; break; case 'q': quiet = 1; break; case 'Q': case 'R': case 'D': action = flag; device = optarg; break; case 'a': agent = optarg; break; case 'l': target = optarg; action = 'L'; break; case 'M': action = flag; break; case 't': timeout = crm_atoi(optarg, NULL); break; case 'B': case 'F': case 'U': case 'C': case 'H': case 'r': case 'd': target = optarg; action = flag; break; case 'i': fence_level = crm_atoi(optarg, NULL); break; case 'v': devices = stonith_key_value_add(devices, NULL, optarg); break; case 'o': crm_info("Scanning: -o %s", optarg); rc = sscanf(optarg, "%[^=]=%[^=]", name, value); if(rc != 2) { crm_err("Invalid option: -o %s", optarg); ++argerr; } else { crm_info("Got: '%s'='%s'", name, value); stonith_key_value_add(params, name, value); } break; case 'e': { char *key = crm_concat("OCF_RESKEY", optarg, '_'); const char *env = getenv(key); if(env == NULL) { crm_err("Invalid option: -e %s", optarg); ++argerr; } else { crm_info("Got: '%s'='%s'", optarg, env); stonith_key_value_add( params, optarg, env); } } break; default: ++argerr; break; } } if (optind > argc) { ++argerr; } if (argerr) { crm_help('?', EX_USAGE); } crm_debug("Create"); st = stonith_api_new(); if(action != 'M' && action != 'I') { rc = st->cmds->connect(st, crm_system_name, NULL); crm_debug("Connect: %d", rc); if(rc < 0) { goto done; } } switch(action) { case 'I': rc = st->cmds->list(st, st_opt_sync_call, NULL, &devices, timeout); for(dIter = devices; dIter; dIter = dIter->next ) { fprintf( stdout, " %s\n", dIter->value ); } if(rc == 0) { fprintf(stderr, "No devices found\n"); } else if(rc > 0) { fprintf(stderr, "%d devices found\n", rc); rc = 0; } stonith_key_value_freeall(devices, 1, 1); break; case 'L': rc = st->cmds->query(st, st_opts, target, &devices, timeout); for(dIter = devices; dIter; dIter = dIter->next ) { fprintf( stdout, " %s\n", dIter->value ); } if(rc == 0) { fprintf(stderr, "No devices found\n"); } else if(rc > 0) { fprintf(stderr, "%d devices found\n", rc); rc = 0; } stonith_key_value_freeall(devices, 1, 1); break; case 'Q': rc = st->cmds->call(st, st_opts, device, "monitor", NULL, timeout); if(rc < 0) { rc = st->cmds->call(st, st_opts, device, "list", NULL, timeout); } break; case 'R': rc = st->cmds->register_device(st, st_opts, device, "stonith-ng", agent, params); break; case 'D': rc = st->cmds->remove_device(st, st_opts, device); break; case 'r': rc = st->cmds->register_level(st, st_opts, target, fence_level, devices); break; case 'd': rc = st->cmds->remove_level(st, st_opts, target, fence_level); break; case 'M': if (agent == NULL) { printf("Please specify an agent to query using -a,--agent [value]\n"); return -1; } else { char *buffer = NULL; rc = st->cmds->metadata(st, st_opt_sync_call, agent, NULL, &buffer, timeout); if(rc == pcmk_ok) { printf("%s\n", buffer); } free(buffer); } break; case 'C': rc = st->cmds->confirm(st, st_opts, target); break; case 'B': rc = st->cmds->fence(st, st_opts, target, "reboot", timeout); break; case 'F': rc = st->cmds->fence(st, st_opts, target, "off", timeout); break; case 'U': rc = st->cmds->fence(st, st_opts, target, "on", timeout); break; case 'H': { stonith_history_t *history, *hp, *latest = NULL; rc = st->cmds->history(st, st_opts, target, &history, timeout); for(hp = history; hp; hp = hp->next) { char *action_s = NULL; time_t complete = hp->completed; if(hp->state == st_done) { latest = hp; } if(quiet || !verbose) { continue; } else if(hp->action == NULL) { action_s = strdup("unknown"); } else if(hp->action[0] != 'r') { action_s = crm_concat("turn", hp->action, ' '); } else { action_s = strdup(hp->action); } if(hp->state == st_failed) { printf("%s failed to %s node %s on behalf of %s at %s\n", hp->delegate?hp->delegate:"We", action_s, hp->target, hp->origin, ctime(&complete)); } else if(hp->state == st_done && hp->delegate) { printf("%s was able to %s node %s on behalf of %s at %s\n", hp->delegate, action_s, hp->target, hp->origin, ctime(&complete)); } else if(hp->state == st_done) { printf("We were able to %s node %s on behalf of %s at %s\n", action_s, hp->target, hp->origin, ctime(&complete)); } else { printf("%s wishes to %s node %s - %d %d\n", hp->origin, action_s, hp->target, hp->state, hp->completed); } free(action_s); } if(latest) { if(quiet) { printf("%d\n", latest->completed); } else { char *action_s = NULL; time_t complete = latest->completed; if(latest->action == NULL) { action_s = strdup("unknown"); } else if(latest->action[0] != 'r') { action_s = crm_concat("turn", latest->action, ' '); } else { action_s = strdup(latest->action); } printf("%s was able to %s node %s on behalf of %s at %s\n", latest->delegate?latest->delegate:"We", action_s, latest->target, latest->origin, ctime(&complete)); free(action_s); } } } break; } done: if(rc < 0) { printf("Command failed: %s\n", pcmk_strerror(rc)); } stonith_key_value_freeall(params, 1, 1); st->cmds->disconnect(st); crm_debug("Disconnect: %d", rc); crm_debug("Destroy"); stonith_api_delete(st); return rc; }
void attrd_peer_message(crm_node_t *peer, xmlNode *xml) { int peer_state = 0; const char *v = crm_element_value(xml, F_ATTRD_VERSION); const char *op = crm_element_value(xml, F_ATTRD_TASK); const char *election_op = crm_element_value(xml, F_CRM_TASK); const char *host = crm_element_value(xml, F_ATTRD_HOST); if(election_op) { enum election_result rc = 0; crm_xml_add(xml, F_CRM_HOST_FROM, peer->uname); rc = election_count_vote(writer, xml, TRUE); switch(rc) { case election_start: free(peer_writer); peer_writer = NULL; election_vote(writer); break; case election_lost: free(peer_writer); peer_writer = strdup(peer->uname); break; default: election_check(writer); break; } return; } else if(v == NULL) { /* From the non-atomic version */ if (safe_str_eq(op, ATTRD_OP_UPDATE)) { const char *name = crm_element_value(xml, F_ATTRD_ATTRIBUTE); crm_trace("Compatibility update of %s from %s", name, peer->uname); attrd_peer_update(peer, xml, host, FALSE); } else if (safe_str_eq(op, ATTRD_OP_FLUSH)) { const char *name = crm_element_value(xml, F_ATTRD_ATTRIBUTE); attribute_t *a = g_hash_table_lookup(attributes, name); if(a) { crm_trace("Compatibility write-out of %s for %s from %s", a->id, op, peer->uname); write_or_elect_attribute(a); } } else if (safe_str_eq(op, ATTRD_OP_REFRESH)) { GHashTableIter aIter; attribute_t *a = NULL; g_hash_table_iter_init(&aIter, attributes); while (g_hash_table_iter_next(&aIter, NULL, (gpointer *) & a)) { crm_trace("Compatibility write-out of %s for %s from %s", a->id, op, peer->uname); write_or_elect_attribute(a); } } } crm_element_value_int(xml, F_ATTRD_WRITER, &peer_state); if(election_state(writer) == election_won && peer_state == election_won && safe_str_neq(peer->uname, attrd_cluster->uname)) { crm_notice("Detected another attribute writer: %s", peer->uname); election_vote(writer); } else if(peer_state == election_won) { if(peer_writer == NULL) { peer_writer = strdup(peer->uname); crm_notice("Recorded attribute writer: %s", peer->uname); } else if(safe_str_neq(peer->uname, peer_writer)) { crm_notice("Recorded new attribute writer: %s (was %s)", peer->uname, peer_writer); free(peer_writer); peer_writer = strdup(peer->uname); } } if (safe_str_eq(op, ATTRD_OP_UPDATE) || safe_str_eq(op, ATTRD_OP_UPDATE_BOTH) || safe_str_eq(op, ATTRD_OP_UPDATE_DELAY)) { attrd_peer_update(peer, xml, host, FALSE); } else if (safe_str_eq(op, ATTRD_OP_SYNC)) { attrd_peer_sync(peer, xml); } else if (safe_str_eq(op, ATTRD_OP_PEER_REMOVE)) { attrd_peer_remove(host, TRUE, peer->uname); } else if (safe_str_eq(op, ATTRD_OP_SYNC_RESPONSE) && safe_str_neq(peer->uname, attrd_cluster->uname)) { xmlNode *child = NULL; crm_info("Processing %s from %s", op, peer->uname); for (child = __xml_first_child(xml); child != NULL; child = __xml_next(child)) { host = crm_element_value(child, F_ATTRD_HOST); attrd_peer_update(peer, child, host, TRUE); } } }
/* Exit code means? */ static int32_t st_ipc_dispatch(qb_ipcs_connection_t * qbc, void *data, size_t size) { uint32_t id = 0; uint32_t flags = 0; int call_options = 0; xmlNode *request = NULL; crm_client_t *c = crm_client_get(qbc); const char *op = NULL; if (c == NULL) { crm_info("Invalid client: %p", qbc); return 0; } request = crm_ipcs_recv(c, data, size, &id, &flags); if (request == NULL) { crm_ipcs_send_ack(c, id, flags, "nack", __FUNCTION__, __LINE__); return 0; } op = crm_element_value(request, F_CRM_TASK); if(safe_str_eq(op, CRM_OP_RM_NODE_CACHE)) { crm_xml_add(request, F_TYPE, T_STONITH_NG); crm_xml_add(request, F_STONITH_OPERATION, op); crm_xml_add(request, F_STONITH_CLIENTID, c->id); crm_xml_add(request, F_STONITH_CLIENTNAME, crm_client_name(c)); crm_xml_add(request, F_STONITH_CLIENTNODE, stonith_our_uname); send_cluster_message(NULL, crm_msg_stonith_ng, request, FALSE); free_xml(request); return 0; } if (c->name == NULL) { const char *value = crm_element_value(request, F_STONITH_CLIENTNAME); if (value == NULL) { value = "unknown"; } c->name = crm_strdup_printf("%s.%u", value, c->pid); } crm_element_value_int(request, F_STONITH_CALLOPTS, &call_options); crm_trace("Flags %u/%u for command %u from %s", flags, call_options, id, crm_client_name(c)); if (is_set(call_options, st_opt_sync_call)) { CRM_ASSERT(flags & crm_ipc_client_response); CRM_LOG_ASSERT(c->request_id == 0); /* This means the client has two synchronous events in-flight */ c->request_id = id; /* Reply only to the last one */ } crm_xml_add(request, F_STONITH_CLIENTID, c->id); crm_xml_add(request, F_STONITH_CLIENTNAME, crm_client_name(c)); crm_xml_add(request, F_STONITH_CLIENTNODE, stonith_our_uname); crm_log_xml_trace(request, "Client[inbound]"); stonith_command(c, id, flags, request, NULL); free_xml(request); return 0; }
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; } } } }
/* 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; }
/* * Useful Examples: * http://www.kernel.org/pub/linux/libs/pam/Linux-PAM-html * http://developer.apple.com/samplecode/CryptNoMore/index.html */ static int construct_pam_passwd(int num_msg, const struct pam_message **msg, struct pam_response **response, void *data) { int count = 0; struct pam_response *reply; char *string = (char *)data; CRM_CHECK(data, return PAM_CONV_ERR); CRM_CHECK(num_msg == 1, return PAM_CONV_ERR); /* We only want to handle one message */ reply = calloc(1, sizeof(struct pam_response)); CRM_ASSERT(reply != NULL); for (count = 0; count < num_msg; ++count) { switch (msg[count]->msg_style) { case PAM_TEXT_INFO: crm_info("PAM: %s\n", msg[count]->msg); break; case PAM_PROMPT_ECHO_OFF: case PAM_PROMPT_ECHO_ON: reply[count].resp_retcode = 0; reply[count].resp = string; /* We already made a copy */ case PAM_ERROR_MSG: /* In theory we'd want to print this, but then * we see the password prompt in the logs */ /* crm_err("PAM error: %s\n", msg[count]->msg); */ break; default: crm_err("Unhandled conversation type: %d", msg[count]->msg_style); goto bail; } } *response = reply; reply = NULL; return PAM_SUCCESS; bail: for (count = 0; count < num_msg; ++count) { if (reply[count].resp != NULL) { switch (msg[count]->msg_style) { case PAM_PROMPT_ECHO_ON: case PAM_PROMPT_ECHO_OFF: /* Erase the data - it contained a password */ while (*(reply[count].resp)) { *(reply[count].resp)++ = '\0'; } free(reply[count].resp); break; } reply[count].resp = NULL; } } free(reply); reply = NULL; return PAM_CONV_ERR; }
static void tengine_stonith_notify(stonith_t * st, stonith_event_t * st_event) { if(te_client_id == NULL) { te_client_id = g_strdup_printf("%s.%d", crm_system_name, getpid()); } if (st_event == NULL) { crm_err("Notify data not found"); return; } if (st_event->result == pcmk_ok && safe_str_eq("on", st_event->action)) { crm_notice("%s was successfully unfenced by %s (at the request of %s)", st_event->target, st_event->executioner ? st_event->executioner : "<anyone>", st_event->origin); /* TODO: Hook up st_event->device */ return; } else if (safe_str_eq("on", st_event->action)) { crm_err("Unfencing of %s by %s failed: %s (%d)", st_event->target, st_event->executioner ? st_event->executioner : "<anyone>", pcmk_strerror(st_event->result), st_event->result); return; } else if (st_event->result == pcmk_ok && crm_str_eq(st_event->target, fsa_our_uname, TRUE)) { crm_crit("We were allegedly just fenced by %s for %s!", st_event->executioner ? st_event->executioner : "<anyone>", st_event->origin); /* Dumps blackbox if enabled */ qb_log_fini(); /* Try to get the above log message to disk - somehow */ /* Get out ASAP and do not come back up. * * Triggering a reboot is also not the worst idea either since * the rest of the cluster thinks we're safely down */ #ifdef RB_HALT_SYSTEM reboot(RB_HALT_SYSTEM); #endif /* * If reboot() fails or is not supported, coming back up will * probably lead to a situation where the other nodes set our * status to 'lost' because of the fencing callback and will * discard subsequent election votes with: * * Election 87 (current: 5171, owner: 103): Processed vote from east-03 (Peer is not part of our cluster) * * So just stay dead, something is seriously messed up anyway. * */ exit(100); /* None of our wrappers since we already called qb_log_fini() */ return; } if (st_event->result == pcmk_ok && safe_str_eq(st_event->operation, T_STONITH_NOTIFY_FENCE)) { st_fail_count_reset(st_event->target); } crm_notice("Peer %s was%s terminated (%s) by %s for %s: %s (ref=%s) by client %s", st_event->target, st_event->result == pcmk_ok ? "" : " not", st_event->action, st_event->executioner ? st_event->executioner : "<anyone>", st_event->origin, pcmk_strerror(st_event->result), st_event->id, st_event->client_origin ? st_event->client_origin : "<unknown>"); #if SUPPORT_CMAN if (st_event->result == pcmk_ok && is_cman_cluster()) { int local_rc = 0; int confirm = 0; char *target_copy = strdup(st_event->target); /* In case fenced hasn't noticed yet * * Any fencing that has been inititated will be completed by way of the fence_pcmk redirect */ local_rc = fenced_external(target_copy); if (local_rc != 0) { crm_err("Could not notify CMAN that '%s' is now fenced: %d", st_event->target, local_rc); } else { crm_notice("Notified CMAN that '%s' is now fenced", st_event->target); } /* In case fenced is already trying to shoot it */ confirm = open("/var/run/cluster/fenced_override", O_NONBLOCK|O_WRONLY); if (confirm > 0) { int ignore = 0; int len = strlen(target_copy); errno = 0; local_rc = write(confirm, target_copy, len); ignore = write(confirm, "\n", 1); if(ignore < 0 && errno == EBADF) { crm_trace("CMAN not expecting %s to be fenced (yet)", st_event->target); } else if (local_rc < len) { crm_perror(LOG_ERR, "Confirmation of CMAN fencing event for '%s' failed: %d", st_event->target, local_rc); } else { fsync(confirm); crm_notice("Confirmed CMAN fencing event for '%s'", st_event->target); } close(confirm); } free(target_copy); } #endif if (st_event->result == pcmk_ok) { crm_node_t *peer = crm_find_peer_full(0, st_event->target, CRM_GET_PEER_REMOTE | CRM_GET_PEER_CLUSTER); const char *uuid = NULL; gboolean we_are_executioner = safe_str_eq(st_event->executioner, fsa_our_uname); if (peer == NULL) { return; } uuid = crm_peer_uuid(peer); crm_trace("target=%s dc=%s", st_event->target, fsa_our_dc); if(AM_I_DC) { /* The DC always sends updates */ send_stonith_update(NULL, st_event->target, uuid); if (st_event->client_origin && safe_str_neq(st_event->client_origin, te_client_id)) { /* Abort the current transition graph if it wasn't us * that invoked stonith to fence someone */ crm_info("External fencing operation from %s fenced %s", st_event->client_origin, st_event->target); abort_transition(INFINITY, tg_restart, "External Fencing Operation", NULL); } /* Assume it was our leader if we dont currently have one */ } else if (fsa_our_dc == NULL || safe_str_eq(fsa_our_dc, st_event->target)) { crm_notice("Target %s our leader %s (recorded: %s)", fsa_our_dc ? "was" : "may have been", st_event->target, fsa_our_dc ? fsa_our_dc : "<unset>"); /* Given the CIB resyncing that occurs around elections, * have one node update the CIB now and, if the new DC is different, * have them do so too after the election */ if (we_are_executioner) { send_stonith_update(NULL, st_event->target, uuid); } stonith_cleanup_list = g_list_append(stonith_cleanup_list, strdup(st_event->target)); } crmd_peer_down(peer, TRUE); } }
/* coverity[-alloc] Memory is referenced in one or both hashtables */ crm_node_t * crm_get_peer(unsigned int id, const char *uname) { GHashTableIter iter; crm_node_t *node = NULL; crm_node_t *by_id = NULL; crm_node_t *by_name = NULL; CRM_ASSERT(id > 0 || uname != NULL); crm_peer_init(); if (uname != NULL) { g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { if(node->uname && strcasecmp(node->uname, uname) == 0) { crm_trace("Name match: %s = %p", node->uname, node); by_name = node; break; } } } if (id > 0) { g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { if(node->id == id) { crm_trace("ID match: %u = %p", node->id, node); by_id = node; break; } } } node = by_id; /* Good default */ if(by_id == by_name) { /* Nothing to do if they match (both NULL counts) */ crm_trace("Consistent: %p for %u/%s", by_id, id, uname); } else if(by_id == NULL && by_name) { crm_trace("Only one: %p for %u/%s", by_name, id, uname); if(id && by_name->id) { crm_dump_peer_hash(LOG_WARNING, __FUNCTION__); crm_crit("Node %u and %u share the same name '%s'", id, by_name->id, uname); node = NULL; /* Create a new one */ } else { node = by_name; } } else if(by_name == NULL && by_id) { crm_trace("Only one: %p for %u/%s", by_id, id, uname); if(uname && by_id->uname) { crm_dump_peer_hash(LOG_WARNING, __FUNCTION__); crm_crit("Node '%s' and '%s' share the same cluster nodeid %u: assuming '%s' is correct", uname, by_id->uname, id, uname); } } else if(uname && by_id->uname) { crm_warn("Node '%s' and '%s' share the same cluster nodeid: %u", by_id->uname, by_name->uname, id); } else if(id && by_name->id) { crm_warn("Node %u and %u share the same name: '%s'", by_id->id, by_name->id, uname); } else { /* Simple merge */ /* Only corosync based clusters use nodeid's * * The functions that call crm_update_peer_state() only know nodeid * so 'by_id' is authorative when merging * * Same for crm_update_peer_proc() */ crm_dump_peer_hash(LOG_DEBUG, __FUNCTION__); crm_info("Merging %p into %p", by_name, by_id); g_hash_table_foreach_remove(crm_peer_cache, crm_hash_find_by_data, by_name); } if (node == NULL) { char *uniqueid = crm_generate_uuid(); node = calloc(1, sizeof(crm_node_t)); CRM_ASSERT(node); crm_info("Created entry %s/%p for node %s/%u (%d total)", uniqueid, node, uname, id, 1 + g_hash_table_size(crm_peer_cache)); g_hash_table_replace(crm_peer_cache, uniqueid, node); } if(id > 0 && uname && (node->id == 0 || node->uname == NULL)) { crm_info("Node %u is now known as %s", id, uname); } if(id > 0 && node->id == 0) { node->id = id; } if(uname && node->uname == NULL) { int lpc, len = strlen(uname); for (lpc = 0; lpc < len; lpc++) { if (uname[lpc] >= 'A' && uname[lpc] <= 'Z') { crm_warn("Node names with capitals are discouraged, consider changing '%s' to something else", uname); break; } } node->uname = strdup(uname); if (crm_status_callback) { crm_status_callback(crm_status_uname, node, NULL); } } if(node->uuid == NULL) { const char *uuid = crm_peer_uuid(node); if (uuid) { crm_info("Node %u has uuid %s", id, uuid); } else { crm_info("Cannot obtain a UUID for node %d/%s", id, node->uname); } } return node; }
void abort_transition_graph(int abort_priority, enum transition_action abort_action, const char *abort_text, xmlNode * reason, const char *fn, int line) { int add[] = { 0, 0, 0 }; int del[] = { 0, 0, 0 }; int level = LOG_INFO; xmlNode *diff = NULL; xmlNode *change = NULL; CRM_CHECK(transition_graph != NULL, return); switch (fsa_state) { case S_STARTING: case S_PENDING: case S_NOT_DC: case S_HALT: case S_ILLEGAL: case S_STOPPING: case S_TERMINATE: crm_info("Abort %s suppressed: state=%s (complete=%d)", abort_text, fsa_state2string(fsa_state), transition_graph->complete); return; default: break; } /* Make sure any queued calculations are discarded ASAP */ free(fsa_pe_ref); fsa_pe_ref = NULL; if (transition_graph->complete == FALSE) { if(update_abort_priority(transition_graph, abort_priority, abort_action, abort_text)) { level = LOG_NOTICE; } } if(reason) { xmlNode *search = NULL; for(search = reason; search; search = search->parent) { if (safe_str_eq(XML_TAG_DIFF, TYPE(search))) { diff = search; break; } } if(diff) { xml_patch_versions(diff, add, del); for(search = reason; search; search = search->parent) { if (safe_str_eq(XML_DIFF_CHANGE, TYPE(search))) { change = search; break; } } } } if(reason == NULL) { do_crm_log(level, "Transition aborted: %s (source=%s:%d, %d)", abort_text, fn, line, transition_graph->complete); } else if(change == NULL) { char *local_path = xml_get_path(reason); do_crm_log(level, "Transition aborted by %s.%s: %s (cib=%d.%d.%d, source=%s:%d, path=%s, %d)", TYPE(reason), ID(reason), abort_text, add[0], add[1], add[2], fn, line, local_path, transition_graph->complete); free(local_path); } else { const char *kind = NULL; const char *op = crm_element_value(change, XML_DIFF_OP); const char *path = crm_element_value(change, XML_DIFF_PATH); if(change == reason) { if(strcmp(op, "create") == 0) { reason = reason->children; } else if(strcmp(op, "modify") == 0) { reason = first_named_child(reason, XML_DIFF_RESULT); if(reason) { reason = reason->children; } } } kind = TYPE(reason); if(strcmp(op, "delete") == 0) { const char *shortpath = strrchr(path, '/'); do_crm_log(level, "Transition aborted by deletion of %s: %s (cib=%d.%d.%d, source=%s:%d, path=%s, %d)", shortpath?shortpath+1:path, abort_text, add[0], add[1], add[2], fn, line, path, transition_graph->complete); } else if (safe_str_eq(XML_CIB_TAG_NVPAIR, kind)) { do_crm_log(level, "Transition aborted by %s, %s=%s: %s (%s cib=%d.%d.%d, source=%s:%d, path=%s, %d)", crm_element_value(reason, XML_ATTR_ID), crm_element_value(reason, XML_NVPAIR_ATTR_NAME), crm_element_value(reason, XML_NVPAIR_ATTR_VALUE), abort_text, op, add[0], add[1], add[2], fn, line, path, transition_graph->complete); } else if (safe_str_eq(XML_LRM_TAG_RSC_OP, kind)) { const char *magic = crm_element_value(reason, XML_ATTR_TRANSITION_MAGIC); do_crm_log(level, "Transition aborted by %s '%s' on %s: %s (magic=%s, cib=%d.%d.%d, source=%s:%d, %d)", crm_element_value(reason, XML_LRM_ATTR_TASK_KEY), op, crm_element_value(reason, XML_LRM_ATTR_TARGET), abort_text, magic, add[0], add[1], add[2], fn, line, transition_graph->complete); } else if (safe_str_eq(XML_CIB_TAG_STATE, kind) || safe_str_eq(XML_CIB_TAG_NODE, kind)) { const char *uname = crm_peer_uname(ID(reason)); do_crm_log(level, "Transition aborted by %s '%s' on %s: %s (cib=%d.%d.%d, source=%s:%d, %d)", kind, op, uname ? uname : ID(reason), abort_text, add[0], add[1], add[2], fn, line, transition_graph->complete); } else { do_crm_log(level, "Transition aborted by %s.%s '%s': %s (cib=%d.%d.%d, source=%s:%d, path=%s, %d)", TYPE(reason), ID(reason), op?op:"change", abort_text, add[0], add[1], add[2], fn, line, path, transition_graph->complete); } } if (transition_graph->complete) { if (transition_timer->period_ms > 0) { crm_timer_stop(transition_timer); crm_timer_start(transition_timer); } else { register_fsa_input(C_FSA_INTERNAL, I_PE_CALC, NULL); } return; } mainloop_set_trigger(transition_trigger); }
/* * returns the ID of the action if a match is found * returns -1 if a match was not found * returns -2 if a match was found but the action failed (and was * not allowed to) */ int match_graph_event(int action_id, xmlNode * event, const char *event_node, int op_status, int op_rc, int target_rc) { const char *target = NULL; const char *allow_fail = NULL; const char *this_event = NULL; crm_action_t *action = NULL; action = get_action(action_id, FALSE); if (action == NULL) { return -1; } op_status = status_from_rc(action, op_status, op_rc, target_rc); if (op_status != PCMK_LRM_OP_DONE) { update_failcount(event, event_node, op_rc, target_rc, FALSE); } /* Process OP status */ switch (op_status) { case PCMK_LRM_OP_PENDING: crm_debug("Ignoring pending operation"); return action->id; break; case PCMK_LRM_OP_DONE: break; case PCMK_LRM_OP_ERROR: case PCMK_LRM_OP_TIMEOUT: case PCMK_LRM_OP_NOTSUPPORTED: action->failed = TRUE; break; case PCMK_LRM_OP_CANCELLED: /* do nothing?? */ crm_err("Dont know what to do for cancelled ops yet"); break; default: action->failed = TRUE; crm_err("Unsupported action result: %d", op_status); } /* stop this event's timer if it had one */ stop_te_timer(action->timer); action->confirmed = TRUE; update_graph(transition_graph, action); trigger_graph(); if (action->failed) { allow_fail = crm_meta_value(action->params, XML_ATTR_TE_ALLOWFAIL); if (crm_is_true(allow_fail)) { action->failed = FALSE; } } if (action->failed) { abort_transition(action->synapse->priority + 1, tg_restart, "Event failed", event); } this_event = crm_element_value(event, XML_LRM_ATTR_TASK_KEY); target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); crm_info("Action %s (%d) confirmed on %s (rc=%d)", crm_str(this_event), action->id, crm_str(target), op_status); /* determine if this action affects a remote-node's online/offline status */ process_remote_node_action(action, event); return action->id; }
static gboolean te_crm_command(crm_graph_t * graph, crm_action_t * action) { char *counter = NULL; xmlNode *cmd = NULL; gboolean is_local = FALSE; const char *id = NULL; const char *task = NULL; const char *value = NULL; const char *on_node = NULL; const char *router_node = NULL; gboolean rc = TRUE; gboolean no_wait = FALSE; id = ID(action->xml); task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); on_node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); router_node = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if (!router_node) { router_node = on_node; } CRM_CHECK(on_node != NULL && strlen(on_node) != 0, crm_err("Corrupted command (id=%s) %s: no node", crm_str(id), crm_str(task)); return FALSE); crm_info("Executing crm-event (%s): %s on %s%s%s", crm_str(id), crm_str(task), on_node, is_local ? " (local)" : "", no_wait ? " - no waiting" : ""); if (safe_str_eq(router_node, fsa_our_uname)) { is_local = TRUE; } value = crm_meta_value(action->params, XML_ATTR_TE_NOWAIT); if (crm_is_true(value)) { no_wait = TRUE; } if (is_local && safe_str_eq(task, CRM_OP_SHUTDOWN)) { /* defer until everything else completes */ crm_info("crm-event (%s) is a local shutdown", crm_str(id)); graph->completion_action = tg_shutdown; graph->abort_reason = "local shutdown"; action->confirmed = TRUE; update_graph(graph, action); trigger_graph(); return TRUE; } else if (safe_str_eq(task, CRM_OP_SHUTDOWN)) { crm_node_t *peer = crm_get_peer(0, router_node); crm_update_peer_expected(__FUNCTION__, peer, CRMD_JOINSTATE_DOWN); } cmd = create_request(task, action->xml, router_node, CRM_SYSTEM_CRMD, CRM_SYSTEM_TENGINE, NULL); counter = generate_transition_key(transition_graph->id, action->id, get_target_rc(action), te_uuid); crm_xml_add(cmd, XML_ATTR_TRANSITION_KEY, counter); rc = send_cluster_message(crm_get_peer(0, router_node), crm_msg_crmd, cmd, TRUE); free(counter); free_xml(cmd); if (rc == FALSE) { crm_err("Action %d failed: send", action->id); return FALSE; } else if (no_wait) { action->confirmed = TRUE; update_graph(graph, action); trigger_graph(); } else { if (action->timeout <= 0) { crm_err("Action %d: %s on %s had an invalid timeout (%dms). Using %dms instead", action->id, task, on_node, action->timeout, graph->network_delay); action->timeout = graph->network_delay; } te_start_action_timer(graph, action); } return TRUE; }
int main(int argc, char **argv) { msgtype type = -1; char *request = NULL; gboolean request_only = FALSE; char *reqid = NULL; int loglevel = LOG_INFO; gboolean verbose = FALSE; gboolean quiet = FALSE; int argerr = 0, flag; #ifdef HAVE_GETOPT_H int opt_idx = 0; static struct option long_opts[] = { {"type", 1, 0, 't'}, {"request", 1, 0, 'r'}, {"request-only", 1, 0, 'R'}, {"reqid", 1, 0, 'i'}, {"verbose", 0, 0, 'V'}, {"quiet", 0, 0, 'q'}, {"help", 0, 0, '?'}, {0, 0, 0, 0} }; #endif int sockfd, rc, ret = 0; vm_message msg; while (1) { #ifdef HAVE_GETOPT_H flag = getopt_long(argc, argv, OPTARGS, long_opts, &opt_idx); #else flag = getopt(argc, argv, OPTARGS); #endif if (flag == -1) break; switch (flag) { case 't': if (safe_str_eq(optarg, "monitor")) type = T_MOD_MONITOR; else if (safe_str_eq(optarg, "stonith")) type = T_MOD_STONITH; break; case 'r': request = crm_strdup(optarg); break; case 'R': request_only = TRUE; request = crm_strdup(optarg); break; case 'i': reqid = crm_strdup(optarg); break; case 'V': loglevel++; verbose = TRUE; break; case 'q': quiet = TRUE; break; case '?': usage(basename(argv[0]), LSB_EXIT_GENERIC); break; default: fprintf(stderr, "Argument code 0%o (%c) is not (?yet?) supported", flag, flag); argerr++; break; } } if (!quiet) crm_log_init(basename(argv[0]), loglevel, TRUE, verbose, argc, argv); else crm_log_init(basename(argv[0]), loglevel, TRUE, verbose, 0, NULL); if (optind < argc) { fprintf(stderr, "non-option ARGV-elements: "); while (optind < argc) fprintf(stderr, "%s", argv[optind++]); fprintf(stderr, "\n"); argerr++; } if (argerr || type == -1 || !request) usage(crm_system_name, LSB_EXIT_GENERIC); on_host = FALSE; if ((sockfd = connect_to(SOCK_PATH, 0)) < 0) return 1; if (send_message(sockfd, type, reqid, request) < 0) { ret = 1; goto end; } if (request_only == TRUE) goto end; while (1) { rc = receive_msg(sockfd, &msg); if (rc == 0) { if (reqid && safe_str_neq(reqid, msg.info.id)) { crm_info("request ID (%s) is unmatch : %s", reqid, msg.info.id); continue; } else { fprintf(stdout, "%s\n", msg.data); goto end; } } else if (rc == 2) { /* migration occurred.. */ ret = 2; goto end; } ret = 1; goto end; } end: close(sockfd); return ret; }
static gboolean te_rsc_command(crm_graph_t * graph, crm_action_t * action) { /* never overwrite stop actions in the CIB with * anything other than completed results * * Writing pending stops makes it look like the * resource is running again */ xmlNode *cmd = NULL; xmlNode *rsc_op = NULL; gboolean rc = TRUE; gboolean no_wait = FALSE; gboolean is_local = FALSE; char *counter = NULL; const char *task = NULL; const char *value = NULL; const char *on_node = NULL; const char *router_node = NULL; const char *task_uuid = NULL; CRM_ASSERT(action != NULL); CRM_ASSERT(action->xml != NULL); action->executed = FALSE; on_node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); CRM_CHECK(on_node != NULL && strlen(on_node) != 0, crm_err("Corrupted command(id=%s) %s: no node", ID(action->xml), crm_str(task)); return FALSE); rsc_op = action->xml; task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); task_uuid = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY); router_node = crm_element_value(rsc_op, XML_LRM_ATTR_ROUTER_NODE); if (!router_node) { router_node = on_node; } counter = generate_transition_key(transition_graph->id, action->id, get_target_rc(action), te_uuid); crm_xml_add(rsc_op, XML_ATTR_TRANSITION_KEY, counter); if (safe_str_eq(router_node, fsa_our_uname)) { is_local = TRUE; } value = crm_meta_value(action->params, XML_ATTR_TE_NOWAIT); if (crm_is_true(value)) { no_wait = TRUE; } crm_notice("Initiating action %d: %s %s on %s%s%s", action->id, task, task_uuid, on_node, is_local ? " (local)" : "", no_wait ? " - no waiting" : ""); cmd = create_request(CRM_OP_INVOKE_LRM, rsc_op, router_node, CRM_SYSTEM_LRMD, CRM_SYSTEM_TENGINE, NULL); if (is_local) { /* shortcut local resource commands */ ha_msg_input_t data = { .msg = cmd, .xml = rsc_op, }; fsa_data_t msg = { .id = 0, .data = &data, .data_type = fsa_dt_ha_msg, .fsa_input = I_NULL, .fsa_cause = C_FSA_INTERNAL, .actions = A_LRM_INVOKE, .origin = __FUNCTION__, }; do_lrm_invoke(A_LRM_INVOKE, C_FSA_INTERNAL, fsa_state, I_NULL, &msg); } else { rc = send_cluster_message(crm_get_peer(0, router_node), crm_msg_lrmd, cmd, TRUE); } free(counter); free_xml(cmd); action->executed = TRUE; if (rc == FALSE) { crm_err("Action %d failed: send", action->id); return FALSE; } else if (no_wait) { crm_info("Action %d confirmed - no wait", action->id); action->confirmed = TRUE; update_graph(transition_graph, action); trigger_graph(); } else { if (action->timeout <= 0) { crm_err("Action %d: %s %s on %s had an invalid timeout (%dms). Using %dms instead", action->id, task, task_uuid, on_node, action->timeout, graph->network_delay); action->timeout = graph->network_delay; } te_start_action_timer(graph, action); } value = crm_meta_value(action->params, XML_OP_ATTR_PENDING); if (crm_is_true(value) && safe_str_neq(task, CRMD_ACTION_CANCEL) && safe_str_neq(task, CRMD_ACTION_DELETE)) { /* write a "pending" entry to the CIB, inhibit notification */ crm_debug("Recording pending op %s in the CIB", task_uuid); cib_action_update(action, PCMK_LRM_OP_PENDING, PCMK_OCF_UNKNOWN); } return TRUE; }
/* A_DC_JOIN_OFFER_ONE */ void do_dc_join_offer_one(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) { crm_node_t *member; ha_msg_input_t *welcome = NULL; const char *op = NULL; const char *join_to = NULL; if(msg_data->data) { welcome = fsa_typed_data(fsa_dt_ha_msg); } else { crm_info("A new node joined - wait until it contacts us"); return; } if(welcome == NULL) { crm_err("Attempt to send welcome message " "without a message to reply to!"); return; } join_to = crm_element_value(welcome->msg, F_CRM_HOST_FROM); if(join_to == NULL) { crm_err("Attempt to send welcome message " "without a host to reply to!"); return; } member = crm_get_peer(0, join_to); if(member == NULL || crm_is_member_active(member) == FALSE) { crm_err("Attempt to send welcome message " "to a node not part of our partition!"); return; } op = crm_element_value(welcome->msg, F_CRM_TASK); if(join_to != NULL && (cur_state == S_INTEGRATION || cur_state == S_FINALIZE_JOIN)) { /* note: it _is_ possible that a node will have been * sick or starting up when the original offer was made. * however, it will either re-announce itself in due course * _or_ we can re-store the original offer on the client. */ crm_debug("(Re-)offering membership to %s...", join_to); } crm_info("join-%d: Processing %s request from %s in state %s", current_join_id, op, join_to, fsa_state2string(cur_state)); join_make_offer(NULL, member, NULL); /* always offer to the DC (ourselves) * this ensures the correct value for max_generation_from */ member = crm_get_peer(0, fsa_our_uname); join_make_offer(NULL, member, NULL); /* this was a genuine join request, cancel any existing * transition and invoke the PE */ start_transition(fsa_state); /* dont waste time by invoking the pe yet; */ crm_debug("Waiting on %d outstanding join acks for join-%d", g_hash_table_size(welcomed_nodes), current_join_id); }
/*! * \internal * \brief Respond to a client update request * * \param[in] xml Root of request XML * * \return void */ void attrd_client_update(xmlNode *xml) { attribute_t *a = NULL; attribute_value_t *v = NULL; char *key = crm_element_value_copy(xml, F_ATTRD_KEY); char *set = crm_element_value_copy(xml, F_ATTRD_SET); char *host = crm_element_value_copy(xml, F_ATTRD_HOST); const char *attr = crm_element_value(xml, F_ATTRD_ATTRIBUTE); const char *value = crm_element_value(xml, F_ATTRD_VALUE); const char *regex = crm_element_value(xml, F_ATTRD_REGEX); /* If a regex was specified, broadcast a message for each match */ if ((attr == NULL) && regex) { GHashTableIter aIter; regex_t *r_patt = calloc(1, sizeof(regex_t)); crm_debug("Setting %s to %s", regex, value); if (regcomp(r_patt, regex, REG_EXTENDED)) { crm_err("Bad regex '%s' for update", regex); } else { g_hash_table_iter_init(&aIter, attributes); while (g_hash_table_iter_next(&aIter, (gpointer *) & attr, NULL)) { int status = regexec(r_patt, attr, 0, NULL, 0); if (status == 0) { crm_trace("Matched %s with %s", attr, regex); crm_xml_add(xml, F_ATTRD_ATTRIBUTE, attr); send_attrd_message(NULL, xml); } } } free(key); free(set); free(host); regfree(r_patt); free(r_patt); return; } else if (attr == NULL) { crm_err("Update request did not specify attribute or regular expression"); return; } if (host == NULL) { crm_trace("Inferring host"); host = strdup(attrd_cluster->uname); crm_xml_add(xml, F_ATTRD_HOST, host); crm_xml_add_int(xml, F_ATTRD_HOST_ID, attrd_cluster->nodeid); } a = g_hash_table_lookup(attributes, attr); /* If value was specified using ++ or += notation, expand to real value */ if (value) { int offset = 1; int int_value = 0; static const int plus_plus_len = 5; if ((strlen(value) >= (plus_plus_len + 2)) && (value[plus_plus_len] == '+') && ((value[plus_plus_len + 1] == '+') || (value[plus_plus_len + 1] == '='))) { if (a) { v = g_hash_table_lookup(a->values, host); } if (v) { int_value = char2score(v->current); } if (value[plus_plus_len + 1] != '+') { const char *offset_s = value + (plus_plus_len + 2); offset = char2score(offset_s); } int_value += offset; if (int_value > INFINITY) { int_value = INFINITY; } crm_info("Expanded %s=%s to %d", attr, value, int_value); crm_xml_add_int(xml, F_ATTRD_VALUE, int_value); /* Replacing the value frees the previous memory, so re-query it */ value = crm_element_value(xml, F_ATTRD_VALUE); } } if ((peer_writer == NULL) && (election_state(writer) != election_in_progress)) { crm_info("Starting an election to determine the writer"); election_vote(writer); } crm_debug("Broadcasting %s[%s] = %s%s", attr, host, value, ((election_state(writer) == election_won)? " (writer)" : "")); free(key); free(set); free(host); send_attrd_message(NULL, xml); /* ends up at attrd_peer_message() */ }
/* JOIN状態チェック処理 */ gboolean check_join_state(enum crmd_fsa_state cur_state, const char *source) { crm_debug("Invoked by %s in state: %s", source, fsa_state2string(cur_state)); if(saved_ccm_membership_id != crm_peer_seq) { crm_info("%s: Membership changed since join started: %llu -> %llu", source, saved_ccm_membership_id, crm_peer_seq); register_fsa_input_before(C_FSA_INTERNAL, I_NODE_JOIN, NULL); } else if(cur_state == S_INTEGRATION) { /* 現在のstateがS_INTEGRATIONの時 */ if(g_hash_table_size(welcomed_nodes) == 0) { /* welcomed_nodesハッシュテーブルのサイズが0(すべての認識済みのノードからCRM_OP_JOIN_REQUESTを受信した)の場合 */ crm_debug("join-%d: Integration of %d peers complete: %s", current_join_id, g_hash_table_size(integrated_nodes), source); /* I_INTEGRATEDへ */ register_fsa_input_before( C_FSA_INTERNAL, I_INTEGRATED, NULL); return TRUE; } } else if(cur_state == S_FINALIZE_JOIN) { /* 現在のstateがS_FINALIZE_JOINの時 */ if(is_set(fsa_input_register, R_HAVE_CIB) == FALSE) { crm_debug("join-%d: Delaying I_FINALIZED until we have the CIB", current_join_id); return TRUE; } else if(g_hash_table_size(integrated_nodes) == 0 && g_hash_table_size(finalized_nodes) == 0) { crm_debug("join-%d complete: %s", current_join_id, source); /* I_FINALIZEDへ */ register_fsa_input_later(C_FSA_INTERNAL, I_FINALIZED, NULL); } else if(g_hash_table_size(integrated_nodes) != 0 && g_hash_table_size(finalized_nodes) != 0) { char *msg = NULL; crm_err("join-%d: Waiting on %d integrated nodes" " AND %d finalized nodes", current_join_id, g_hash_table_size(integrated_nodes), g_hash_table_size(finalized_nodes)); msg = crm_strdup("Integrated node"); g_hash_table_foreach(integrated_nodes, ghash_print_node, msg); crm_free(msg); msg = crm_strdup("Finalized node"); g_hash_table_foreach(finalized_nodes, ghash_print_node, msg); crm_free(msg); } else if(g_hash_table_size(integrated_nodes) != 0) { crm_debug("join-%d: Still waiting on %d integrated nodes", current_join_id, g_hash_table_size(integrated_nodes)); } else if(g_hash_table_size(finalized_nodes) != 0) { crm_debug("join-%d: Still waiting on %d finalized nodes", current_join_id, g_hash_table_size(finalized_nodes)); } } return FALSE; }
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 }