diff --git a/include/crm/common/internal.h b/include/crm/common/internal.h index b1d58283d7..2b0c294cf4 100644 --- a/include/crm/common/internal.h +++ b/include/crm/common/internal.h @@ -1,235 +1,235 @@ /* * Copyright 2015-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef CRM_COMMON_INTERNAL__H #define CRM_COMMON_INTERNAL__H #include // getpid() #include // bool #include // strcmp() #include // uid_t, gid_t, pid_t #include // guint, GList, GHashTable #include // xmlNode #include // crm_strdup_printf() // Internal ACL-related utilities (from acl.c) char *pcmk__uid2username(uid_t uid); const char *pcmk__update_acl_user(xmlNode *request, const char *field, const char *peer_user); #if ENABLE_ACL # include static inline bool pcmk__is_privileged(const char *user) { return user && (!strcmp(user, CRM_DAEMON_USER) || !strcmp(user, "root")); } #endif #if SUPPORT_CIBSECRETS // Internal CIB utilities (from cib_secrets.c) */ int pcmk__substitute_secrets(const char *rsc_id, GHashTable *params); #endif /* internal digest-related utilities (from digest.c) */ bool pcmk__verify_digest(xmlNode *input, const char *expected); /* internal I/O utilities (from io.c) */ int pcmk__real_path(const char *path, char **resolved_path); char *pcmk__series_filename(const char *directory, const char *series, int sequence, bool bzip); int pcmk__read_series_sequence(const char *directory, const char *series, unsigned int *seq); void pcmk__write_series_sequence(const char *directory, const char *series, unsigned int sequence, int max); int pcmk__chown_series_sequence(const char *directory, const char *series, uid_t uid, gid_t gid); bool pcmk__daemon_can_write(const char *dir, const char *file); void pcmk__sync_directory(const char *name); int pcmk__file_contents(const char *filename, char **contents); int pcmk__write_sync(int fd, const char *contents); int pcmk__set_nonblocking(int fd); const char *pcmk__get_tmpdir(void); void pcmk__close_fds_in_child(bool); /* internal procfs utilities (from procfs.c) */ pid_t pcmk__procfs_pid_of(const char *name); unsigned int pcmk__procfs_num_cores(void); /* internal XML schema functions (from xml.c) */ void crm_schema_init(void); void crm_schema_cleanup(void); /* internal functions related to process IDs (from pid.c) */ /*! * \internal * \brief Check whether process exists (by PID and optionally executable path) * * \param[in] pid PID of process to check * \param[in] daemon If not NULL, path component to match with procfs entry * * \return Standard Pacemaker return code * \note Particular return codes of interest include pcmk_rc_ok for alive, * ESRCH for process is not alive (verified by kill and/or executable path * match), EACCES for caller unable or not allowed to check. A result of * "alive" is less reliable when \p daemon is not provided or procfs is * not available, since there is no guarantee that the PID has not been * recycled for another process. * \note This function cannot be used to verify \e authenticity of the process. */ int pcmk__pid_active(pid_t pid, const char *daemon); int pcmk__read_pidfile(const char *filename, pid_t *pid); int pcmk__pidfile_matches(const char *filename, pid_t expected_pid, const char *expected_name, pid_t *pid); int pcmk__lock_pidfile(const char *filename, const char *name); /* interal functions related to resource operations (from operations.c) */ // printf-style format to create operation ID from resource, action, interval #define PCMK__OP_FMT "%s_%s_%u" char *pcmk__op_key(const char *rsc_id, const char *op_type, guint interval_ms); char *pcmk__notify_key(const char *rsc_id, const char *notify_type, const char *op_type); char *pcmk__transition_key(int transition_id, int action_id, int target_rc, const char *node); void pcmk__filter_op_for_digest(xmlNode *param_set); // miscellaneous utilities (from utils.c) const char *pcmk_message_name(const char *name); /* internal generic string functions (from strings.c) */ int pcmk__guint_from_hash(GHashTable *table, const char *key, guint default_val, guint *result); bool pcmk__starts_with(const char *str, const char *prefix); bool pcmk__ends_with(const char *s, const char *match); bool pcmk__ends_with_ext(const char *s, const char *match); char *pcmk__add_word(char *list, const char *word); int pcmk__compress(const char *data, unsigned int length, unsigned int max, char **result, unsigned int *result_len); /* Correctly displaying singular or plural is complicated; consider "1 node has" * vs. "2 nodes have". A flexible solution is to pluralize entire strings, e.g. * * if (a == 1) { * crm_info("singular message"): * } else { * crm_info("plural message"); * } * * though even that's not sufficient for all languages besides English (if we * ever desire to do translations of output and log messages). But the following * convenience macros are "good enough" and more concise for many cases. */ /* Example: * crm_info("Found %d %s", nentries, * pcmk__plural_alt(nentries, "entry", "entries")); */ #define pcmk__plural_alt(i, s1, s2) (((i) == 1)? (s1) : (s2)) // Example: crm_info("Found %d node%s", nnodes, pcmk__plural_s(nnodes)); #define pcmk__plural_s(i) pcmk__plural_alt(i, "", "s") static inline int pcmk__str_empty(const char *s) { return (s == NULL) || (s[0] == '\0'); } static inline char * -crm_getpid_s() +pcmk__getpid_s() { return crm_strdup_printf("%lu", (unsigned long) getpid()); } // More efficient than g_list_length(list) == 1 static inline bool pcmk__list_of_1(GList *list) { return list && (list->next == NULL); } // More efficient than g_list_length(list) > 1 static inline bool pcmk__list_of_multiple(GList *list) { return list && (list->next != NULL); } /* convenience functions for failure-related node attributes */ #define PCMK__FAIL_COUNT_PREFIX "fail-count" #define PCMK__LAST_FAILURE_PREFIX "last-failure" /*! * \internal * \brief Generate a failure-related node attribute name for a resource * * \param[in] prefix Start of attribute name * \param[in] rsc_id Resource name * \param[in] op Operation name * \param[in] interval_ms Operation interval * * \return Newly allocated string with attribute name * * \note Failure attributes are named like PREFIX-RSC#OP_INTERVAL (for example, * "fail-count-myrsc#monitor_30000"). The '#' is used because it is not * a valid character in a resource ID, to reliably distinguish where the * operation name begins. The '_' is used simply to be more comparable to * action labels like "myrsc_monitor_30000". */ static inline char * pcmk__fail_attr_name(const char *prefix, const char *rsc_id, const char *op, guint interval_ms) { CRM_CHECK(prefix && rsc_id && op, return NULL); return crm_strdup_printf("%s-%s#%s_%u", prefix, rsc_id, op, interval_ms); } static inline char * pcmk__failcount_name(const char *rsc_id, const char *op, guint interval_ms) { return pcmk__fail_attr_name(PCMK__FAIL_COUNT_PREFIX, rsc_id, op, interval_ms); } static inline char * pcmk__lastfailure_name(const char *rsc_id, const char *op, guint interval_ms) { return pcmk__fail_attr_name(PCMK__LAST_FAILURE_PREFIX, rsc_id, op, interval_ms); } #endif /* CRM_COMMON_INTERNAL__H */ diff --git a/lib/cluster/cpg.c b/lib/cluster/cpg.c index 877d56767e..371bd299bd 100644 --- a/lib/cluster/cpg.c +++ b/lib/cluster/cpg.c @@ -1,810 +1,810 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* PCMK__SPECIAL_PID* */ cpg_handle_t pcmk_cpg_handle = 0; /* TODO: Remove, use cluster.cpg_handle */ static bool cpg_evicted = FALSE; gboolean(*pcmk_cpg_dispatch_fn) (int kind, const char *from, const char *data) = NULL; #define cs_repeat(counter, max, code) do { \ code; \ if(rc == CS_ERR_TRY_AGAIN || rc == CS_ERR_QUEUE_FULL) { \ counter++; \ crm_debug("Retrying operation after %ds", counter); \ sleep(counter); \ } else { \ break; \ } \ } while(counter < max) void cluster_disconnect_cpg(crm_cluster_t *cluster) { pcmk_cpg_handle = 0; if (cluster->cpg_handle) { crm_trace("Disconnecting CPG"); cpg_leave(cluster->cpg_handle, &cluster->group); cpg_finalize(cluster->cpg_handle); cluster->cpg_handle = 0; } else { crm_info("No CPG connection"); } } uint32_t get_local_nodeid(cpg_handle_t handle) { cs_error_t rc = CS_OK; int retries = 0; static uint32_t local_nodeid = 0; cpg_handle_t local_handle = handle; cpg_callbacks_t cb = { }; int fd = -1; uid_t found_uid = 0; gid_t found_gid = 0; pid_t found_pid = 0; int rv; if(local_nodeid != 0) { return local_nodeid; } if(handle == 0) { crm_trace("Creating connection"); cs_repeat(retries, 5, rc = cpg_initialize(&local_handle, &cb)); if (rc != CS_OK) { crm_err("Could not connect to the CPG API: %s (%d)", cs_strerror(rc), rc); return 0; } rc = cpg_fd_get(local_handle, &fd); if (rc != CS_OK) { crm_err("Could not obtain the CPG API connection: %s (%d)", cs_strerror(rc), rc); goto bail; } /* CPG provider run as root (in given user namespace, anyway)? */ if (!(rv = crm_ipc_is_authentic_process(fd, (uid_t) 0,(gid_t) 0, &found_pid, &found_uid, &found_gid))) { crm_err("CPG provider is not authentic:" " process %lld (uid: %lld, gid: %lld)", (long long) PCMK__SPECIAL_PID_AS_0(found_pid), (long long) found_uid, (long long) found_gid); goto bail; } else if (rv < 0) { crm_err("Could not verify authenticity of CPG provider: %s (%d)", strerror(-rv), -rv); goto bail; } } if (rc == CS_OK) { retries = 0; crm_trace("Performing lookup"); cs_repeat(retries, 5, rc = cpg_local_get(local_handle, &local_nodeid)); } if (rc != CS_OK) { crm_err("Could not get local node id from the CPG API: %s (%d)", ais_error2text(rc), rc); } bail: if(handle == 0) { crm_trace("Closing connection"); cpg_finalize(local_handle); } crm_debug("Local nodeid is %u", local_nodeid); return local_nodeid; } GListPtr cs_message_queue = NULL; int cs_message_timer = 0; static ssize_t crm_cs_flush(gpointer data); static gboolean crm_cs_flush_cb(gpointer data) { cs_message_timer = 0; crm_cs_flush(data); return FALSE; } #define CS_SEND_MAX 200 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=%llu", (unsigned long long) 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 (%lld)", sent, queue_len, last_sent, ais_error2text(rc), (long long) rc); } else { crm_trace("Sent %d CPG messages (%d remaining, last=%u): %s (%lld)", sent, queue_len, last_sent, ais_error2text(rc), (long long) 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; } gboolean send_cpg_iov(struct iovec * iov) { static unsigned int queued = 0; queued++; crm_trace("Queueing CPG message %u (%llu bytes)", queued, (unsigned long long) iov->iov_len); cs_message_queue = g_list_append(cs_message_queue, iov); crm_cs_flush(&pcmk_cpg_handle); return TRUE; } static int pcmk_cpg_dispatch(gpointer user_data) { int rc = 0; crm_cluster_t *cluster = (crm_cluster_t*) user_data; rc = cpg_dispatch(cluster->cpg_handle, CS_DISPATCH_ONE); if (rc != CS_OK) { crm_err("Connection to the CPG API failed: %s (%d)", ais_error2text(rc), rc); cluster->cpg_handle = 0; return -1; } else if(cpg_evicted) { crm_err("Evicted from CPG membership"); return -1; } return 0; } char * pcmk_message_common_cs(cpg_handle_t handle, uint32_t nodeid, uint32_t pid, void *content, uint32_t *kind, const char **from) { char *data = NULL; AIS_Message *msg = (AIS_Message *) content; if(handle) { // Do filtering and field massaging uint32_t local_nodeid = get_local_nodeid(handle); const char *local_name = get_local_node_name(); if (msg->sender.id > 0 && msg->sender.id != nodeid) { crm_err("Nodeid mismatch from %d.%d: claimed nodeid=%u", nodeid, pid, msg->sender.id); return NULL; } else if (msg->host.id != 0 && (local_nodeid != msg->host.id)) { /* Not for us */ crm_trace("Not for us: %u != %u", msg->host.id, local_nodeid); return NULL; } else if (msg->host.size != 0 && safe_str_neq(msg->host.uname, local_name)) { /* Not for us */ crm_trace("Not for us: %s != %s", msg->host.uname, local_name); return NULL; } msg->sender.id = nodeid; if (msg->sender.size == 0) { crm_node_t *peer = crm_get_peer(nodeid, NULL); if (peer == NULL) { crm_err("Peer with nodeid=%u is unknown", nodeid); } else if (peer->uname == NULL) { crm_err("No uname for peer with nodeid=%u", nodeid); } else { crm_notice("Fixing uname for peer with nodeid=%u", nodeid); msg->sender.size = strlen(peer->uname); memset(msg->sender.uname, 0, MAX_NAME); memcpy(msg->sender.uname, peer->uname, msg->sender.size); } } } crm_trace("Got new%s message (size=%d, %d, %d)", msg->is_compressed ? " compressed" : "", ais_data_len(msg), msg->size, msg->compressed_size); if (kind != NULL) { *kind = msg->header.id; } if (from != NULL) { *from = msg->sender.uname; } if (msg->is_compressed && msg->size > 0) { int rc = BZ_OK; char *uncompressed = NULL; unsigned int new_size = msg->size + 1; if (check_message_sanity(msg, NULL) == FALSE) { goto badmsg; } crm_trace("Decompressing message data"); uncompressed = calloc(1, new_size); rc = BZ2_bzBuffToBuffDecompress(uncompressed, &new_size, msg->data, msg->compressed_size, 1, 0); if (rc != BZ_OK) { crm_err("Decompression failed: %s " CRM_XS " bzerror=%d", bz2_strerror(rc), rc); free(uncompressed); goto badmsg; } CRM_ASSERT(rc == BZ_OK); CRM_ASSERT(new_size == msg->size); data = uncompressed; } else if (check_message_sanity(msg, data) == FALSE) { goto badmsg; } else if (safe_str_eq("identify", data)) { - char *pid_s = crm_getpid_s(); + char *pid_s = pcmk__getpid_s(); send_cluster_text(crm_class_cluster, pid_s, TRUE, NULL, crm_msg_ais); free(pid_s); return NULL; } else { data = strdup(msg->data); } // Is this necessary? crm_get_peer(msg->sender.id, msg->sender.uname); crm_trace("Payload: %.200s", data); return data; badmsg: crm_err("Invalid message (id=%d, dest=%s:%s, from=%s:%s.%d):" " min=%d, total=%d, size=%d, bz2_size=%d", msg->id, ais_dest(&(msg->host)), msg_type2text(msg->host.type), ais_dest(&(msg->sender)), msg_type2text(msg->sender.type), msg->sender.pid, (int)sizeof(AIS_Message), msg->header.size, msg->size, msg->compressed_size); free(data); return NULL; } static int cmp_member_list_nodeid(const void *first, const void *second) { const struct cpg_address *const a = *((const struct cpg_address **) first), *const b = *((const struct cpg_address **) second); if (a->nodeid < b->nodeid) { return -1; } else if (a->nodeid > b->nodeid) { return 1; } /* don't bother with "reason" nor "pid" */ return 0; } static const char * cpgreason2str(cpg_reason_t reason) { switch (reason) { case CPG_REASON_JOIN: return " via cpg_join"; case CPG_REASON_LEAVE: return " via cpg_leave"; case CPG_REASON_NODEDOWN: return " via cluster exit"; case CPG_REASON_NODEUP: return " via cluster join"; case CPG_REASON_PROCDOWN: return " for unknown reason"; default: break; } return ""; } static inline const char * peer_name(crm_node_t *peer) { if (peer == NULL) { return "unknown node"; } else if (peer->uname == NULL) { return "peer node"; } else { return peer->uname; } } void pcmk_cpg_membership(cpg_handle_t handle, const struct cpg_name *groupName, const struct cpg_address *member_list, size_t member_list_entries, const struct cpg_address *left_list, size_t left_list_entries, const struct cpg_address *joined_list, size_t joined_list_entries) { int i; gboolean found = FALSE; static int counter = 0; uint32_t local_nodeid = get_local_nodeid(handle); const struct cpg_address *key, **sorted; sorted = malloc(member_list_entries * sizeof(const struct cpg_address *)); CRM_ASSERT(sorted != NULL); for (size_t iter = 0; iter < member_list_entries; iter++) { sorted[iter] = member_list + iter; } /* so that the cross-matching multiply-subscribed nodes is then cheap */ qsort(sorted, member_list_entries, sizeof(const struct cpg_address *), cmp_member_list_nodeid); for (i = 0; i < left_list_entries; i++) { crm_node_t *peer = crm_find_peer(left_list[i].nodeid, NULL); const struct cpg_address **rival = NULL; /* in CPG world, NODE:PROCESS-IN-MEMBERSHIP-OF-G is an 1:N relation and not playing by this rule may go wild in case of multiple residual instances of the same pacemaker daemon at the same node -- we must ensure that the possible local rival(s) won't make us cry out and bail (e.g. when they quit themselves), since all the surrounding logic denies this simple fact that the full membership is discriminated also per the PID of the process beside mere node ID (and implicitly, group ID); practically, this will be sound in terms of not preventing progress, since all the CPG joiners are also API end-point carriers, and that's what matters locally (who's the winner); remotely, we will just compare leave_list and member_list and if the left process has its node retained in member_list (under some other PID, anyway) we will just ignore it as well XXX: long-term fix is to establish in-out PID-aware tracking? */ if (peer) { key = &left_list[i]; rival = bsearch(&key, sorted, member_list_entries, sizeof(const struct cpg_address *), cmp_member_list_nodeid); } if (rival == NULL) { crm_info("Group %s event %d: %s (node %u pid %u) left%s", groupName->value, counter, peer_name(peer), left_list[i].nodeid, left_list[i].pid, cpgreason2str(left_list[i].reason)); if (peer) { crm_update_peer_proc(__FUNCTION__, peer, crm_proc_cpg, OFFLINESTATUS); } } else if (left_list[i].nodeid == local_nodeid) { crm_warn("Group %s event %d: duplicate local pid %u left%s", groupName->value, counter, left_list[i].pid, cpgreason2str(left_list[i].reason)); } else { crm_warn("Group %s event %d: " "%s (node %u) duplicate pid %u left%s (%u remains)", groupName->value, counter, peer_name(peer), left_list[i].nodeid, left_list[i].pid, cpgreason2str(left_list[i].reason), (*rival)->pid); } } free(sorted); sorted = NULL; for (i = 0; i < joined_list_entries; i++) { crm_info("Group %s event %d: node %u pid %u joined%s", groupName->value, counter, joined_list[i].nodeid, joined_list[i].pid, cpgreason2str(joined_list[i].reason)); } for (i = 0; i < member_list_entries; i++) { crm_node_t *peer = crm_get_peer(member_list[i].nodeid, NULL); if (member_list[i].nodeid == local_nodeid && member_list[i].pid != getpid()) { /* see the note above */ crm_warn("Group %s event %d: detected duplicate local pid %u", groupName->value, counter, member_list[i].pid); continue; } crm_info("Group %s event %d: %s (node %u pid %u) is member", groupName->value, counter, peer_name(peer), member_list[i].nodeid, member_list[i].pid); /* If the caller left auto-reaping enabled, this will also update the * state to member. */ peer = crm_update_peer_proc(__FUNCTION__, peer, crm_proc_cpg, ONLINESTATUS); if (peer && peer->state && strcmp(peer->state, CRM_NODE_MEMBER)) { /* The node is a CPG member, but we currently think it's not a * cluster member. This is possible only if auto-reaping was * disabled. The node may be joining, and we happened to get the CPG * notification before the quorum notification; or the node may have * just died, and we are processing its final messages; or a bug * has affected the peer cache. */ time_t now = time(NULL); if (peer->when_lost == 0) { // Track when we first got into this contradictory state peer->when_lost = now; } else if (now > (peer->when_lost + 60)) { // If it persists for more than a minute, update the state crm_warn("Node %u is member of group %s but was believed offline", member_list[i].nodeid, groupName->value); crm_update_peer_state(__FUNCTION__, peer, CRM_NODE_MEMBER, 0); } } if (local_nodeid == member_list[i].nodeid) { found = TRUE; } } if (!found) { crm_err("Local node was evicted from group %s", groupName->value); cpg_evicted = TRUE; } counter++; } gboolean cluster_connect_cpg(crm_cluster_t *cluster) { cs_error_t rc; int fd = -1; int retries = 0; uint32_t id = 0; crm_node_t *peer = NULL; cpg_handle_t handle = 0; const char *message_name = pcmk_message_name(crm_system_name); uid_t found_uid = 0; gid_t found_gid = 0; pid_t found_pid = 0; int rv; struct mainloop_fd_callbacks cpg_fd_callbacks = { .dispatch = pcmk_cpg_dispatch, .destroy = cluster->destroy, }; cpg_callbacks_t cpg_callbacks = { .cpg_deliver_fn = cluster->cpg.cpg_deliver_fn, .cpg_confchg_fn = cluster->cpg.cpg_confchg_fn, /* .cpg_deliver_fn = pcmk_cpg_deliver, */ /* .cpg_confchg_fn = pcmk_cpg_membership, */ }; cpg_evicted = FALSE; cluster->group.length = 0; cluster->group.value[0] = 0; /* group.value is char[128] */ strncpy(cluster->group.value, message_name, 127); cluster->group.value[127] = 0; cluster->group.length = 1 + QB_MIN(127, strlen(cluster->group.value)); cs_repeat(retries, 30, rc = cpg_initialize(&handle, &cpg_callbacks)); if (rc != CS_OK) { crm_err("Could not connect to the CPG API: %s (%d)", cs_strerror(rc), rc); goto bail; } rc = cpg_fd_get(handle, &fd); if (rc != CS_OK) { crm_err("Could not obtain the CPG API connection: %s (%d)", cs_strerror(rc), rc); goto bail; } /* CPG provider run as root (in given user namespace, anyway)? */ if (!(rv = crm_ipc_is_authentic_process(fd, (uid_t) 0,(gid_t) 0, &found_pid, &found_uid, &found_gid))) { crm_err("CPG provider is not authentic:" " process %lld (uid: %lld, gid: %lld)", (long long) PCMK__SPECIAL_PID_AS_0(found_pid), (long long) found_uid, (long long) found_gid); rc = CS_ERR_ACCESS; goto bail; } else if (rv < 0) { crm_err("Could not verify authenticity of CPG provider: %s (%d)", strerror(-rv), -rv); rc = CS_ERR_ACCESS; goto bail; } id = get_local_nodeid(handle); if (id == 0) { crm_err("Could not get local node id from the CPG API"); goto bail; } cluster->nodeid = id; retries = 0; cs_repeat(retries, 30, rc = cpg_join(handle, &cluster->group)); if (rc != CS_OK) { crm_err("Could not join the CPG group '%s': %d", message_name, rc); goto bail; } pcmk_cpg_handle = handle; cluster->cpg_handle = handle; mainloop_add_fd("corosync-cpg", G_PRIORITY_MEDIUM, fd, cluster, &cpg_fd_callbacks); bail: if (rc != CS_OK) { cpg_finalize(handle); return FALSE; } peer = crm_get_peer(id, NULL); crm_update_peer_proc(__FUNCTION__, peer, crm_proc_cpg, ONLINESTATUS); return TRUE; } gboolean send_cluster_message_cs(xmlNode * msg, gboolean local, crm_node_t * node, enum crm_ais_msg_types dest) { gboolean rc = TRUE; char *data = NULL; data = dump_xml_unformatted(msg); rc = send_cluster_text(crm_class_cluster, data, local, node, dest); free(data); return rc; } gboolean send_cluster_text(enum crm_ais_msg_class msg_class, const char *data, gboolean local, crm_node_t *node, enum crm_ais_msg_types dest) { static int msg_id = 0; static int local_pid = 0; static int local_name_len = 0; static const char *local_name = NULL; char *target = NULL; struct iovec *iov; AIS_Message *msg = NULL; enum crm_ais_msg_types sender = text2msg_type(crm_system_name); switch (msg_class) { case crm_class_cluster: break; default: crm_err("Invalid message class: %d", msg_class); return FALSE; } CRM_CHECK(dest != crm_msg_ais, return FALSE); if(local_name == NULL) { local_name = get_local_node_name(); } if(local_name_len == 0 && local_name) { local_name_len = strlen(local_name); } if (data == NULL) { data = ""; } if (local_pid == 0) { local_pid = getpid(); } if (sender == crm_msg_none) { sender = local_pid; } msg = calloc(1, sizeof(AIS_Message)); msg_id++; msg->id = msg_id; msg->header.id = msg_class; msg->header.error = CS_OK; msg->host.type = dest; msg->host.local = local; if (node) { if (node->uname) { target = strdup(node->uname); msg->host.size = strlen(node->uname); memset(msg->host.uname, 0, MAX_NAME); memcpy(msg->host.uname, node->uname, msg->host.size); } else { target = crm_strdup_printf("%u", node->id); } msg->host.id = node->id; } else { target = strdup("all"); } msg->sender.id = 0; msg->sender.type = sender; msg->sender.pid = local_pid; msg->sender.size = local_name_len; memset(msg->sender.uname, 0, MAX_NAME); if(local_name && msg->sender.size) { memcpy(msg->sender.uname, local_name, msg->sender.size); } msg->size = 1 + strlen(data); msg->header.size = sizeof(AIS_Message) + msg->size; if (msg->size < CRM_BZ2_THRESHOLD) { msg = realloc_safe(msg, msg->header.size); memcpy(msg->data, data, msg->size); } else { char *compressed = NULL; unsigned int new_size = 0; char *uncompressed = strdup(data); if (pcmk__compress(uncompressed, (unsigned int) msg->size, 0, &compressed, &new_size) == pcmk_rc_ok) { msg->header.size = sizeof(AIS_Message) + new_size; msg = realloc_safe(msg, msg->header.size); memcpy(msg->data, compressed, new_size); msg->is_compressed = TRUE; msg->compressed_size = new_size; } else { msg = realloc_safe(msg, msg->header.size); memcpy(msg->data, data, msg->size); } free(uncompressed); free(compressed); } iov = calloc(1, sizeof(struct iovec)); iov->iov_base = msg; iov->iov_len = msg->header.size; if (msg->compressed_size) { crm_trace("Queueing CPG message %u to %s (%llu bytes, %d bytes compressed payload): %.200s", msg->id, target, (unsigned long long) iov->iov_len, msg->compressed_size, data); } else { crm_trace("Queueing CPG message %u to %s (%llu bytes, %d bytes payload): %.200s", msg->id, target, (unsigned long long) iov->iov_len, msg->size, data); } free(target); send_cpg_iov(iov); return TRUE; } enum crm_ais_msg_types text2msg_type(const char *text) { int type = crm_msg_none; CRM_CHECK(text != NULL, return type); text = pcmk_message_name(text); if (safe_str_eq(text, "ais")) { type = crm_msg_ais; } else if (safe_str_eq(text, CRM_SYSTEM_CIB)) { type = crm_msg_cib; } else if (safe_str_eq(text, CRM_SYSTEM_CRMD) || safe_str_eq(text, CRM_SYSTEM_DC)) { type = crm_msg_crmd; } else if (safe_str_eq(text, CRM_SYSTEM_TENGINE)) { type = crm_msg_te; } else if (safe_str_eq(text, CRM_SYSTEM_PENGINE)) { type = crm_msg_pe; } else if (safe_str_eq(text, CRM_SYSTEM_LRMD)) { type = crm_msg_lrmd; } else if (safe_str_eq(text, CRM_SYSTEM_STONITHD)) { type = crm_msg_stonithd; } else if (safe_str_eq(text, "stonith-ng")) { type = crm_msg_stonith_ng; } else if (safe_str_eq(text, "attrd")) { type = crm_msg_attrd; } else { /* This will normally be a transient client rather than * a cluster daemon. Set the type to the pid of the client */ int scan_rc = sscanf(text, "%d", &type); if (scan_rc != 1 || type <= crm_msg_stonith_ng) { /* Ensure it's sane */ type = crm_msg_none; } } return type; } diff --git a/lib/common/utils.c b/lib/common/utils.c index 243305bde9..99ad6592ec 100644 --- a/lib/common/utils.c +++ b/lib/common/utils.c @@ -1,654 +1,654 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #ifndef _GNU_SOURCE # define _GNU_SOURCE #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifndef MAXLINE # define MAXLINE 512 #endif #ifndef PW_BUFFER_LEN # define PW_BUFFER_LEN 500 #endif CRM_TRACE_INIT_DATA(common); gboolean crm_config_error = FALSE; gboolean crm_config_warning = FALSE; char *crm_system_name = NULL; int node_score_red = 0; int node_score_green = 0; int node_score_yellow = 0; int char2score(const char *score) { int score_f = 0; if (score == NULL) { } else if (safe_str_eq(score, CRM_MINUS_INFINITY_S)) { score_f = -CRM_SCORE_INFINITY; } else if (safe_str_eq(score, CRM_INFINITY_S)) { score_f = CRM_SCORE_INFINITY; } else if (safe_str_eq(score, CRM_PLUS_INFINITY_S)) { score_f = CRM_SCORE_INFINITY; } else if (safe_str_eq(score, "red")) { score_f = node_score_red; } else if (safe_str_eq(score, "yellow")) { score_f = node_score_yellow; } else if (safe_str_eq(score, "green")) { score_f = node_score_green; } else { score_f = crm_parse_int(score, NULL); if (score_f > 0 && score_f > CRM_SCORE_INFINITY) { score_f = CRM_SCORE_INFINITY; } else if (score_f < 0 && score_f < -CRM_SCORE_INFINITY) { score_f = -CRM_SCORE_INFINITY; } } return score_f; } char * score2char_stack(int score, char *buf, size_t len) { if (score >= CRM_SCORE_INFINITY) { strncpy(buf, CRM_INFINITY_S, 9); } else if (score <= -CRM_SCORE_INFINITY) { strncpy(buf, CRM_MINUS_INFINITY_S , 10); } else { return crm_itoa_stack(score, buf, len); } return buf; } char * score2char(int score) { if (score >= CRM_SCORE_INFINITY) { return strdup(CRM_INFINITY_S); } else if (score <= -CRM_SCORE_INFINITY) { return strdup(CRM_MINUS_INFINITY_S); } return crm_itoa(score); } char * generate_hash_key(const char *crm_msg_reference, const char *sys) { char *hash_key = crm_strdup_printf("%s_%s", (sys? sys : "none"), - crm_msg_reference, '_'); + crm_msg_reference); crm_trace("created hash key: (%s)", hash_key); return hash_key; } int crm_user_lookup(const char *name, uid_t * uid, gid_t * gid) { int rc = pcmk_ok; char *buffer = NULL; struct passwd pwd; struct passwd *pwentry = NULL; buffer = calloc(1, PW_BUFFER_LEN); if (buffer == NULL) { return -ENOMEM; } rc = getpwnam_r(name, &pwd, buffer, PW_BUFFER_LEN, &pwentry); if (pwentry) { if (uid) { *uid = pwentry->pw_uid; } if (gid) { *gid = pwentry->pw_gid; } crm_trace("User %s has uid=%d gid=%d", name, pwentry->pw_uid, pwentry->pw_gid); } else { rc = rc? -rc : -EINVAL; crm_info("User %s lookup: %s", name, pcmk_strerror(rc)); } free(buffer); return rc; } /*! * \brief Get user and group IDs of pacemaker daemon user * * \param[out] uid If non-NULL, where to store daemon user ID * \param[out] gid If non-NULL, where to store daemon group ID * * \return pcmk_ok on success, -errno otherwise */ int pcmk_daemon_user(uid_t *uid, gid_t *gid) { static uid_t daemon_uid; static gid_t daemon_gid; static bool found = false; int rc = pcmk_err_generic; if (!found) { rc = crm_user_lookup(CRM_DAEMON_USER, &daemon_uid, &daemon_gid); if (rc == pcmk_ok) { found = true; } } if (found) { if (uid) { *uid = daemon_uid; } if (gid) { *gid = daemon_gid; } } return rc; } static int crm_version_helper(const char *text, const char **end_text) { int atoi_result = -1; CRM_ASSERT(end_text != NULL); errno = 0; if (text != NULL && text[0] != 0) { /* seemingly sacrificing const-correctness -- because while strtol doesn't modify the input, it doesn't want to artificially taint the "end_text" pointer-to-pointer-to-first-char-in-string with constness in case the input wasn't actually constant -- by semantic definition not a single character will get modified so it shall be perfectly safe to make compiler happy with dropping "const" qualifier here */ atoi_result = (int) strtol(text, (char **) end_text, 10); if (errno == EINVAL) { crm_err("Conversion of '%s' %c failed", text, text[0]); atoi_result = -1; } } return atoi_result; } /* * version1 < version2 : -1 * version1 = version2 : 0 * version1 > version2 : 1 */ int compare_version(const char *version1, const char *version2) { int rc = 0; int lpc = 0; const char *ver1_iter, *ver2_iter; if (version1 == NULL && version2 == NULL) { return 0; } else if (version1 == NULL) { return -1; } else if (version2 == NULL) { return 1; } ver1_iter = version1; ver2_iter = version2; while (1) { int digit1 = 0; int digit2 = 0; lpc++; if (ver1_iter == ver2_iter) { break; } if (ver1_iter != NULL) { digit1 = crm_version_helper(ver1_iter, &ver1_iter); } if (ver2_iter != NULL) { digit2 = crm_version_helper(ver2_iter, &ver2_iter); } if (digit1 < digit2) { rc = -1; break; } else if (digit1 > digit2) { rc = 1; break; } if (ver1_iter != NULL && *ver1_iter == '.') { ver1_iter++; } if (ver1_iter != NULL && *ver1_iter == '\0') { ver1_iter = NULL; } if (ver2_iter != NULL && *ver2_iter == '.') { ver2_iter++; } if (ver2_iter != NULL && *ver2_iter == 0) { ver2_iter = NULL; } } if (rc == 0) { crm_trace("%s == %s (%d)", version1, version2, lpc); } else if (rc < 0) { crm_trace("%s < %s (%d)", version1, version2, lpc); } else if (rc > 0) { crm_trace("%s > %s (%d)", version1, version2, lpc); } return rc; } /*! * \brief Parse milliseconds from a Pacemaker interval specification * * \param[in] input Pacemaker time interval specification (a bare number of * seconds, a number with a unit optionally with whitespace * before and/or after the number, or an ISO 8601 duration) * * \return Milliseconds equivalent of given specification on success (limited * to the range of an unsigned integer), 0 if input is NULL, * or 0 (and set errno to EINVAL) on error */ guint crm_parse_interval_spec(const char *input) { long long msec = -1; errno = 0; if (input == NULL) { return 0; } else if (input[0] == 'P') { crm_time_t *period_s = crm_time_parse_duration(input); if (period_s) { msec = 1000 * crm_time_get_seconds(period_s); crm_time_free(period_s); } } else { msec = crm_get_msec(input); } if (msec < 0) { crm_warn("Using 0 instead of '%s'", input); errno = EINVAL; return 0; } return (msec >= G_MAXUINT)? G_MAXUINT : (guint) msec; } extern bool crm_is_daemon; /* coverity[+kill] */ void crm_abort(const char *file, const char *function, int line, const char *assert_condition, gboolean do_core, gboolean do_fork) { int rc = 0; int pid = 0; int status = 0; /* Implied by the parent's error logging below */ /* crm_write_blackbox(0); */ if(crm_is_daemon == FALSE) { /* This is a command line tool - do not fork */ /* crm_add_logfile(NULL); * Record it to a file? */ crm_enable_stderr(TRUE); /* Make sure stderr is enabled so we can tell the caller */ do_fork = FALSE; /* Just crash if needed */ } if (do_core == FALSE) { crm_err("%s: Triggered assert at %s:%d : %s", function, file, line, assert_condition); return; } else if (do_fork) { pid = fork(); } else { crm_err("%s: Triggered fatal assert at %s:%d : %s", function, file, line, assert_condition); } if (pid == -1) { crm_crit("%s: Cannot create core for non-fatal assert at %s:%d : %s", function, file, line, assert_condition); return; } else if(pid == 0) { /* Child process */ abort(); return; } /* Parent process */ crm_err("%s: Forked child %d to record non-fatal assert at %s:%d : %s", function, pid, file, line, assert_condition); crm_write_blackbox(SIGTRAP, NULL); do { rc = waitpid(pid, &status, 0); if(rc == pid) { return; /* Job done */ } } while(errno == EINTR); if (errno == ECHILD) { /* crm_mon does this */ crm_trace("Cannot wait on forked child %d - SIGCHLD is probably set to SIG_IGN", pid); return; } crm_perror(LOG_ERR, "Cannot wait on forked child %d", pid); } void crm_make_daemon(const char *name, gboolean daemonize, const char *pidfile) { int rc; pid_t pid; const char *devnull = "/dev/null"; if (daemonize == FALSE) { return; } /* Check before we even try... */ rc = pcmk__pidfile_matches(pidfile, 1, name, &pid); if ((rc != pcmk_rc_ok) && (rc != ENOENT)) { crm_err("%s: already running [pid %lld in %s]", name, (long long) pid, pidfile); printf("%s: already running [pid %lld in %s]\n", name, (long long) pid, pidfile); crm_exit(CRM_EX_ERROR); } pid = fork(); if (pid < 0) { fprintf(stderr, "%s: could not start daemon\n", name); crm_perror(LOG_ERR, "fork"); crm_exit(CRM_EX_OSERR); } else if (pid > 0) { crm_exit(CRM_EX_OK); } rc = pcmk__lock_pidfile(pidfile, name); if (rc != pcmk_rc_ok) { crm_err("Could not lock '%s' for %s: %s " CRM_XS " rc=%d", pidfile, name, pcmk_rc_str(rc), rc); printf("Could not lock '%s' for %s: %s (%d)\n", pidfile, name, pcmk_rc_str(rc), rc); crm_exit(CRM_EX_ERROR); } umask(S_IWGRP | S_IWOTH | S_IROTH); close(STDIN_FILENO); (void)open(devnull, O_RDONLY); /* Stdin: fd 0 */ close(STDOUT_FILENO); (void)open(devnull, O_WRONLY); /* Stdout: fd 1 */ close(STDERR_FILENO); (void)open(devnull, O_WRONLY); /* Stderr: fd 2 */ } char * crm_meta_name(const char *field) { int lpc = 0; int max = 0; char *crm_name = NULL; CRM_CHECK(field != NULL, return NULL); crm_name = crm_strdup_printf(CRM_META "_%s", field); /* Massage the names so they can be used as shell variables */ max = strlen(crm_name); for (; lpc < max; lpc++) { switch (crm_name[lpc]) { case '-': crm_name[lpc] = '_'; break; } } return crm_name; } const char * crm_meta_value(GHashTable * hash, const char *field) { char *key = NULL; const char *value = NULL; key = crm_meta_name(field); if (key) { value = g_hash_table_lookup(hash, key); free(key); } return value; } void * find_library_function(void **handle, const char *lib, const char *fn, gboolean fatal) { char *error; void *a_function; if (*handle == NULL) { *handle = dlopen(lib, RTLD_LAZY); } if (!(*handle)) { crm_err("%sCould not open %s: %s", fatal ? "Fatal: " : "", lib, dlerror()); if (fatal) { crm_exit(CRM_EX_FATAL); } return NULL; } a_function = dlsym(*handle, fn); if (a_function == NULL) { error = dlerror(); crm_err("%sCould not find %s in %s: %s", fatal ? "Fatal: " : "", fn, lib, error); if (fatal) { crm_exit(CRM_EX_FATAL); } } return a_function; } #ifdef HAVE_UUID_UUID_H # include #endif char * crm_generate_uuid(void) { unsigned char uuid[16]; char *buffer = malloc(37); /* Including NUL byte */ uuid_generate(uuid); uuid_unparse(uuid, buffer); return buffer; } /*! * \brief Get name to be used as identifier for cluster messages * * \param[in] name Actual system name to check * * \return Non-NULL cluster message identifier corresponding to name * * \note The Pacemaker daemons were renamed in version 2.0.0, but the old names * must continue to be used as the identifier for cluster messages, so * that mixed-version clusters are possible during a rolling upgrade. */ const char * pcmk_message_name(const char *name) { if (name == NULL) { return "unknown"; } else if (!strcmp(name, "pacemaker-attrd")) { return "attrd"; } else if (!strcmp(name, "pacemaker-based")) { return CRM_SYSTEM_CIB; } else if (!strcmp(name, "pacemaker-controld")) { return CRM_SYSTEM_CRMD; } else if (!strcmp(name, "pacemaker-execd")) { return CRM_SYSTEM_LRMD; } else if (!strcmp(name, "pacemaker-fenced")) { return "stonith-ng"; } else if (!strcmp(name, "pacemaker-schedulerd")) { return CRM_SYSTEM_PENGINE; } else { return name; } } /*! * \brief Check whether a string represents a cluster daemon name * * \param[in] name String to check * * \return TRUE if name is standard client name used by daemons, FALSE otherwise */ bool crm_is_daemon_name(const char *name) { name = pcmk_message_name(name); return (!strcmp(name, CRM_SYSTEM_CRMD) || !strcmp(name, CRM_SYSTEM_STONITHD) || !strcmp(name, "stonith-ng") || !strcmp(name, "attrd") || !strcmp(name, CRM_SYSTEM_CIB) || !strcmp(name, CRM_SYSTEM_MCP) || !strcmp(name, CRM_SYSTEM_DC) || !strcmp(name, CRM_SYSTEM_TENGINE) || !strcmp(name, CRM_SYSTEM_LRMD)); } #include char * crm_md5sum(const char *buffer) { int lpc = 0, len = 0; char *digest = NULL; unsigned char raw_digest[MD5_DIGEST_SIZE]; if (buffer == NULL) { buffer = ""; } len = strlen(buffer); crm_trace("Beginning digest of %d bytes", len); digest = malloc(2 * MD5_DIGEST_SIZE + 1); if(digest) { md5_buffer(buffer, len, raw_digest); for (lpc = 0; lpc < MD5_DIGEST_SIZE; lpc++) { sprintf(digest + (2 * lpc), "%02x", raw_digest[lpc]); } digest[(2 * MD5_DIGEST_SIZE)] = 0; crm_trace("Digest %s.", digest); } else { crm_err("Could not create digest"); } return digest; } #ifdef HAVE_GNUTLS_GNUTLS_H void crm_gnutls_global_init(void) { signal(SIGPIPE, SIG_IGN); gnutls_global_init(); } #endif /*! * \brief Get the local hostname * * \return Newly allocated string with name, or NULL (and set errno) on error */ char * pcmk_hostname() { struct utsname hostinfo; return (uname(&hostinfo) < 0)? NULL : strdup(hostinfo.nodename); } diff --git a/tools/crm_node.c b/tools/crm_node.c index d750b0a6ee..3fb9856fac 100644 --- a/tools/crm_node.c +++ b/tools/crm_node.c @@ -1,665 +1,665 @@ /* - * Copyright 2004-2019 the Pacemaker project contributors + * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #define SUMMARY "crm_node - Tool for displaying low-level node information" struct { gboolean corosync; gboolean dangerous_cmd; gboolean force_flag; char command; int nodeid; char *target_uname; } options = { .command = '\0', .force_flag = FALSE }; gboolean command_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean name_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean remove_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); static char *pid_s = NULL; static GMainLoop *mainloop = NULL; static crm_exit_t exit_code = CRM_EX_OK; #define INDENT " " static GOptionEntry command_entries[] = { { "cluster-id", 'i', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display this node's cluster id", NULL }, { "list", 'l', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display all known members (past and present) of this cluster", NULL }, { "name", 'n', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display the name used by the cluster for this node", NULL }, { "partition", 'p', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display the members of this partition", NULL }, { "quorum", 'q', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display a 1 if our partition has quorum, 0 if not", NULL }, { "name-for-id", 'N', 0, G_OPTION_ARG_CALLBACK, name_cb, "Display the name used by the cluster for the node with the specified ID", "ID" }, { "remove", 'R', 0, G_OPTION_ARG_CALLBACK, remove_cb, "(Advanced) Remove the (stopped) node with the specified name from Pacemaker's\n" INDENT "configuration and caches (the node must already have been removed from\n" INDENT "the underlying cluster stack configuration", "NAME" }, { NULL } }; static GOptionEntry addl_entries[] = { { "force", 'f', 0, G_OPTION_ARG_NONE, &options.force_flag, NULL, NULL }, #if SUPPORT_COROSYNC /* Unused and deprecated */ { "corosync", 'C', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_NONE, &options.corosync, NULL, NULL }, #endif // @TODO add timeout option for when IPC replies are needed { NULL } }; gboolean command_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (safe_str_eq("-i", option_name) || safe_str_eq("--cluster-id", option_name)) { options.command = 'i'; } else if (safe_str_eq("-l", option_name) || safe_str_eq("--list", option_name)) { options.command = 'l'; } else if (safe_str_eq("-n", option_name) || safe_str_eq("--name", option_name)) { options.command = 'n'; } else if (safe_str_eq("-p", option_name) || safe_str_eq("--partition", option_name)) { options.command = 'p'; } else if (safe_str_eq("-q", option_name) || safe_str_eq("--quorum", option_name)) { options.command = 'q'; } else { g_set_error(error, G_OPTION_ERROR, CRM_EX_INVALID_PARAM, "Unknown param passed to command_cb: %s\n", option_name); return FALSE; } return TRUE; } gboolean name_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.command = 'N'; options.nodeid = crm_parse_int(optarg, NULL); return TRUE; } gboolean remove_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (optarg == NULL) { crm_err("-R option requires an argument"); g_set_error(error, G_OPTION_ERROR, CRM_EX_INVALID_PARAM, "-R option requires an argument"); return FALSE; } options.command = 'R'; options.dangerous_cmd = TRUE; options.target_uname = strdup(optarg); return TRUE; } /*! * \internal * \brief Exit crm_node * Clean up memory, and either quit mainloop (if running) or exit * * \param[in] value Exit status */ static void crm_node_exit(crm_exit_t value) { if (pid_s) { free(pid_s); pid_s = NULL; } exit_code = value; if (mainloop && g_main_loop_is_running(mainloop)) { g_main_loop_quit(mainloop); } else { crm_exit(exit_code); } } static void exit_disconnect(gpointer user_data) { fprintf(stderr, "error: Lost connection to cluster\n"); crm_node_exit(CRM_EX_DISCONNECT); } typedef int (*ipc_dispatch_fn) (const char *buffer, ssize_t length, gpointer userdata); static crm_ipc_t * new_mainloop_for_ipc(const char *system, ipc_dispatch_fn dispatch) { mainloop_io_t *source = NULL; crm_ipc_t *ipc = NULL; struct ipc_client_callbacks ipc_callbacks = { .dispatch = dispatch, .destroy = exit_disconnect }; mainloop = g_main_loop_new(NULL, FALSE); source = mainloop_add_ipc_client(system, G_PRIORITY_DEFAULT, 0, NULL, &ipc_callbacks); ipc = mainloop_get_ipc_client(source); if (ipc == NULL) { fprintf(stderr, "error: Could not connect to cluster (is it running?)\n"); crm_node_exit(CRM_EX_DISCONNECT); } return ipc; } static void run_mainloop_and_exit() { g_main_loop_run(mainloop); g_main_loop_unref(mainloop); mainloop = NULL; crm_node_exit(exit_code); } static int send_controller_hello(crm_ipc_t *controller) { xmlNode *hello = NULL; int rc; - pid_s = crm_getpid_s(); + pid_s = pcmk__getpid_s(); hello = create_hello_message(pid_s, crm_system_name, "1", "0"); rc = crm_ipc_send(controller, hello, 0, 0, NULL); free_xml(hello); return (rc < 0)? rc : 0; } static int send_node_info_request(crm_ipc_t *controller, uint32_t nodeid) { xmlNode *ping = NULL; int rc; ping = create_request(CRM_OP_NODE_INFO, NULL, NULL, CRM_SYSTEM_CRMD, crm_system_name, pid_s); if (nodeid > 0) { crm_xml_add_int(ping, XML_ATTR_ID, nodeid); } rc = crm_ipc_send(controller, ping, 0, 0, NULL); free_xml(ping); return (rc < 0)? rc : 0; } static int dispatch_controller(const char *buffer, ssize_t length, gpointer userdata) { xmlNode *message = string2xml(buffer); xmlNode *data = NULL; const char *value = NULL; if (message == NULL) { fprintf(stderr, "error: Could not understand reply from controller\n"); crm_node_exit(CRM_EX_PROTOCOL); return 0; } crm_log_xml_trace(message, "controller reply"); exit_code = CRM_EX_PROTOCOL; // Validate reply value = crm_element_value(message, F_CRM_MSG_TYPE); if (safe_str_neq(value, XML_ATTR_RESPONSE)) { fprintf(stderr, "error: Message from controller was not a reply\n"); goto done; } value = crm_element_value(message, XML_ATTR_REFERENCE); if (value == NULL) { fprintf(stderr, "error: Controller reply did not specify original message\n"); goto done; } data = get_message_xml(message, F_CRM_DATA); if (data == NULL) { fprintf(stderr, "error: Controller reply did not contain any data\n"); goto done; } switch (options.command) { case 'i': value = crm_element_value(data, XML_ATTR_ID); if (value == NULL) { fprintf(stderr, "error: Controller reply did not contain node ID\n"); } else { printf("%s\n", value); exit_code = CRM_EX_OK; } break; case 'n': case 'N': value = crm_element_value(data, XML_ATTR_UNAME); if (value == NULL) { fprintf(stderr, "Node is not known to cluster\n"); exit_code = CRM_EX_NOHOST; } else { printf("%s\n", value); exit_code = CRM_EX_OK; } break; case 'q': value = crm_element_value(data, XML_ATTR_HAVE_QUORUM); if (value == NULL) { fprintf(stderr, "error: Controller reply did not contain quorum status\n"); } else { bool quorum = crm_is_true(value); printf("%d\n", quorum); exit_code = quorum? CRM_EX_OK : CRM_EX_QUORUM; } break; default: fprintf(stderr, "internal error: Controller reply not expected\n"); exit_code = CRM_EX_SOFTWARE; break; } done: free_xml(message); crm_node_exit(exit_code); return 0; } static void run_controller_mainloop(uint32_t nodeid) { crm_ipc_t *controller = NULL; int rc; controller = new_mainloop_for_ipc(CRM_SYSTEM_CRMD, dispatch_controller); rc = send_controller_hello(controller); if (rc < 0) { fprintf(stderr, "error: Could not register with controller: %s\n", pcmk_strerror(rc)); crm_node_exit(crm_errno2exit(rc)); } rc = send_node_info_request(controller, nodeid); if (rc < 0) { fprintf(stderr, "error: Could not ping controller: %s\n", pcmk_strerror(rc)); crm_node_exit(crm_errno2exit(rc)); } // Run main loop to get controller reply via dispatch_controller() run_mainloop_and_exit(); } static void print_node_name() { // Check environment first (i.e. when called by resource agent) const char *name = getenv("OCF_RESKEY_" CRM_META "_" XML_LRM_ATTR_TARGET); if (name != NULL) { printf("%s\n", name); crm_node_exit(CRM_EX_OK); } else { // Otherwise ask the controller run_controller_mainloop(0); } } static int cib_remove_node(long id, const char *name) { int rc; cib_t *cib = NULL; xmlNode *node = NULL; xmlNode *node_state = NULL; crm_trace("Removing %s from the CIB", name); if(name == NULL && id == 0) { return -ENOTUNIQ; } node = create_xml_node(NULL, XML_CIB_TAG_NODE); node_state = create_xml_node(NULL, XML_CIB_TAG_STATE); crm_xml_add(node, XML_ATTR_UNAME, name); crm_xml_add(node_state, XML_ATTR_UNAME, name); if (id > 0) { crm_xml_set_id(node, "%ld", id); crm_xml_add(node_state, XML_ATTR_ID, ID(node)); } cib = cib_new(); cib->cmds->signon(cib, crm_system_name, cib_command); rc = cib->cmds->remove(cib, XML_CIB_TAG_NODES, node, cib_sync_call); if (rc != pcmk_ok) { printf("Could not remove %s[%ld] from " XML_CIB_TAG_NODES ": %s", name, id, pcmk_strerror(rc)); } rc = cib->cmds->remove(cib, XML_CIB_TAG_STATUS, node_state, cib_sync_call); if (rc != pcmk_ok) { printf("Could not remove %s[%ld] from " XML_CIB_TAG_STATUS ": %s", name, id, pcmk_strerror(rc)); } cib->cmds->signoff(cib); cib_delete(cib); return rc; } static int tools_remove_node_cache(const char *node_name, long nodeid, const char *target) { int rc = -1; crm_ipc_t *conn = crm_ipc_new(target, 0); xmlNode *cmd = NULL; if (!conn) { return -ENOTCONN; } if (!crm_ipc_connect(conn)) { crm_perror(LOG_ERR, "Connection to %s failed", target); crm_ipc_destroy(conn); return -ENOTCONN; } if(safe_str_eq(target, CRM_SYSTEM_CRMD)) { // The controller requires a hello message before sending a request rc = send_controller_hello(conn); if (rc < 0) { fprintf(stderr, "error: Could not register with controller: %s\n", pcmk_strerror(rc)); return rc; } } crm_trace("Removing %s[%ld] from the %s membership cache", node_name, nodeid, target); if(safe_str_eq(target, T_ATTRD)) { cmd = create_xml_node(NULL, __FUNCTION__); crm_xml_add(cmd, F_TYPE, T_ATTRD); crm_xml_add(cmd, F_ORIG, crm_system_name); crm_xml_add(cmd, PCMK__XA_TASK, PCMK__ATTRD_CMD_PEER_REMOVE); crm_xml_add(cmd, PCMK__XA_ATTR_NODE_NAME, node_name); if (nodeid > 0) { crm_xml_add_int(cmd, PCMK__XA_ATTR_NODE_ID, (int) nodeid); } } else { cmd = create_request(CRM_OP_RM_NODE_CACHE, NULL, NULL, target, crm_system_name, pid_s); if (nodeid > 0) { crm_xml_set_id(cmd, "%ld", nodeid); } crm_xml_add(cmd, XML_ATTR_UNAME, node_name); } rc = crm_ipc_send(conn, cmd, 0, 0, NULL); crm_debug("%s peer cache cleanup for %s (%ld): %d", target, node_name, nodeid, rc); if (rc > 0) { rc = cib_remove_node(nodeid, node_name); } if (conn) { crm_ipc_close(conn); crm_ipc_destroy(conn); } free_xml(cmd); return rc > 0 ? 0 : rc; } static void remove_node(const char *target_uname) { int d = 0; long nodeid = 0; const char *node_name = NULL; char *endptr = NULL; const char *daemons[] = { CRM_SYSTEM_CRMD, "stonith-ng", T_ATTRD, CRM_SYSTEM_MCP, }; // Check whether node was specified by name or numeric ID errno = 0; nodeid = strtol(target_uname, &endptr, 10); if ((errno != 0) || (endptr == target_uname) || (*endptr != '\0') || (nodeid <= 0)) { // It's not a positive integer, so assume it's a node name nodeid = 0; node_name = target_uname; } for (d = 0; d < DIMOF(daemons); d++) { if (tools_remove_node_cache(node_name, nodeid, daemons[d])) { crm_err("Failed to connect to %s to remove node '%s'", daemons[d], target_uname); crm_node_exit(CRM_EX_ERROR); return; } } crm_node_exit(CRM_EX_OK); } static gint compare_node_xml(gconstpointer a, gconstpointer b) { const char *a_name = crm_element_value((xmlNode*) a, "uname"); const char *b_name = crm_element_value((xmlNode*) b, "uname"); return strcmp((a_name? a_name : ""), (b_name? b_name : "")); } static int node_mcp_dispatch(const char *buffer, ssize_t length, gpointer userdata) { GList *nodes = NULL; xmlNode *node = NULL; xmlNode *msg = string2xml(buffer); const char *uname; const char *state; if (msg == NULL) { fprintf(stderr, "error: Could not understand pacemakerd response\n"); crm_node_exit(CRM_EX_PROTOCOL); return 0; } crm_log_xml_trace(msg, "message"); for (node = __xml_first_child(msg); node != NULL; node = __xml_next(node)) { nodes = g_list_insert_sorted(nodes, node, compare_node_xml); } for (GList *iter = nodes; iter; iter = iter->next) { node = (xmlNode*) iter->data; uname = crm_element_value(node, "uname"); state = crm_element_value(node, "state"); if (options.command == 'l') { int id = 0; crm_element_value_int(node, "id", &id); printf("%d %s %s\n", id, (uname? uname : ""), (state? state : "")); // This is CRM_NODE_MEMBER but we don't want to include cluster header } else if ((options.command == 'p') && safe_str_eq(state, "member")) { printf("%s ", (uname? uname : "")); } } if (options.command == 'p') { fprintf(stdout, "\n"); } free_xml(msg); crm_node_exit(CRM_EX_OK); return 0; } static void run_pacemakerd_mainloop() { crm_ipc_t *ipc = NULL; xmlNode *poke = NULL; ipc = new_mainloop_for_ipc(CRM_SYSTEM_MCP, node_mcp_dispatch); // Sending anything will get us a list of nodes poke = create_xml_node(NULL, "poke"); crm_ipc_send(ipc, poke, 0, 0, NULL); free_xml(poke); // Handle reply via node_mcp_dispatch() run_mainloop_and_exit(); } static GOptionContext * build_arg_context(pcmk__common_args_t *args, GOptionGroup *group) { GOptionContext *context = NULL; GOptionEntry extra_prog_entries[] = { { "quiet", 'Q', 0, G_OPTION_ARG_NONE, &(args->quiet), "Be less descriptive in output.", NULL }, { NULL } }; context = pcmk__build_arg_context(args, NULL, &group); /* Add the -q option, which cannot be part of the globally supported options * because some tools use that flag for something else. */ pcmk__add_main_args(context, extra_prog_entries); pcmk__add_arg_group(context, "commands", "Commands:", "Show command help", command_entries); pcmk__add_arg_group(context, "additional", "Additional Options:", "Show additional options", addl_entries); return context; } int main(int argc, char **argv) { pcmk__common_args_t *args = pcmk__new_common_args(SUMMARY); GError *error = NULL; GOptionContext *context = NULL; GOptionGroup *output_group = NULL; gchar **processed_args = NULL; context = build_arg_context(args, output_group); crm_log_cli_init("crm_node"); processed_args = pcmk__cmdline_preproc(argv, "NR"); if (!g_option_context_parse_strv(context, &processed_args, &error)) { fprintf(stderr, "%s: %s\n", g_get_prgname(), error->message); exit_code = CRM_EX_USAGE; goto done; } for (int i = 0; i < args->verbosity; i++) { crm_bump_log_level(argc, argv); } if (args->version) { /* FIXME: When crm_node is converted to use formatted output, this can go. */ pcmk__cli_help('v', CRM_EX_USAGE); } if (optind > argc || options.command == 0) { fprintf(stderr, "%s", g_option_context_get_help(context, TRUE, NULL)); exit_code = CRM_EX_USAGE; goto done; } if (options.dangerous_cmd && options.force_flag == FALSE) { fprintf(stderr, "The supplied command is considered dangerous." " To prevent accidental destruction of the cluster," " the --force flag is required in order to proceed.\n"); exit_code = CRM_EX_USAGE; goto done; } switch (options.command) { case 'n': print_node_name(); break; case 'R': remove_node(options.target_uname); break; case 'i': case 'q': case 'N': run_controller_mainloop(options.nodeid); break; case 'l': case 'p': run_pacemakerd_mainloop(); break; default: break; } done: g_strfreev(processed_args); g_clear_error(&error); pcmk__free_arg_context(context); crm_node_exit(exit_code); return exit_code; } diff --git a/tools/crm_resource.c b/tools/crm_resource.c index 82aaadce03..8f0c7c1d77 100644 --- a/tools/crm_resource.c +++ b/tools/crm_resource.c @@ -1,1558 +1,1558 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include bool BE_QUIET = FALSE; bool scope_master = FALSE; int cib_options = cib_sync_call; static GMainLoop *mainloop = NULL; // Things that should be cleaned up on exit static cib_t *cib_conn = NULL; static pcmk_controld_api_t *controld_api = NULL; static pe_working_set_t *data_set = NULL; #define MESSAGE_TIMEOUT_S 60 // Clean up and exit static void bye(crm_exit_t exit_code) { static bool crm_resourece_shutdown_flag = FALSE; if (crm_resourece_shutdown_flag) { return; } crm_resourece_shutdown_flag = TRUE; if (cib_conn != NULL) { cib_conn->cmds->signoff(cib_conn); cib_delete(cib_conn); cib_conn = NULL; } if (controld_api != NULL) { pcmk_free_controld_api(controld_api); } pe_free_working_set(data_set); data_set = NULL; crm_exit(exit_code); return; } static gboolean resource_ipc_timeout(gpointer data) { fprintf(stderr, "Aborting because no messages received in %d seconds\n", MESSAGE_TIMEOUT_S); crm_err("No messages received in %d seconds", MESSAGE_TIMEOUT_S); bye(CRM_EX_TIMEOUT); return FALSE; } static void handle_controller_reply(pcmk_controld_api_t *capi, void *api_data, void *user_data) { fprintf(stderr, "."); if ((capi->replies_expected(capi) == 0) && mainloop && g_main_loop_is_running(mainloop)) { fprintf(stderr, " OK\n"); crm_debug("Got all the replies we expected"); bye(CRM_EX_OK); } } static void handle_controller_drop(pcmk_controld_api_t *capi, void *api_data, void *user_data) { crm_info("Connection to controller was terminated"); bye(CRM_EX_DISCONNECT); } static void start_mainloop(pcmk_controld_api_t *capi) { if (capi->replies_expected(capi) > 0) { unsigned int count = capi->replies_expected(capi); fprintf(stderr, "Waiting for %d %s from the controller", count, pcmk__plural_alt(count, "reply", "replies")); mainloop = g_main_loop_new(NULL, FALSE); g_timeout_add(MESSAGE_TIMEOUT_S * 1000, resource_ipc_timeout, NULL); g_main_loop_run(mainloop); } } static int compare_id(gconstpointer a, gconstpointer b) { return strcmp((const char *)a, (const char *)b); } static GListPtr build_constraint_list(xmlNode *root) { GListPtr retval = NULL; xmlNode *cib_constraints = NULL; xmlXPathObjectPtr xpathObj = NULL; int ndx = 0; cib_constraints = get_object_root(XML_CIB_TAG_CONSTRAINTS, root); xpathObj = xpath_search(cib_constraints, "//" XML_CONS_TAG_RSC_LOCATION); for (ndx = 0; ndx < numXpathResults(xpathObj); ndx++) { xmlNode *match = getXpathResult(xpathObj, ndx); retval = g_list_insert_sorted(retval, (gpointer) ID(match), compare_id); } freeXpathObject(xpathObj); return retval; } /* short option letters still available: eEJkKXyYZ */ static pcmk__cli_option_t long_options[] = { // long option, argument type, storage, short option, description, flags { "help", no_argument, NULL, '?', "\t\tDisplay this text and exit", pcmk__option_default }, { "version", no_argument, NULL, '$', "\t\tDisplay version information and exit", pcmk__option_default }, { "verbose", no_argument, NULL, 'V', "\t\tIncrease debug output (may be specified multiple times)", pcmk__option_default }, { "quiet", no_argument, NULL, 'Q', "\t\tBe less descriptive in results", pcmk__option_default }, { "resource", required_argument, NULL, 'r', "\tResource ID", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\nQueries:", pcmk__option_default }, { "list", no_argument, NULL, 'L', "\t\tList all cluster resources with status", pcmk__option_default }, { "list-raw", no_argument, NULL, 'l', "\t\tList IDs of all instantiated resources (individual members rather " "than groups etc.)", pcmk__option_default }, { "list-cts", no_argument, NULL, 'c', NULL, pcmk__option_hidden }, { "list-operations", no_argument, NULL, 'O', "\tList active resource operations, optionally filtered by --resource " "and/or --node", pcmk__option_default }, { "list-all-operations", no_argument, NULL, 'o', "List all resource operations, optionally filtered by --resource " "and/or --node", pcmk__option_default }, { "list-standards", no_argument, NULL, 0, "\tList supported standards", pcmk__option_default }, { "list-ocf-providers", no_argument, NULL, 0, "List all available OCF providers", pcmk__option_default }, { "list-agents", required_argument, NULL, 0, "List all agents available for the named standard and/or provider", pcmk__option_default }, { "list-ocf-alternatives", required_argument, NULL, 0, "List all available providers for the named OCF agent", pcmk__option_default }, { "show-metadata", required_argument, NULL, 0, "Show the metadata for the named class:provider:agent", pcmk__option_default }, { "query-xml", no_argument, NULL, 'q', "\tShow XML configuration of resource (after any template expansion)", pcmk__option_default }, { "query-xml-raw", no_argument, NULL, 'w', "\tShow XML configuration of resource (before any template expansion)", pcmk__option_default }, { "get-parameter", required_argument, NULL, 'g', "Display named parameter for resource (use instance attribute unless " "--meta or --utilization is specified)", pcmk__option_default }, { "get-property", required_argument, NULL, 'G', "Display named property of resource ('class', 'type', or 'provider') " "(requires --resource)", pcmk__option_hidden }, { "locate", no_argument, NULL, 'W', "\t\tShow node(s) currently running resource", pcmk__option_default }, { "stack", no_argument, NULL, 'A', "\t\tDisplay the prerequisites and dependents of a resource", pcmk__option_default }, { "constraints", no_argument, NULL, 'a', "\tDisplay the (co)location constraints that apply to a resource", pcmk__option_default }, { "why", no_argument, NULL, 'Y', "\t\tShow why resources are not running, optionally filtered by " "--resource and/or --node", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\nCommands:", pcmk__option_default }, { "validate", no_argument, NULL, 0, "\t\tValidate resource configuration by calling agent's validate-all " "action. The configuration may be specified either by giving an " "existing resource name with -r, or by specifying --class, " "--agent, and --provider arguments, along with any number of " "--option arguments.", pcmk__option_default }, { "cleanup", no_argument, NULL, 'C', "\t\tIf resource has any past failures, clear its history and " "fail count. Optionally filtered by --resource, --node, " "--operation, and --interval (otherwise all). --operation and " "--interval apply to fail counts, but entire history is always " "cleared, to allow current state to be rechecked. If the named " "resource is part of a group, or one numbered instance of a clone " "or bundled resource, the clean-up applies to the whole collective " "resource unless --force is given.", pcmk__option_default }, { "refresh", no_argument, NULL, 'R', "\t\tDelete resource's history (including failures) so its current " "state is rechecked. Optionally filtered by --resource and --node " "(otherwise all). If the named resource is part of a group, or one " "numbered instance of a clone or bundled resource, the clean-up " "applies to the whole collective resource unless --force is given.", pcmk__option_default }, { "set-parameter", required_argument, NULL, 'p', "Set named parameter for resource (requires -v). Use instance " "attribute unless --meta or --utilization is specified.", pcmk__option_default }, { "delete-parameter", required_argument, NULL, 'd', "Delete named parameter for resource. Use instance attribute unless " "--meta or --utilization is specified.", pcmk__option_default }, { "set-property", required_argument, NULL, 'S', "Set named property of resource ('class', 'type', or 'provider') " "(requires -r, -t, -v)", pcmk__option_hidden }, { "-spacer-", no_argument, NULL, '-', "\nResource location:", pcmk__option_default }, { "move", no_argument, NULL, 'M', "\t\tCreate a constraint to move resource. If --node is specified, the " "constraint will be to move to that node, otherwise it will be to " "ban the current node. Unless --force is specified, this will " "return an error if the resource is already running on the " "specified node. If --force is specified, this will always ban the " "current node. Optional: --lifetime, --master. NOTE: This may " "prevent the resource from running on its previous location until " "the implicit constraint expires or is removed with --clear.", pcmk__option_default }, { "ban", no_argument, NULL, 'B', "\t\tCreate a constraint to keep resource off a node. Optional: " "--node, --lifetime, --master. NOTE: This will prevent the " "resource from running on the affected node until the implicit " "constraint expires or is removed with --clear. If --node is not " "specified, it defaults to the node currently running the resource " "for primitives and groups, or the master for promotable clones " "with promoted-max=1 (all other situations result in an error as " "there is no sane default).", pcmk__option_default }, { "clear", no_argument, NULL, 'U', "\t\tRemove all constraints created by the --ban and/or --move " "commands. Requires: --resource. Optional: --node, --master, " "--expired. If --node is not specified, all constraints created " "by --ban and --move will be removed for the named resource. If " "--node and --force are specified, any constraint created by " "--move will be cleared, even if it is not for the specified node. " "If --expired is specified, only those constraints whose lifetimes " "have expired will be removed.", pcmk__option_default }, { "expired", no_argument, NULL, 'e', "\t\tModifies the --clear argument to remove constraints with " "expired lifetimes.", pcmk__option_default }, { "lifetime", required_argument, NULL, 'u', "\tLifespan (as ISO 8601 duration) of created constraints (with -B, " "-M) (see https://en.wikipedia.org/wiki/ISO_8601#Durations)", pcmk__option_default }, { "master", no_argument, NULL, 0, "\t\tLimit scope of command to Master role (with -B, -M, -U). For -B " "and -M, the previous master may remain active in the Slave role.", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\nAdvanced Commands:", pcmk__option_default }, { "delete", no_argument, NULL, 'D', "\t\t(Advanced) Delete a resource from the CIB. Required: -t", pcmk__option_default }, { "fail", no_argument, NULL, 'F', "\t\t(Advanced) Tell the cluster this resource has failed", pcmk__option_default }, { "restart", no_argument, NULL, 0, "\t\t(Advanced) Tell the cluster to restart this resource and " "anything that depends on it", pcmk__option_default }, { "wait", no_argument, NULL, 0, "\t\t(Advanced) Wait until the cluster settles into a stable state", pcmk__option_default }, { "force-demote", no_argument, NULL, 0, "\t(Advanced) Bypass the cluster and demote a resource on the local " "node. Unless --force is specified, this will refuse to do so if " "the cluster believes the resource is a clone instance already " "running on the local node.", pcmk__option_default }, { "force-stop", no_argument, NULL, 0, "\t(Advanced) Bypass the cluster and stop a resource on the local node", pcmk__option_default }, { "force-start", no_argument, NULL, 0, "\t(Advanced) Bypass the cluster and start a resource on the local " "node. Unless --force is specified, this will refuse to do so if " "the cluster believes the resource is a clone instance already " "running on the local node.", pcmk__option_default }, { "force-promote", no_argument, NULL, 0, "\t(Advanced) Bypass the cluster and promote a resource on the local " "node. Unless --force is specified, this will refuse to do so if " "the cluster believes the resource is a clone instance already " "running on the local node.", pcmk__option_default }, { "force-check", no_argument, NULL, 0, "\t(Advanced) Bypass the cluster and check the state of a resource on " "the local node", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\nValidate Options:", pcmk__option_default }, { "class", required_argument, NULL, 0, "\tThe standard the resource agent confirms to (for example, ocf). " "Use with --agent, --provider, --option, and --validate.", pcmk__option_default }, { "agent", required_argument, NULL, 0, "\tThe agent to use (for example, IPaddr). Use with --class, " "--provider, --option, and --validate.", pcmk__option_default }, { "provider", required_argument, NULL, 0, "\tThe vendor that supplies the resource agent (for example, " "heartbeat). Use with --class, --agent, --option, and --validate.", pcmk__option_default }, { "option", required_argument, NULL, 0, "\tSpecify a device configuration parameter as NAME=VALUE (may be " "specified multiple times). Use with --validate and without the " "-r option.", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\nAdditional Options:", pcmk__option_default }, { "node", required_argument, NULL, 'N', "\tNode name", pcmk__option_default }, { "recursive", no_argument, NULL, 0, "\tFollow colocation chains when using --set-parameter", pcmk__option_default }, { "resource-type", required_argument, NULL, 't', "Resource XML element (primitive, group, etc.) (with -D)", pcmk__option_default }, { "parameter-value", required_argument, NULL, 'v', "Value to use with -p", pcmk__option_default }, { "meta", no_argument, NULL, 'm', "\t\tUse resource meta-attribute instead of instance attribute " "(with -p, -g, -d)", pcmk__option_default }, { "utilization", no_argument, NULL, 'z', "\tUse resource utilization attribute instead of instance attribute " "(with -p, -g, -d)", pcmk__option_default }, { "operation", required_argument, NULL, 'n', "\tOperation to clear instead of all (with -C -r)", pcmk__option_default }, { "interval", required_argument, NULL, 'I', "\tInterval of operation to clear (default 0) (with -C -r -n)", pcmk__option_default }, { "set-name", required_argument, NULL, 's', "\t(Advanced) XML ID of attributes element to use (with -p, -d)", pcmk__option_default }, { "nvpair", required_argument, NULL, 'i', "\t(Advanced) XML ID of nvpair element to use (with -p, -d)", pcmk__option_default }, { "timeout", required_argument, NULL, 'T', "\t(Advanced) Abort if command does not finish in this time (with " "--restart, --wait, --force-*)", pcmk__option_default }, { "force", no_argument, NULL, 'f', "\t\tIf making CIB changes, do so regardless of quorum. See help for " "individual commands for additional behavior.", pcmk__option_default }, { "xml-file", required_argument, NULL, 'x', NULL, pcmk__option_hidden }, /* legacy options */ { "host-uname", required_argument, NULL, 'H', NULL, pcmk__option_hidden }, { "-spacer-", 1, NULL, '-', "\nExamples:", pcmk__option_paragraph }, { "-spacer-", 1, NULL, '-', "List the available OCF agents:", pcmk__option_paragraph }, { "-spacer-", 1, NULL, '-', " crm_resource --list-agents ocf", pcmk__option_example }, { "-spacer-", 1, NULL, '-', "List the available OCF agents from the linux-ha project:", pcmk__option_paragraph }, { "-spacer-", 1, NULL, '-', " crm_resource --list-agents ocf:heartbeat", pcmk__option_example }, { "-spacer-", 1, NULL, '-', "Move 'myResource' to a specific node:", pcmk__option_paragraph }, { "-spacer-", 1, NULL, '-', " crm_resource --resource myResource --move --node altNode", pcmk__option_example }, { "-spacer-", 1, NULL, '-', "Allow (but not force) 'myResource' to move back to its original " "location:", pcmk__option_paragraph }, { "-spacer-", 1, NULL, '-', " crm_resource --resource myResource --clear", pcmk__option_example }, { "-spacer-", 1, NULL, '-', "Stop 'myResource' (and anything that depends on it):", pcmk__option_paragraph }, { "-spacer-", 1, NULL, '-', " crm_resource --resource myResource --set-parameter target-role " "--meta --parameter-value Stopped", pcmk__option_example }, { "-spacer-", 1, NULL, '-', "Tell the cluster not to manage 'myResource' (the cluster will not " "attempt to start or stop the resource under any circumstances; " "useful when performing maintenance tasks on a resource):", pcmk__option_paragraph }, { "-spacer-", 1, NULL, '-', " crm_resource --resource myResource --set-parameter is-managed " "--meta --parameter-value false", pcmk__option_example }, { "-spacer-", 1, NULL, '-', "Erase the operation history of 'myResource' on 'aNode' (the cluster " "will 'forget' the existing resource state, including any " "errors, and attempt to recover the resource; useful when a " "resource had failed permanently and has been repaired " "by an administrator):", pcmk__option_paragraph }, { "-spacer-", 1, NULL, '-', " crm_resource --resource myResource --cleanup --node aNode", pcmk__option_example }, { 0, 0, 0, 0 } }; int main(int argc, char **argv) { char rsc_cmd = 'L'; const char *v_class = NULL; const char *v_agent = NULL; const char *v_provider = NULL; char *name = NULL; char *value = NULL; GHashTable *validate_options = NULL; const char *rsc_id = NULL; const char *host_uname = NULL; const char *prop_name = NULL; const char *prop_value = NULL; const char *rsc_type = NULL; const char *prop_id = NULL; const char *prop_set = NULL; const char *rsc_long_cmd = NULL; const char *longname = NULL; const char *operation = NULL; const char *interval_spec = NULL; const char *cib_file = getenv("CIB_file"); GHashTable *override_params = NULL; char *xml_file = NULL; xmlNode *cib_xml_copy = NULL; resource_t *rsc = NULL; bool recursive = FALSE; bool validate_cmdline = FALSE; /* whether we are just validating based on command line options */ bool require_resource = TRUE; /* whether command requires that resource be specified */ bool require_dataset = TRUE; /* whether command requires populated dataset instance */ bool require_crmd = FALSE; // whether command requires controller connection bool clear_expired = FALSE; int rc = pcmk_ok; int is_ocf_rc = 0; int option_index = 0; int timeout_ms = 0; int argerr = 0; int flag; int find_flags = 0; // Flags to use when searching for resource crm_exit_t exit_code = CRM_EX_OK; crm_log_cli_init("crm_resource"); pcmk__set_cli_options(NULL, "| [options]", long_options, "perform tasks related to Pacemaker " "cluster resources"); validate_options = crm_str_table_new(); while (1) { flag = pcmk__next_cli_option(argc, argv, &option_index, &longname); if (flag == -1) break; switch (flag) { case 0: /* long options with no short equivalent */ if (safe_str_eq("master", longname)) { scope_master = TRUE; } else if(safe_str_eq(longname, "recursive")) { recursive = TRUE; } else if (safe_str_eq("wait", longname)) { rsc_cmd = flag; rsc_long_cmd = longname; require_resource = FALSE; require_dataset = FALSE; } else if ( safe_str_eq("validate", longname) || safe_str_eq("restart", longname) || safe_str_eq("force-demote", longname) || safe_str_eq("force-stop", longname) || safe_str_eq("force-start", longname) || safe_str_eq("force-promote", longname) || safe_str_eq("force-check", longname)) { rsc_cmd = flag; rsc_long_cmd = longname; find_flags = pe_find_renamed|pe_find_anon; crm_log_args(argc, argv); } else if (safe_str_eq("list-ocf-providers", longname) || safe_str_eq("list-ocf-alternatives", longname) || safe_str_eq("list-standards", longname)) { const char *text = NULL; lrmd_list_t *list = NULL; lrmd_list_t *iter = NULL; lrmd_t *lrmd_conn = lrmd_api_new(); if (safe_str_eq("list-ocf-providers", longname) || safe_str_eq("list-ocf-alternatives", longname)) { rc = lrmd_conn->cmds->list_ocf_providers(lrmd_conn, optarg, &list); text = "OCF providers"; } else if (safe_str_eq("list-standards", longname)) { rc = lrmd_conn->cmds->list_standards(lrmd_conn, &list); text = "standards"; } if (rc > 0) { for (iter = list; iter != NULL; iter = iter->next) { printf("%s\n", iter->val); } lrmd_list_freeall(list); } else if (optarg) { fprintf(stderr, "No %s found for %s\n", text, optarg); exit_code = CRM_EX_NOSUCH; } else { fprintf(stderr, "No %s found\n", text); exit_code = CRM_EX_NOSUCH; } lrmd_api_delete(lrmd_conn); bye(exit_code); } else if (safe_str_eq("show-metadata", longname)) { char *standard = NULL; char *provider = NULL; char *type = NULL; char *metadata = NULL; lrmd_t *lrmd_conn = lrmd_api_new(); rc = crm_parse_agent_spec(optarg, &standard, &provider, &type); if (rc == pcmk_ok) { rc = lrmd_conn->cmds->get_metadata(lrmd_conn, standard, provider, type, &metadata, 0); } else { fprintf(stderr, "'%s' is not a valid agent specification\n", optarg); rc = -ENXIO; } if (metadata) { printf("%s\n", metadata); } else { fprintf(stderr, "Metadata query for %s failed: %s\n", optarg, pcmk_strerror(rc)); exit_code = crm_errno2exit(rc); } lrmd_api_delete(lrmd_conn); bye(exit_code); } else if (safe_str_eq("list-agents", longname)) { lrmd_list_t *list = NULL; lrmd_list_t *iter = NULL; char *provider = strchr (optarg, ':'); lrmd_t *lrmd_conn = lrmd_api_new(); if (provider) { *provider++ = 0; } rc = lrmd_conn->cmds->list_agents(lrmd_conn, &list, optarg, provider); if (rc > 0) { for (iter = list; iter != NULL; iter = iter->next) { printf("%s\n", iter->val); } lrmd_list_freeall(list); } else { fprintf(stderr, "No agents found for standard=%s, provider=%s\n", optarg, (provider? provider : "*")); exit_code = CRM_EX_NOSUCH; } lrmd_api_delete(lrmd_conn); bye(exit_code); } else if (safe_str_eq("class", longname)) { if (!(pcmk_get_ra_caps(optarg) & pcmk_ra_cap_params)) { if (BE_QUIET == FALSE) { fprintf(stdout, "Standard %s does not support parameters\n", optarg); } bye(exit_code); } else { v_class = optarg; } validate_cmdline = TRUE; require_resource = FALSE; } else if (safe_str_eq("agent", longname)) { validate_cmdline = TRUE; require_resource = FALSE; v_agent = optarg; } else if (safe_str_eq("provider", longname)) { validate_cmdline = TRUE; require_resource = FALSE; v_provider = optarg; } else if (safe_str_eq("option", longname)) { crm_info("Scanning: --option %s", optarg); rc = pcmk_scan_nvpair(optarg, &name, &value); if (rc != 2) { fprintf(stderr, "Invalid option: --option %s: %s", optarg, pcmk_strerror(rc)); argerr++; } else { crm_info("Got: '%s'='%s'", name, value); } g_hash_table_replace(validate_options, name, value); } else { crm_err("Unhandled long option: %s", longname); } break; case 'V': resource_verbose++; crm_bump_log_level(argc, argv); break; case '$': case '?': pcmk__cli_help(flag, CRM_EX_OK); break; case 'x': xml_file = strdup(optarg); break; case 'Q': BE_QUIET = TRUE; break; case 'm': attr_set_type = XML_TAG_META_SETS; break; case 'z': attr_set_type = XML_TAG_UTILIZATION; break; case 'u': move_lifetime = strdup(optarg); break; case 'f': do_force = TRUE; crm_log_args(argc, argv); break; case 'i': prop_id = optarg; break; case 's': prop_set = optarg; break; case 'r': rsc_id = optarg; break; case 'v': prop_value = optarg; break; case 't': rsc_type = optarg; break; case 'T': timeout_ms = crm_get_msec(optarg); break; case 'e': clear_expired = TRUE; require_resource = FALSE; break; case 'C': case 'R': crm_log_args(argc, argv); require_resource = FALSE; if (cib_file == NULL) { require_crmd = TRUE; } rsc_cmd = flag; find_flags = pe_find_renamed|pe_find_anon; break; case 'n': operation = optarg; break; case 'I': interval_spec = optarg; break; case 'D': require_dataset = FALSE; crm_log_args(argc, argv); rsc_cmd = flag; find_flags = pe_find_renamed|pe_find_any; break; case 'F': require_crmd = TRUE; crm_log_args(argc, argv); rsc_cmd = flag; break; case 'U': case 'B': case 'M': crm_log_args(argc, argv); rsc_cmd = flag; find_flags = pe_find_renamed|pe_find_anon; break; case 'c': case 'L': case 'l': case 'O': case 'o': require_resource = FALSE; rsc_cmd = flag; break; case 'Y': require_resource = FALSE; rsc_cmd = flag; find_flags = pe_find_renamed|pe_find_anon; break; case 'q': case 'w': rsc_cmd = flag; find_flags = pe_find_renamed|pe_find_any; break; case 'W': case 'A': case 'a': rsc_cmd = flag; find_flags = pe_find_renamed|pe_find_anon; break; case 'S': require_dataset = FALSE; crm_log_args(argc, argv); prop_name = optarg; rsc_cmd = flag; find_flags = pe_find_renamed|pe_find_any; break; case 'p': case 'd': crm_log_args(argc, argv); prop_name = optarg; rsc_cmd = flag; find_flags = pe_find_renamed|pe_find_any; break; case 'G': case 'g': prop_name = optarg; rsc_cmd = flag; find_flags = pe_find_renamed|pe_find_any; break; case 'H': case 'N': crm_trace("Option %c => %s", flag, optarg); host_uname = optarg; break; default: CMD_ERR("Argument code 0%o (%c) is not (?yet?) supported", flag, flag); ++argerr; break; } } // Catch the case where the user didn't specify a command if (rsc_cmd == 'L') { require_resource = FALSE; } // --expired without --clear/-U doesn't make sense if (clear_expired == TRUE && rsc_cmd != 'U') { CMD_ERR("--expired requires --clear or -U"); argerr++; } if (optind < argc && argv[optind] != NULL && rsc_cmd == 0 && rsc_long_cmd) { override_params = crm_str_table_new(); while (optind < argc && argv[optind] != NULL) { char *name = calloc(1, strlen(argv[optind])); char *value = calloc(1, strlen(argv[optind])); int rc = sscanf(argv[optind], "%[^=]=%s", name, value); if(rc == 2) { g_hash_table_replace(override_params, name, value); } else { CMD_ERR("Error parsing '%s' as a name=value pair for --%s", argv[optind], rsc_long_cmd); free(value); free(name); argerr++; } optind++; } } else if (optind < argc && argv[optind] != NULL && rsc_cmd == 0) { CMD_ERR("non-option ARGV-elements: "); while (optind < argc && argv[optind] != NULL) { CMD_ERR("[%d of %d] %s ", optind, argc, argv[optind]); optind++; argerr++; } } if (optind > argc) { ++argerr; } // Sanity check validating from command line parameters. If everything checks out, // go ahead and run the validation. This way we don't need a CIB connection. if (validate_cmdline == TRUE) { // -r cannot be used with any of --class, --agent, or --provider if (rsc_id != NULL) { CMD_ERR("--resource cannot be used with --class, --agent, and --provider"); argerr++; // If --class, --agent, or --provider are given, --validate must also be given. } else if (!safe_str_eq(rsc_long_cmd, "validate")) { CMD_ERR("--class, --agent, and --provider require --validate"); argerr++; // Not all of --class, --agent, and --provider need to be given. Not all // classes support the concept of a provider. Check that what we were given // is valid. } else if (crm_str_eq(v_class, "stonith", TRUE)) { if (v_provider != NULL) { CMD_ERR("stonith does not support providers"); argerr++; } else if (stonith_agent_exists(v_agent, 0) == FALSE) { CMD_ERR("%s is not a known stonith agent", v_agent ? v_agent : ""); argerr++; } } else if (resources_agent_exists(v_class, v_provider, v_agent) == FALSE) { CMD_ERR("%s:%s:%s is not a known resource", v_class ? v_class : "", v_provider ? v_provider : "", v_agent ? v_agent : ""); argerr++; } if (argerr == 0) { rc = cli_resource_execute_from_params("test", v_class, v_provider, v_agent, "validate-all", validate_options, override_params, timeout_ms); exit_code = crm_errno2exit(rc); bye(exit_code); } } if (argerr) { CMD_ERR("Invalid option(s) supplied, use --help for valid usage"); bye(CRM_EX_USAGE); } if (do_force) { crm_debug("Forcing..."); cib_options |= cib_quorum_override; } if (require_resource && !rsc_id) { CMD_ERR("Must supply a resource id with -r"); rc = -ENXIO; goto bail; } if (find_flags && rsc_id) { require_dataset = TRUE; } /* Establish a connection to the CIB manager */ cib_conn = cib_new(); rc = cib_conn->cmds->signon(cib_conn, crm_system_name, cib_command); if (rc != pcmk_ok) { CMD_ERR("Error connecting to the CIB manager: %s", pcmk_strerror(rc)); goto bail; } /* Populate working set from XML file if specified or CIB query otherwise */ if (require_dataset) { if (xml_file != NULL) { cib_xml_copy = filename2xml(xml_file); } else { rc = cib_conn->cmds->query(cib_conn, NULL, &cib_xml_copy, cib_scope_local | cib_sync_call); } if(rc != pcmk_ok) { goto bail; } /* Populate the working set instance */ data_set = pe_new_working_set(); if (data_set == NULL) { rc = -ENOMEM; goto bail; } set_bit(data_set->flags, pe_flag_no_counts); set_bit(data_set->flags, pe_flag_no_compat); rc = update_working_set_xml(data_set, &cib_xml_copy); if (rc != pcmk_ok) { goto bail; } cluster_status(data_set); } // If command requires that resource exist if specified, find it if (find_flags && rsc_id) { rsc = pe_find_resource_with_flags(data_set->resources, rsc_id, find_flags); if (rsc == NULL) { CMD_ERR("Resource '%s' not found", rsc_id); rc = -ENXIO; goto bail; } } // Establish a connection to the controller if needed if (require_crmd) { char *client_uuid; pcmk_controld_api_cb_t dispatch_cb = { handle_controller_reply, NULL }; pcmk_controld_api_cb_t destroy_cb = { handle_controller_drop, NULL }; - client_uuid = crm_getpid_s(); + client_uuid = pcmk__getpid_s(); controld_api = pcmk_new_controld_api(crm_system_name, client_uuid); free(client_uuid); rc = controld_api->connect(controld_api, true, &dispatch_cb, &destroy_cb); if (rc != pcmk_rc_ok) { CMD_ERR("Error connecting to the controller: %s", pcmk_rc_str(rc)); rc = pcmk_rc2legacy(rc); goto bail; } } /* Handle rsc_cmd appropriately */ if (rsc_cmd == 'L') { rc = pcmk_ok; cli_resource_print_list(data_set, FALSE); } else if (rsc_cmd == 'l') { int found = 0; GListPtr lpc = NULL; rc = pcmk_ok; for (lpc = data_set->resources; lpc != NULL; lpc = lpc->next) { rsc = (resource_t *) lpc->data; found++; cli_resource_print_raw(rsc); } if (found == 0) { printf("NO resources configured\n"); rc = -ENXIO; } } else if (rsc_cmd == 0 && rsc_long_cmd && safe_str_eq(rsc_long_cmd, "restart")) { /* We don't pass data_set because rsc needs to stay valid for the entire * lifetime of cli_resource_restart(), but it will reset and update the * working set multiple times, so it needs to use its own copy. */ rc = cli_resource_restart(rsc, host_uname, timeout_ms, cib_conn); } else if (rsc_cmd == 0 && rsc_long_cmd && safe_str_eq(rsc_long_cmd, "wait")) { rc = wait_till_stable(timeout_ms, cib_conn); } else if (rsc_cmd == 0 && rsc_long_cmd) { // validate, force-(stop|start|demote|promote|check) rc = cli_resource_execute(rsc, rsc_id, rsc_long_cmd, override_params, timeout_ms, cib_conn, data_set); if (rc >= 0) { is_ocf_rc = 1; } } else if (rsc_cmd == 'A' || rsc_cmd == 'a') { GListPtr lpc = NULL; xmlNode *cib_constraints = get_object_root(XML_CIB_TAG_CONSTRAINTS, data_set->input); unpack_constraints(cib_constraints, data_set); // Constraints apply to group/clone, not member/instance rsc = uber_parent(rsc); for (lpc = data_set->resources; lpc != NULL; lpc = lpc->next) { resource_t *r = (resource_t *) lpc->data; clear_bit(r->flags, pe_rsc_allocating); } cli_resource_print_colocation(rsc, TRUE, rsc_cmd == 'A', 1); fprintf(stdout, "* %s\n", rsc->id); cli_resource_print_location(rsc, NULL); for (lpc = data_set->resources; lpc != NULL; lpc = lpc->next) { resource_t *r = (resource_t *) lpc->data; clear_bit(r->flags, pe_rsc_allocating); } cli_resource_print_colocation(rsc, FALSE, rsc_cmd == 'A', 1); } else if (rsc_cmd == 'c') { GListPtr lpc = NULL; rc = pcmk_ok; for (lpc = data_set->resources; lpc != NULL; lpc = lpc->next) { rsc = (resource_t *) lpc->data; cli_resource_print_cts(rsc); } cli_resource_print_cts_constraints(data_set); } else if (rsc_cmd == 'F') { rc = cli_resource_fail(controld_api, host_uname, rsc_id, data_set); if (rc == pcmk_rc_ok) { start_mainloop(controld_api); } rc = pcmk_rc2legacy(rc); } else if (rsc_cmd == 'O') { rc = cli_resource_print_operations(rsc_id, host_uname, TRUE, data_set); } else if (rsc_cmd == 'o') { rc = cli_resource_print_operations(rsc_id, host_uname, FALSE, data_set); } else if (rsc_cmd == 'W') { rc = cli_resource_search(rsc, rsc_id, data_set); if (rc >= 0) { rc = pcmk_ok; } } else if (rsc_cmd == 'q') { rc = cli_resource_print(rsc, data_set, TRUE); } else if (rsc_cmd == 'w') { rc = cli_resource_print(rsc, data_set, FALSE); } else if (rsc_cmd == 'Y') { node_t *dest = NULL; if (host_uname) { dest = pe_find_node(data_set->nodes, host_uname); if (dest == NULL) { rc = -pcmk_err_node_unknown; goto bail; } } cli_resource_why(cib_conn, data_set->resources, rsc, dest); rc = pcmk_ok; } else if (rsc_cmd == 'U') { GListPtr before = NULL; GListPtr after = NULL; GListPtr remaining = NULL; GListPtr ele = NULL; node_t *dest = NULL; if (BE_QUIET == FALSE) { before = build_constraint_list(data_set->input); } if (clear_expired == TRUE) { rc = cli_resource_clear_all_expired(data_set->input, cib_conn, rsc_id, host_uname, scope_master); } else if (host_uname) { dest = pe_find_node(data_set->nodes, host_uname); if (dest == NULL) { rc = -pcmk_err_node_unknown; if (BE_QUIET == FALSE) { g_list_free(before); } goto bail; } rc = cli_resource_clear(rsc_id, dest->details->uname, NULL, cib_conn, TRUE); } else { rc = cli_resource_clear(rsc_id, NULL, data_set->nodes, cib_conn, TRUE); } if (BE_QUIET == FALSE) { rc = cib_conn->cmds->query(cib_conn, NULL, &cib_xml_copy, cib_scope_local | cib_sync_call); if (rc != pcmk_ok) { CMD_ERR("Could not get modified CIB: %s\n", pcmk_strerror(rc)); g_list_free(before); goto bail; } data_set->input = cib_xml_copy; cluster_status(data_set); after = build_constraint_list(data_set->input); remaining = subtract_lists(before, after, (GCompareFunc) strcmp); for (ele = remaining; ele != NULL; ele = ele->next) { printf("Removing constraint: %s\n", (char *) ele->data); } g_list_free(before); g_list_free(after); g_list_free(remaining); } } else if (rsc_cmd == 'M' && host_uname) { rc = cli_resource_move(rsc, rsc_id, host_uname, cib_conn, data_set); } else if (rsc_cmd == 'B' && host_uname) { node_t *dest = pe_find_node(data_set->nodes, host_uname); if (dest == NULL) { rc = -pcmk_err_node_unknown; goto bail; } rc = cli_resource_ban(rsc_id, dest->details->uname, NULL, cib_conn); } else if (rsc_cmd == 'B' || rsc_cmd == 'M') { pe_node_t *current = NULL; unsigned int nactive = 0; current = pe__find_active_requires(rsc, &nactive); if (nactive == 1) { rc = cli_resource_ban(rsc_id, current->details->uname, NULL, cib_conn); } else if (is_set(rsc->flags, pe_rsc_promotable)) { int count = 0; GListPtr iter = NULL; current = NULL; for(iter = rsc->children; iter; iter = iter->next) { resource_t *child = (resource_t *)iter->data; enum rsc_role_e child_role = child->fns->state(child, TRUE); if(child_role == RSC_ROLE_MASTER) { count++; current = pe__current_node(child); } } if(count == 1 && current) { rc = cli_resource_ban(rsc_id, current->details->uname, NULL, cib_conn); } else { rc = -EINVAL; exit_code = CRM_EX_USAGE; CMD_ERR("Resource '%s' not moved: active in %d locations (promoted in %d).", rsc_id, nactive, count); CMD_ERR("To prevent '%s' from running on a specific location, " "specify a node.", rsc_id); CMD_ERR("To prevent '%s' from being promoted at a specific " "location, specify a node and the master option.", rsc_id); } } else { rc = -EINVAL; exit_code = CRM_EX_USAGE; CMD_ERR("Resource '%s' not moved: active in %d locations.", rsc_id, nactive); CMD_ERR("To prevent '%s' from running on a specific location, " "specify a node.", rsc_id); } } else if (rsc_cmd == 'G') { rc = cli_resource_print_property(rsc, prop_name, data_set); } else if (rsc_cmd == 'S') { xmlNode *msg_data = NULL; if ((rsc_type == NULL) || !strlen(rsc_type)) { CMD_ERR("Must specify -t with resource type"); rc = -ENXIO; goto bail; } else if ((prop_value == NULL) || !strlen(prop_value)) { CMD_ERR("Must supply -v with new value"); rc = -EINVAL; goto bail; } CRM_LOG_ASSERT(prop_name != NULL); msg_data = create_xml_node(NULL, rsc_type); crm_xml_add(msg_data, XML_ATTR_ID, rsc_id); crm_xml_add(msg_data, prop_name, prop_value); rc = cib_conn->cmds->modify(cib_conn, XML_CIB_TAG_RESOURCES, msg_data, cib_options); free_xml(msg_data); } else if (rsc_cmd == 'g') { rc = cli_resource_print_attribute(rsc, prop_name, data_set); } else if (rsc_cmd == 'p') { if (prop_value == NULL || strlen(prop_value) == 0) { CMD_ERR("You need to supply a value with the -v option"); rc = -EINVAL; goto bail; } /* coverity[var_deref_model] False positive */ rc = cli_resource_update_attribute(rsc, rsc_id, prop_set, prop_id, prop_name, prop_value, recursive, cib_conn, data_set); } else if (rsc_cmd == 'd') { /* coverity[var_deref_model] False positive */ rc = cli_resource_delete_attribute(rsc, rsc_id, prop_set, prop_id, prop_name, cib_conn, data_set); } else if ((rsc_cmd == 'C') && rsc) { if (do_force == FALSE) { rsc = uber_parent(rsc); } crm_debug("Erasing failures of %s (%s requested) on %s", rsc->id, rsc_id, (host_uname? host_uname: "all nodes")); rc = cli_resource_delete(controld_api, host_uname, rsc, operation, interval_spec, TRUE, data_set); if ((rc == pcmk_ok) && !BE_QUIET) { // Show any reasons why resource might stay stopped cli_resource_check(cib_conn, rsc); } if (rc == pcmk_ok) { start_mainloop(controld_api); } } else if (rsc_cmd == 'C') { rc = cli_cleanup_all(controld_api, host_uname, operation, interval_spec, data_set); if (rc == pcmk_ok) { start_mainloop(controld_api); } } else if ((rsc_cmd == 'R') && rsc) { if (do_force == FALSE) { rsc = uber_parent(rsc); } crm_debug("Re-checking the state of %s (%s requested) on %s", rsc->id, rsc_id, (host_uname? host_uname: "all nodes")); rc = cli_resource_delete(controld_api, host_uname, rsc, NULL, 0, FALSE, data_set); if ((rc == pcmk_ok) && !BE_QUIET) { // Show any reasons why resource might stay stopped cli_resource_check(cib_conn, rsc); } if (rc == pcmk_ok) { start_mainloop(controld_api); } } else if (rsc_cmd == 'R') { const char *router_node = host_uname; int attr_options = pcmk__node_attr_none; if (host_uname) { node_t *node = pe_find_node(data_set->nodes, host_uname); if (pe__is_guest_or_remote_node(node)) { node = pe__current_node(node->details->remote_rsc); if (node == NULL) { CMD_ERR("No cluster connection to Pacemaker Remote node %s detected", host_uname); rc = -ENXIO; goto bail; } router_node = node->details->uname; attr_options |= pcmk__node_attr_remote; } } if (controld_api == NULL) { printf("Dry run: skipping clean-up of %s due to CIB_file\n", host_uname? host_uname : "all nodes"); rc = pcmk_ok; goto bail; } crm_debug("Re-checking the state of all resources on %s", host_uname?host_uname:"all nodes"); rc = pcmk_rc2legacy(pcmk__node_attr_request_clear(NULL, host_uname, NULL, NULL, NULL, NULL, attr_options)); if (controld_api->reprobe(controld_api, host_uname, router_node) == pcmk_rc_ok) { start_mainloop(controld_api); } } else if (rsc_cmd == 'D') { xmlNode *msg_data = NULL; if (rsc_type == NULL) { CMD_ERR("You need to specify a resource type with -t"); rc = -ENXIO; goto bail; } msg_data = create_xml_node(NULL, rsc_type); crm_xml_add(msg_data, XML_ATTR_ID, rsc_id); rc = cib_conn->cmds->remove(cib_conn, XML_CIB_TAG_RESOURCES, msg_data, cib_options); free_xml(msg_data); } else { CMD_ERR("Unknown command: %c", rsc_cmd); } bail: if (is_ocf_rc) { exit_code = rc; } else if (rc != pcmk_ok) { CMD_ERR("Error performing operation: %s", pcmk_strerror(rc)); if (rc == -pcmk_err_no_quorum) { CMD_ERR("To ignore quorum, use the force option"); } if (exit_code == CRM_EX_OK) { exit_code = crm_errno2exit(rc); } } bye(exit_code); } diff --git a/tools/crm_resource_runtime.c b/tools/crm_resource_runtime.c index e7e77c0077..e204e5cfa0 100644 --- a/tools/crm_resource_runtime.c +++ b/tools/crm_resource_runtime.c @@ -1,2089 +1,2089 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include int resource_verbose = 0; bool do_force = FALSE; const char *attr_set_type = XML_TAG_ATTR_SETS; static int do_find_resource(const char *rsc, resource_t * the_rsc, pe_working_set_t * data_set) { int found = 0; GListPtr lpc = NULL; for (lpc = the_rsc->running_on; lpc != NULL; lpc = lpc->next) { node_t *node = (node_t *) lpc->data; if (BE_QUIET) { fprintf(stdout, "%s\n", node->details->uname); } else { const char *state = ""; if (!pe_rsc_is_clone(the_rsc) && the_rsc->fns->state(the_rsc, TRUE) == RSC_ROLE_MASTER) { state = "Master"; } fprintf(stdout, "resource %s is running on: %s %s\n", rsc, node->details->uname, state); } found++; } if (BE_QUIET == FALSE && found == 0) { fprintf(stderr, "resource %s is NOT running\n", rsc); } return found; } int cli_resource_search(resource_t *rsc, const char *requested_name, pe_working_set_t *data_set) { int found = 0; resource_t *parent = uber_parent(rsc); if (pe_rsc_is_clone(rsc)) { for (GListPtr iter = rsc->children; iter != NULL; iter = iter->next) { found += do_find_resource(requested_name, iter->data, data_set); } /* The anonymous clone children's common ID is supplied */ } else if (pe_rsc_is_clone(parent) && is_not_set(rsc->flags, pe_rsc_unique) && rsc->clone_name && safe_str_eq(requested_name, rsc->clone_name) && safe_str_neq(requested_name, rsc->id)) { for (GListPtr iter = parent->children; iter; iter = iter->next) { found += do_find_resource(requested_name, iter->data, data_set); } } else { found += do_find_resource(requested_name, rsc, data_set); } return found; } #define XPATH_MAX 1024 static int find_resource_attr(cib_t * the_cib, const char *attr, const char *rsc, const char *set_type, const char *set_name, const char *attr_id, const char *attr_name, char **value) { int offset = 0; int rc = pcmk_ok; xmlNode *xml_search = NULL; char *xpath_string = NULL; if(value) { *value = NULL; } if(the_cib == NULL) { return -ENOTCONN; } xpath_string = calloc(1, XPATH_MAX); offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "%s", get_object_path("resources")); offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "//*[@id=\"%s\"]", rsc); if (set_type) { offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "/%s", set_type); if (set_name) { offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "[@id=\"%s\"]", set_name); } } offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "//nvpair["); if (attr_id) { offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "@id=\"%s\"", attr_id); } if (attr_name) { if (attr_id) { offset += snprintf(xpath_string + offset, XPATH_MAX - offset, " and "); } offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "@name=\"%s\"", attr_name); } offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "]"); CRM_LOG_ASSERT(offset > 0); rc = the_cib->cmds->query(the_cib, xpath_string, &xml_search, cib_sync_call | cib_scope_local | cib_xpath); if (rc != pcmk_ok) { goto bail; } crm_log_xml_debug(xml_search, "Match"); if (xml_has_children(xml_search)) { xmlNode *child = NULL; rc = -EINVAL; printf("Multiple attributes match name=%s\n", attr_name); for (child = __xml_first_child(xml_search); child != NULL; child = __xml_next(child)) { printf(" Value: %s \t(id=%s)\n", crm_element_value(child, XML_NVPAIR_ATTR_VALUE), ID(child)); } } else if(value) { const char *tmp = crm_element_value(xml_search, attr); if (tmp) { *value = strdup(tmp); } } bail: free(xpath_string); free_xml(xml_search); return rc; } /* PRIVATE. Use the find_matching_attr_resources instead. */ static void find_matching_attr_resources_recursive(GList/* */ ** result, resource_t * rsc, const char * rsc_id, const char * attr_set, const char * attr_id, const char * attr_name, cib_t * cib, const char * cmd, int depth) { int rc = pcmk_ok; char *lookup_id = clone_strip(rsc->id); char *local_attr_id = NULL; /* visit the children */ for(GList *gIter = rsc->children; gIter; gIter = gIter->next) { find_matching_attr_resources_recursive(result, (resource_t*)gIter->data, rsc_id, attr_set, attr_id, attr_name, cib, cmd, depth+1); /* do it only once for clones */ if(pe_clone == rsc->variant) { break; } } rc = find_resource_attr(cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); /* Post-order traversal. * The root is always on the list and it is the last item. */ if((0 == depth) || (pcmk_ok == rc)) { /* push the head */ *result = g_list_append(*result, rsc); } free(local_attr_id); free(lookup_id); } /* The result is a linearized pre-ordered tree of resources. */ static GList/**/ * find_matching_attr_resources(resource_t * rsc, const char * rsc_id, const char * attr_set, const char * attr_id, const char * attr_name, cib_t * cib, const char * cmd) { int rc = pcmk_ok; char *lookup_id = NULL; char *local_attr_id = NULL; GList * result = NULL; /* If --force is used, update only the requested resource (clone or primitive). * Otherwise, if the primitive has the attribute, use that. * Otherwise use the clone. */ if(do_force == TRUE) { return g_list_append(result, rsc); } if(rsc->parent && pe_clone == rsc->parent->variant) { int rc = pcmk_ok; char *local_attr_id = NULL; rc = find_resource_attr(cib, XML_ATTR_ID, rsc_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); free(local_attr_id); if(rc != pcmk_ok) { rsc = rsc->parent; if (BE_QUIET == FALSE) { printf("Performing %s of '%s' on '%s', the parent of '%s'\n", cmd, attr_name, rsc->id, rsc_id); } } return g_list_append(result, rsc); } else if(rsc->parent == NULL && rsc->children && pe_clone == rsc->variant) { resource_t *child = rsc->children->data; if(child->variant == pe_native) { lookup_id = clone_strip(child->id); /* Could be a cloned group! */ rc = find_resource_attr(cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); if(rc == pcmk_ok) { rsc = child; if (BE_QUIET == FALSE) { printf("A value for '%s' already exists in child '%s', performing %s on that instead of '%s'\n", attr_name, lookup_id, cmd, rsc_id); } } free(local_attr_id); free(lookup_id); } return g_list_append(result, rsc); } /* If the resource is a group ==> children inherit the attribute if defined. */ find_matching_attr_resources_recursive(&result, rsc, rsc_id, attr_set, attr_id, attr_name, cib, cmd, 0); return result; } int cli_resource_update_attribute(resource_t *rsc, const char *requested_name, const char *attr_set, const char *attr_id, const char *attr_name, const char *attr_value, bool recursive, cib_t *cib, pe_working_set_t *data_set) { int rc = pcmk_ok; static bool need_init = TRUE; char *local_attr_id = NULL; char *local_attr_set = NULL; GList/**/ *resources = NULL; const char *common_attr_id = attr_id; if(attr_id == NULL && do_force == FALSE && find_resource_attr( cib, XML_ATTR_ID, uber_parent(rsc)->id, NULL, NULL, NULL, attr_name, NULL) == -EINVAL) { printf("\n"); } if (safe_str_eq(attr_set_type, XML_TAG_ATTR_SETS)) { if (do_force == FALSE) { rc = find_resource_attr(cib, XML_ATTR_ID, uber_parent(rsc)->id, XML_TAG_META_SETS, attr_set, attr_id, attr_name, &local_attr_id); if (rc == pcmk_ok && BE_QUIET == FALSE) { printf("WARNING: There is already a meta attribute for '%s' called '%s' (id=%s)\n", uber_parent(rsc)->id, attr_name, local_attr_id); printf(" Delete '%s' first or use the force option to override\n", local_attr_id); } free(local_attr_id); if (rc == pcmk_ok) { return -ENOTUNIQ; } } resources = g_list_append(resources, rsc); } else { resources = find_matching_attr_resources(rsc, requested_name, attr_set, attr_id, attr_name, cib, "update"); } /* If either attr_set or attr_id is specified, * one clearly intends to modify a single resource. * It is the last item on the resource list.*/ for(GList *gIter = (attr_set||attr_id) ? g_list_last(resources) : resources ; gIter; gIter = gIter->next) { char *lookup_id = NULL; xmlNode *xml_top = NULL; xmlNode *xml_obj = NULL; local_attr_id = NULL; local_attr_set = NULL; rsc = (resource_t*)gIter->data; attr_id = common_attr_id; lookup_id = clone_strip(rsc->id); /* Could be a cloned group! */ rc = find_resource_attr(cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); if (rc == pcmk_ok) { crm_debug("Found a match for name=%s: id=%s", attr_name, local_attr_id); attr_id = local_attr_id; } else if (rc != -ENXIO) { free(lookup_id); free(local_attr_id); g_list_free(resources); return rc; } else { const char *tag = crm_element_name(rsc->xml); if (attr_set == NULL) { local_attr_set = crm_strdup_printf("%s-%s", lookup_id, attr_set_type); attr_set = local_attr_set; } if (attr_id == NULL) { local_attr_id = crm_strdup_printf("%s-%s", attr_set, attr_name); attr_id = local_attr_id; } xml_top = create_xml_node(NULL, tag); crm_xml_add(xml_top, XML_ATTR_ID, lookup_id); xml_obj = create_xml_node(xml_top, attr_set_type); crm_xml_add(xml_obj, XML_ATTR_ID, attr_set); } xml_obj = crm_create_nvpair_xml(xml_obj, attr_id, attr_name, attr_value); if (xml_top == NULL) { xml_top = xml_obj; } crm_log_xml_debug(xml_top, "Update"); rc = cib->cmds->modify(cib, XML_CIB_TAG_RESOURCES, xml_top, cib_options); if (rc == pcmk_ok && BE_QUIET == FALSE) { printf("Set '%s' option: id=%s%s%s%s%s value=%s\n", lookup_id, local_attr_id, attr_set ? " set=" : "", attr_set ? attr_set : "", attr_name ? " name=" : "", attr_name ? attr_name : "", attr_value); } free_xml(xml_top); free(lookup_id); free(local_attr_id); free(local_attr_set); if(recursive && safe_str_eq(attr_set_type, XML_TAG_META_SETS)) { GListPtr lpc = NULL; if(need_init) { xmlNode *cib_constraints = get_object_root(XML_CIB_TAG_CONSTRAINTS, data_set->input); need_init = FALSE; unpack_constraints(cib_constraints, data_set); for (lpc = data_set->resources; lpc != NULL; lpc = lpc->next) { resource_t *r = (resource_t *) lpc->data; clear_bit(r->flags, pe_rsc_allocating); } } crm_debug("Looking for dependencies %p", rsc->rsc_cons_lhs); set_bit(rsc->flags, pe_rsc_allocating); for (lpc = rsc->rsc_cons_lhs; lpc != NULL; lpc = lpc->next) { rsc_colocation_t *cons = (rsc_colocation_t *) lpc->data; resource_t *peer = cons->rsc_lh; crm_debug("Checking %s %d", cons->id, cons->score); if (cons->score > 0 && is_not_set(peer->flags, pe_rsc_allocating)) { /* Don't get into colocation loops */ crm_debug("Setting %s=%s for dependent resource %s", attr_name, attr_value, peer->id); cli_resource_update_attribute(peer, peer->id, NULL, NULL, attr_name, attr_value, recursive, cib, data_set); } } } } g_list_free(resources); return rc; } int cli_resource_delete_attribute(resource_t *rsc, const char *requested_name, const char *attr_set, const char *attr_id, const char *attr_name, cib_t *cib, pe_working_set_t *data_set) { int rc = pcmk_ok; GList/**/ *resources = NULL; if(attr_id == NULL && do_force == FALSE && find_resource_attr( cib, XML_ATTR_ID, uber_parent(rsc)->id, NULL, NULL, NULL, attr_name, NULL) == -EINVAL) { printf("\n"); } if(safe_str_eq(attr_set_type, XML_TAG_META_SETS)) { resources = find_matching_attr_resources(rsc, requested_name, attr_set, attr_id, attr_name, cib, "delete"); } else { resources = g_list_append(resources, rsc); } for(GList *gIter = resources; gIter; gIter = gIter->next) { char *lookup_id = NULL; xmlNode *xml_obj = NULL; char *local_attr_id = NULL; rsc = (resource_t*)gIter->data; lookup_id = clone_strip(rsc->id); rc = find_resource_attr(cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); if (rc == -ENXIO) { free(lookup_id); rc = pcmk_ok; continue; } else if (rc != pcmk_ok) { free(lookup_id); g_list_free(resources); return rc; } if (attr_id == NULL) { attr_id = local_attr_id; } xml_obj = crm_create_nvpair_xml(NULL, attr_id, attr_name, NULL); crm_log_xml_debug(xml_obj, "Delete"); CRM_ASSERT(cib); rc = cib->cmds->remove(cib, XML_CIB_TAG_RESOURCES, xml_obj, cib_options); if (rc == pcmk_ok && BE_QUIET == FALSE) { printf("Deleted '%s' option: id=%s%s%s%s%s\n", lookup_id, local_attr_id, attr_set ? " set=" : "", attr_set ? attr_set : "", attr_name ? " name=" : "", attr_name ? attr_name : ""); } free(lookup_id); free_xml(xml_obj); free(local_attr_id); } g_list_free(resources); return rc; } // \return Standard Pacemaker return code static int send_lrm_rsc_op(pcmk_controld_api_t *controld_api, bool do_fail_resource, const char *host_uname, const char *rsc_id, pe_working_set_t *data_set) { const char *router_node = host_uname; const char *rsc_api_id = NULL; const char *rsc_long_id = NULL; const char *rsc_class = NULL; const char *rsc_provider = NULL; const char *rsc_type = NULL; bool cib_only = false; resource_t *rsc = pe_find_resource(data_set->resources, rsc_id); if (rsc == NULL) { CMD_ERR("Resource %s not found", rsc_id); return ENXIO; } else if (rsc->variant != pe_native) { CMD_ERR("We can only process primitive resources, not %s", rsc_id); return EINVAL; } rsc_class = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS); rsc_provider = crm_element_value(rsc->xml, XML_AGENT_ATTR_PROVIDER), rsc_type = crm_element_value(rsc->xml, XML_ATTR_TYPE); if ((rsc_class == NULL) || (rsc_type == NULL)) { CMD_ERR("Resource %s does not have a class and type", rsc_id); return EINVAL; } if (host_uname == NULL) { CMD_ERR("Please specify a node name"); return EINVAL; } else { pe_node_t *node = pe_find_node(data_set->nodes, host_uname); if (node == NULL) { CMD_ERR("Node %s not found", host_uname); return pcmk_rc_node_unknown; } if (!(node->details->online)) { if (do_fail_resource) { CMD_ERR("Node %s is not online", host_uname); return ENOTCONN; } else { cib_only = true; } } if (!cib_only && pe__is_guest_or_remote_node(node)) { node = pe__current_node(node->details->remote_rsc); if (node == NULL) { CMD_ERR("No cluster connection to Pacemaker Remote node %s detected", host_uname); return ENOTCONN; } router_node = node->details->uname; } } if (rsc->clone_name) { rsc_api_id = rsc->clone_name; rsc_long_id = rsc->id; } else { rsc_api_id = rsc->id; } if (do_fail_resource) { return controld_api->fail_resource(controld_api, host_uname, router_node, rsc_api_id, rsc_long_id, rsc_class, rsc_provider, rsc_type); } else { return controld_api->refresh_resource(controld_api, host_uname, router_node, rsc_api_id, rsc_long_id, rsc_class, rsc_provider, rsc_type, cib_only); } } /*! * \internal * \brief Get resource name as used in failure-related node attributes * * \param[in] rsc Resource to check * * \return Newly allocated string containing resource's fail name * \note The caller is responsible for freeing the result. */ static inline char * rsc_fail_name(resource_t *rsc) { const char *name = (rsc->clone_name? rsc->clone_name : rsc->id); return is_set(rsc->flags, pe_rsc_unique)? strdup(name) : clone_strip(name); } // \return Standard Pacemaker return code static int clear_rsc_history(pcmk_controld_api_t *controld_api, const char *host_uname, const char *rsc_id, pe_working_set_t *data_set) { int rc = pcmk_ok; /* Erase the resource's entire LRM history in the CIB, even if we're only * clearing a single operation's fail count. If we erased only entries for a * single operation, we might wind up with a wrong idea of the current * resource state, and we might not re-probe the resource. */ rc = send_lrm_rsc_op(controld_api, false, host_uname, rsc_id, data_set); if (rc != pcmk_rc_ok) { return rc; } crm_trace("Processing %d mainloop inputs", controld_api->replies_expected(controld_api)); while (g_main_context_iteration(NULL, FALSE)) { crm_trace("Processed mainloop input, %d still remaining", controld_api->replies_expected(controld_api)); } return rc; } static int clear_rsc_failures(pcmk_controld_api_t *controld_api, const char *node_name, const char *rsc_id, const char *operation, const char *interval_spec, pe_working_set_t *data_set) { int rc = pcmk_ok; const char *failed_value = NULL; const char *failed_id = NULL; const char *interval_ms_s = NULL; GHashTable *rscs = NULL; GHashTableIter iter; /* Create a hash table to use as a set of resources to clean. This lets us * clean each resource only once (per node) regardless of how many failed * operations it has. */ rscs = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, NULL); // Normalize interval to milliseconds for comparison to history entry if (operation) { interval_ms_s = crm_strdup_printf("%u", crm_parse_interval_spec(interval_spec)); } for (xmlNode *xml_op = __xml_first_child(data_set->failed); xml_op != NULL; xml_op = __xml_next(xml_op)) { failed_id = crm_element_value(xml_op, XML_LRM_ATTR_RSCID); if (failed_id == NULL) { // Malformed history entry, should never happen continue; } // No resource specified means all resources match if (rsc_id) { resource_t *fail_rsc = pe_find_resource_with_flags(data_set->resources, failed_id, pe_find_renamed|pe_find_anon); if (!fail_rsc || safe_str_neq(rsc_id, fail_rsc->id)) { continue; } } // Host name should always have been provided by this point failed_value = crm_element_value(xml_op, XML_ATTR_UNAME); if (safe_str_neq(node_name, failed_value)) { continue; } // No operation specified means all operations match if (operation) { failed_value = crm_element_value(xml_op, XML_LRM_ATTR_TASK); if (safe_str_neq(operation, failed_value)) { continue; } // Interval (if operation was specified) defaults to 0 (not all) failed_value = crm_element_value(xml_op, XML_LRM_ATTR_INTERVAL_MS); if (safe_str_neq(interval_ms_s, failed_value)) { continue; } } /* not available until glib 2.32 g_hash_table_add(rscs, (gpointer) failed_id); */ g_hash_table_insert(rscs, (gpointer) failed_id, (gpointer) failed_id); } g_hash_table_iter_init(&iter, rscs); while (g_hash_table_iter_next(&iter, (gpointer *) &failed_id, NULL)) { crm_debug("Erasing failures of %s on %s", failed_id, node_name); rc = clear_rsc_history(controld_api, node_name, failed_id, data_set); if (rc != pcmk_rc_ok) { return pcmk_rc2legacy(rc); } } g_hash_table_destroy(rscs); return rc; } static int clear_rsc_fail_attrs(resource_t *rsc, const char *operation, const char *interval_spec, node_t *node) { int rc = pcmk_ok; int attr_options = pcmk__node_attr_none; char *rsc_name = rsc_fail_name(rsc); if (pe__is_guest_or_remote_node(node)) { attr_options |= pcmk__node_attr_remote; } rc = pcmk__node_attr_request_clear(NULL, node->details->uname, rsc_name, operation, interval_spec, NULL, attr_options); free(rsc_name); return rc; } int cli_resource_delete(pcmk_controld_api_t *controld_api, const char *host_uname, resource_t *rsc, const char *operation, const char *interval_spec, bool just_failures, pe_working_set_t *data_set) { int rc = pcmk_ok; node_t *node = NULL; if (rsc == NULL) { return -ENXIO; } else if (rsc->children) { GListPtr lpc = NULL; for (lpc = rsc->children; lpc != NULL; lpc = lpc->next) { resource_t *child = (resource_t *) lpc->data; rc = cli_resource_delete(controld_api, host_uname, child, operation, interval_spec, just_failures, data_set); if (rc != pcmk_ok) { return rc; } } return pcmk_ok; } else if (host_uname == NULL) { GListPtr lpc = NULL; GListPtr nodes = g_hash_table_get_values(rsc->known_on); if(nodes == NULL && do_force) { nodes = node_list_dup(data_set->nodes, FALSE, FALSE); } else if(nodes == NULL && rsc->exclusive_discover) { GHashTableIter iter; pe_node_t *node = NULL; g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void**)&node)) { if(node->weight >= 0) { nodes = g_list_prepend(nodes, node); } } } else if(nodes == NULL) { nodes = g_hash_table_get_values(rsc->allowed_nodes); } for (lpc = nodes; lpc != NULL; lpc = lpc->next) { node = (node_t *) lpc->data; if (node->details->online) { rc = cli_resource_delete(controld_api, node->details->uname, rsc, operation, interval_spec, just_failures, data_set); } if (rc != pcmk_ok) { g_list_free(nodes); return rc; } } g_list_free(nodes); return pcmk_ok; } node = pe_find_node(data_set->nodes, host_uname); if (node == NULL) { printf("Unable to clean up %s because node %s not found\n", rsc->id, host_uname); return -ENODEV; } if (!node->details->rsc_discovery_enabled) { printf("Unable to clean up %s because resource discovery disabled on %s\n", rsc->id, host_uname); return -EOPNOTSUPP; } if (controld_api == NULL) { printf("Dry run: skipping clean-up of %s on %s due to CIB_file\n", rsc->id, host_uname); return pcmk_ok; } rc = clear_rsc_fail_attrs(rsc, operation, interval_spec, node); if (rc != pcmk_rc_ok) { printf("Unable to clean up %s failures on %s: %s\n", rsc->id, host_uname, pcmk_rc_str(rc)); return pcmk_rc2legacy(rc); } if (just_failures) { rc = clear_rsc_failures(controld_api, host_uname, rsc->id, operation, interval_spec, data_set); } else { rc = clear_rsc_history(controld_api, host_uname, rsc->id, data_set); rc = pcmk_rc2legacy(rc); } if (rc != pcmk_ok) { printf("Cleaned %s failures on %s, but unable to clean history: %s\n", rsc->id, host_uname, pcmk_strerror(rc)); } else { printf("Cleaned up %s on %s\n", rsc->id, host_uname); } return rc; } int cli_cleanup_all(pcmk_controld_api_t *controld_api, const char *node_name, const char *operation, const char *interval_spec, pe_working_set_t *data_set) { int rc = pcmk_ok; int attr_options = pcmk__node_attr_none; const char *display_name = node_name? node_name : "all nodes"; if (controld_api == NULL) { printf("Dry run: skipping clean-up of %s due to CIB_file\n", display_name); return pcmk_ok; } if (node_name) { node_t *node = pe_find_node(data_set->nodes, node_name); if (node == NULL) { CMD_ERR("Unknown node: %s", node_name); return -ENXIO; } if (pe__is_guest_or_remote_node(node)) { attr_options |= pcmk__node_attr_remote; } } rc = pcmk__node_attr_request_clear(NULL, node_name, NULL, operation, interval_spec, NULL, attr_options); if (rc != pcmk_rc_ok) { printf("Unable to clean up all failures on %s: %s\n", display_name, pcmk_rc_str(rc)); return pcmk_rc2legacy(rc); } if (node_name) { rc = clear_rsc_failures(controld_api, node_name, NULL, operation, interval_spec, data_set); if (rc != pcmk_ok) { printf("Cleaned all resource failures on %s, but unable to clean history: %s\n", node_name, pcmk_strerror(rc)); return rc; } } else { for (GList *iter = data_set->nodes; iter; iter = iter->next) { pe_node_t *node = (pe_node_t *) iter->data; rc = clear_rsc_failures(controld_api, node->details->uname, NULL, operation, interval_spec, data_set); if (rc != pcmk_ok) { printf("Cleaned all resource failures on all nodes, but unable to clean history: %s\n", pcmk_strerror(rc)); return rc; } } } printf("Cleaned up all resources on %s\n", display_name); return pcmk_ok; } void cli_resource_check(cib_t * cib_conn, resource_t *rsc) { bool printed = false; char *role_s = NULL; char *managed = NULL; resource_t *parent = uber_parent(rsc); find_resource_attr(cib_conn, XML_NVPAIR_ATTR_VALUE, parent->id, NULL, NULL, NULL, XML_RSC_ATTR_MANAGED, &managed); find_resource_attr(cib_conn, XML_NVPAIR_ATTR_VALUE, parent->id, NULL, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, &role_s); if(role_s) { enum rsc_role_e role = text2role(role_s); free(role_s); if(role == RSC_ROLE_UNKNOWN) { // Treated as if unset } else if(role == RSC_ROLE_STOPPED) { printf("\n * Configuration specifies '%s' should remain stopped\n", parent->id); printed = true; } else if (is_set(parent->flags, pe_rsc_promotable) && (role == RSC_ROLE_SLAVE)) { printf("\n * Configuration specifies '%s' should not be promoted\n", parent->id); printed = true; } } if (managed && !crm_is_true(managed)) { printf("%s * Configuration prevents cluster from stopping or starting unmanaged '%s'\n", (printed? "" : "\n"), parent->id); printed = true; } free(managed); if (rsc->lock_node) { printf("%s * '%s' is locked to node %s due to shutdown\n", (printed? "" : "\n"), parent->id, rsc->lock_node->details->uname); } if (printed) { printf("\n"); } } // \return Standard Pacemaker return code int cli_resource_fail(pcmk_controld_api_t *controld_api, const char *host_uname, const char *rsc_id, pe_working_set_t *data_set) { crm_notice("Failing %s on %s", rsc_id, host_uname); return send_lrm_rsc_op(controld_api, true, host_uname, rsc_id, data_set); } static GHashTable * generate_resource_params(resource_t * rsc, pe_working_set_t * data_set) { GHashTable *params = NULL; GHashTable *meta = NULL; GHashTable *combined = NULL; GHashTableIter iter; if (!rsc) { crm_err("Resource does not exist in config"); return NULL; } params = crm_str_table_new(); meta = crm_str_table_new(); combined = crm_str_table_new(); get_rsc_attributes(params, rsc, NULL /* TODO: Pass in local node */ , data_set); get_meta_attributes(meta, rsc, NULL /* TODO: Pass in local node */ , data_set); if (params) { char *key = NULL; char *value = NULL; g_hash_table_iter_init(&iter, params); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { g_hash_table_insert(combined, strdup(key), strdup(value)); } g_hash_table_destroy(params); } if (meta) { char *key = NULL; char *value = NULL; g_hash_table_iter_init(&iter, meta); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { char *crm_name = crm_meta_name(key); g_hash_table_insert(combined, crm_name, strdup(value)); } g_hash_table_destroy(meta); } return combined; } static bool resource_is_running_on(resource_t *rsc, const char *host) { bool found = TRUE; GListPtr hIter = NULL; GListPtr hosts = NULL; if(rsc == NULL) { return FALSE; } rsc->fns->location(rsc, &hosts, TRUE); for (hIter = hosts; host != NULL && hIter != NULL; hIter = hIter->next) { pe_node_t *node = (pe_node_t *) hIter->data; if(strcmp(host, node->details->uname) == 0) { crm_trace("Resource %s is running on %s\n", rsc->id, host); goto done; } else if(strcmp(host, node->details->id) == 0) { crm_trace("Resource %s is running on %s\n", rsc->id, host); goto done; } } if(host != NULL) { crm_trace("Resource %s is not running on: %s\n", rsc->id, host); found = FALSE; } else if(host == NULL && hosts == NULL) { crm_trace("Resource %s is not running\n", rsc->id); found = FALSE; } done: g_list_free(hosts); return found; } /*! * \internal * \brief Create a list of all resources active on host from a given list * * \param[in] host Name of host to check whether resources are active * \param[in] rsc_list List of resources to check * * \return New list of resources from list that are active on host */ static GList * get_active_resources(const char *host, GList *rsc_list) { GList *rIter = NULL; GList *active = NULL; for (rIter = rsc_list; rIter != NULL; rIter = rIter->next) { resource_t *rsc = (resource_t *) rIter->data; /* Expand groups to their members, because if we're restarting a member * other than the first, we can't otherwise tell which resources are * stopping and starting. */ if (rsc->variant == pe_group) { active = g_list_concat(active, get_active_resources(host, rsc->children)); } else if (resource_is_running_on(rsc, host)) { active = g_list_append(active, strdup(rsc->id)); } } return active; } GList* subtract_lists(GList *from, GList *items, GCompareFunc cmp) { GList *item = NULL; GList *result = g_list_copy(from); for (item = items; item != NULL; item = item->next) { GList *candidate = NULL; for (candidate = from; candidate != NULL; candidate = candidate->next) { crm_info("Comparing %s with %s", (const char *) candidate->data, (const char *) item->data); if(cmp(candidate->data, item->data) == 0) { result = g_list_remove(result, candidate->data); break; } } } return result; } static void dump_list(GList *items, const char *tag) { int lpc = 0; GList *item = NULL; for (item = items; item != NULL; item = item->next) { crm_trace("%s[%d]: %s", tag, lpc, (char*)item->data); lpc++; } } static void display_list(GList *items, const char *tag) { GList *item = NULL; for (item = items; item != NULL; item = item->next) { fprintf(stdout, "%s%s\n", tag, (const char *)item->data); } } /*! * \internal * \brief Upgrade XML to latest schema version and use it as working set input * * This also updates the working set timestamp to the current time. * * \param[in] data_set Working set instance to update * \param[in] xml XML to use as input * * \return pcmk_ok on success, -ENOKEY if unable to upgrade XML * \note On success, caller is responsible for freeing memory allocated for * data_set->now. * \todo This follows the example of other callers of cli_config_update() * and returns -ENOKEY ("Required key not available") if that fails, * but perhaps -pcmk_err_schema_validation would be better in that case. */ int update_working_set_xml(pe_working_set_t *data_set, xmlNode **xml) { if (cli_config_update(xml, NULL, FALSE) == FALSE) { return -ENOKEY; } data_set->input = *xml; data_set->now = crm_time_new(NULL); return pcmk_ok; } /*! * \internal * \brief Update a working set's XML input based on a CIB query * * \param[in] data_set Data set instance to initialize * \param[in] cib Connection to the CIB manager * * \return pcmk_ok on success, -errno on failure * \note On success, caller is responsible for freeing memory allocated for * data_set->input and data_set->now. */ static int update_working_set_from_cib(pe_working_set_t * data_set, cib_t *cib) { xmlNode *cib_xml_copy = NULL; int rc; rc = cib->cmds->query(cib, NULL, &cib_xml_copy, cib_scope_local | cib_sync_call); if (rc != pcmk_ok) { fprintf(stderr, "Could not obtain the current CIB: %s (%d)\n", pcmk_strerror(rc), rc); return rc; } rc = update_working_set_xml(data_set, &cib_xml_copy); if (rc != pcmk_ok) { fprintf(stderr, "Could not upgrade the current CIB XML\n"); free_xml(cib_xml_copy); return rc; } return pcmk_ok; } static int update_dataset(cib_t *cib, pe_working_set_t * data_set, bool simulate) { char *pid = NULL; char *shadow_file = NULL; cib_t *shadow_cib = NULL; int rc; pe_reset_working_set(data_set); rc = update_working_set_from_cib(data_set, cib); if (rc != pcmk_ok) { return rc; } if(simulate) { - pid = crm_getpid_s(); + pid = pcmk__getpid_s(); shadow_cib = cib_shadow_new(pid); shadow_file = get_shadow_file(pid); if (shadow_cib == NULL) { fprintf(stderr, "Could not create shadow cib: '%s'\n", pid); rc = -ENXIO; goto cleanup; } rc = write_xml_file(data_set->input, shadow_file, FALSE); if (rc < 0) { fprintf(stderr, "Could not populate shadow cib: %s (%d)\n", pcmk_strerror(rc), rc); goto cleanup; } rc = shadow_cib->cmds->signon(shadow_cib, crm_system_name, cib_command); if(rc != pcmk_ok) { fprintf(stderr, "Could not connect to shadow cib: %s (%d)\n", pcmk_strerror(rc), rc); goto cleanup; } pcmk__schedule_actions(data_set, data_set->input, NULL); run_simulation(data_set, shadow_cib, NULL, TRUE); rc = update_dataset(shadow_cib, data_set, FALSE); } else { cluster_status(data_set); } cleanup: /* Do not free data_set->input here, we need rsc->xml to be valid later on */ cib_delete(shadow_cib); free(pid); if(shadow_file) { unlink(shadow_file); free(shadow_file); } return rc; } static int max_delay_for_resource(pe_working_set_t * data_set, resource_t *rsc) { int delay = 0; int max_delay = 0; if(rsc && rsc->children) { GList *iter = NULL; for(iter = rsc->children; iter; iter = iter->next) { resource_t *child = (resource_t *)iter->data; delay = max_delay_for_resource(data_set, child); if(delay > max_delay) { double seconds = delay / 1000.0; crm_trace("Calculated new delay of %.1fs due to %s", seconds, child->id); max_delay = delay; } } } else if(rsc) { char *key = crm_strdup_printf("%s_%s_0", rsc->id, RSC_STOP); action_t *stop = custom_action(rsc, key, RSC_STOP, NULL, TRUE, FALSE, data_set); const char *value = g_hash_table_lookup(stop->meta, XML_ATTR_TIMEOUT); max_delay = value? (int) crm_parse_ll(value, NULL) : -1; pe_free_action(stop); } return max_delay; } static int max_delay_in(pe_working_set_t * data_set, GList *resources) { int max_delay = 0; GList *item = NULL; for (item = resources; item != NULL; item = item->next) { int delay = 0; resource_t *rsc = pe_find_resource(data_set->resources, (const char *)item->data); delay = max_delay_for_resource(data_set, rsc); if(delay > max_delay) { double seconds = delay / 1000.0; crm_trace("Calculated new delay of %.1fs due to %s", seconds, rsc->id); max_delay = delay; } } return 5 + (max_delay / 1000); } #define waiting_for_starts(d, r, h) ((d != NULL) || \ (resource_is_running_on((r), (h)) == FALSE)) /*! * \internal * \brief Restart a resource (on a particular host if requested). * * \param[in] rsc The resource to restart * \param[in] host The host to restart the resource on (or NULL for all) * \param[in] timeout_ms Consider failed if actions do not complete in this time * (specified in milliseconds, but a two-second * granularity is actually used; if 0, a timeout will be * calculated based on the resource timeout) * \param[in] cib Connection to the CIB manager * * \return pcmk_ok on success, -errno on failure (exits on certain failures) */ int cli_resource_restart(pe_resource_t *rsc, const char *host, int timeout_ms, cib_t *cib) { int rc = 0; int lpc = 0; int before = 0; int step_timeout_s = 0; int sleep_interval = 2; int timeout = timeout_ms / 1000; bool stop_via_ban = FALSE; char *rsc_id = NULL; char *orig_target_role = NULL; GList *list_delta = NULL; GList *target_active = NULL; GList *current_active = NULL; GList *restart_target_active = NULL; pe_working_set_t *data_set = NULL; if(resource_is_running_on(rsc, host) == FALSE) { const char *id = rsc->clone_name?rsc->clone_name:rsc->id; if(host) { printf("%s is not running on %s and so cannot be restarted\n", id, host); } else { printf("%s is not running anywhere and so cannot be restarted\n", id); } return -ENXIO; } /* We might set the target-role meta-attribute */ attr_set_type = XML_TAG_META_SETS; rsc_id = strdup(rsc->id); if ((pe_rsc_is_clone(rsc) || pe_bundle_replicas(rsc)) && host) { stop_via_ban = TRUE; } /* grab full cib determine originally active resources disable or ban poll cib and watch for affected resources to get stopped without --timeout, calculate the stop timeout for each step and wait for that if we hit --timeout or the service timeout, re-enable or un-ban, report failure and indicate which resources we couldn't take down if everything stopped, re-enable or un-ban poll cib and watch for affected resources to get started without --timeout, calculate the start timeout for each step and wait for that if we hit --timeout or the service timeout, report (different) failure and indicate which resources we couldn't bring back up report success Optimizations: - use constraints to determine ordered list of affected resources - Allow a --no-deps option (aka. --force-restart) */ data_set = pe_new_working_set(); if (data_set == NULL) { crm_perror(LOG_ERR, "Could not allocate working set"); rc = -ENOMEM; goto done; } set_bit(data_set->flags, pe_flag_no_counts); set_bit(data_set->flags, pe_flag_no_compat); rc = update_dataset(cib, data_set, FALSE); if(rc != pcmk_ok) { fprintf(stdout, "Could not get new resource list: %s (%d)\n", pcmk_strerror(rc), rc); goto done; } restart_target_active = get_active_resources(host, data_set->resources); current_active = get_active_resources(host, data_set->resources); dump_list(current_active, "Origin"); if (stop_via_ban) { /* Stop the clone or bundle instance by banning it from the host */ BE_QUIET = TRUE; rc = cli_resource_ban(rsc_id, host, NULL, cib); } else { /* Stop the resource by setting target-role to Stopped. * Remember any existing target-role so we can restore it later * (though it only makes any difference if it's Slave). */ char *lookup_id = clone_strip(rsc->id); find_resource_attr(cib, XML_NVPAIR_ATTR_VALUE, lookup_id, NULL, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, &orig_target_role); free(lookup_id); rc = cli_resource_update_attribute(rsc, rsc_id, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, RSC_STOPPED, FALSE, cib, data_set); } if(rc != pcmk_ok) { fprintf(stderr, "Could not set target-role for %s: %s (%d)\n", rsc_id, pcmk_strerror(rc), rc); if (current_active) { g_list_free_full(current_active, free); } if (restart_target_active) { g_list_free_full(restart_target_active, free); } goto done; } rc = update_dataset(cib, data_set, TRUE); if(rc != pcmk_ok) { fprintf(stderr, "Could not determine which resources would be stopped\n"); goto failure; } target_active = get_active_resources(host, data_set->resources); dump_list(target_active, "Target"); list_delta = subtract_lists(current_active, target_active, (GCompareFunc) strcmp); fprintf(stdout, "Waiting for %d resources to stop:\n", g_list_length(list_delta)); display_list(list_delta, " * "); step_timeout_s = timeout / sleep_interval; while (list_delta != NULL) { before = g_list_length(list_delta); if(timeout_ms == 0) { step_timeout_s = max_delay_in(data_set, list_delta) / sleep_interval; } /* We probably don't need the entire step timeout */ for(lpc = 0; (lpc < step_timeout_s) && (list_delta != NULL); lpc++) { sleep(sleep_interval); if(timeout) { timeout -= sleep_interval; crm_trace("%ds remaining", timeout); } rc = update_dataset(cib, data_set, FALSE); if(rc != pcmk_ok) { fprintf(stderr, "Could not determine which resources were stopped\n"); goto failure; } if (current_active) { g_list_free_full(current_active, free); } current_active = get_active_resources(host, data_set->resources); g_list_free(list_delta); list_delta = subtract_lists(current_active, target_active, (GCompareFunc) strcmp); dump_list(current_active, "Current"); dump_list(list_delta, "Delta"); } crm_trace("%d (was %d) resources remaining", g_list_length(list_delta), before); if(before == g_list_length(list_delta)) { /* aborted during stop phase, print the contents of list_delta */ fprintf(stderr, "Could not complete shutdown of %s, %d resources remaining\n", rsc_id, g_list_length(list_delta)); display_list(list_delta, " * "); rc = -ETIME; goto failure; } } if (stop_via_ban) { rc = cli_resource_clear(rsc_id, host, NULL, cib, TRUE); } else if (orig_target_role) { rc = cli_resource_update_attribute(rsc, rsc_id, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, orig_target_role, FALSE, cib, data_set); free(orig_target_role); orig_target_role = NULL; } else { rc = cli_resource_delete_attribute(rsc, rsc_id, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, cib, data_set); } if(rc != pcmk_ok) { fprintf(stderr, "Could not unset target-role for %s: %s (%d)\n", rsc_id, pcmk_strerror(rc), rc); goto done; } if (target_active) { g_list_free_full(target_active, free); } target_active = restart_target_active; list_delta = subtract_lists(target_active, current_active, (GCompareFunc) strcmp); fprintf(stdout, "Waiting for %d resources to start again:\n", g_list_length(list_delta)); display_list(list_delta, " * "); step_timeout_s = timeout / sleep_interval; while (waiting_for_starts(list_delta, rsc, host)) { before = g_list_length(list_delta); if(timeout_ms == 0) { step_timeout_s = max_delay_in(data_set, list_delta) / sleep_interval; } /* We probably don't need the entire step timeout */ for (lpc = 0; (lpc < step_timeout_s) && waiting_for_starts(list_delta, rsc, host); lpc++) { sleep(sleep_interval); if(timeout) { timeout -= sleep_interval; crm_trace("%ds remaining", timeout); } rc = update_dataset(cib, data_set, FALSE); if(rc != pcmk_ok) { fprintf(stderr, "Could not determine which resources were started\n"); goto failure; } if (current_active) { g_list_free_full(current_active, free); } /* It's OK if dependent resources moved to a different node, * so we check active resources on all nodes. */ current_active = get_active_resources(NULL, data_set->resources); g_list_free(list_delta); list_delta = subtract_lists(target_active, current_active, (GCompareFunc) strcmp); dump_list(current_active, "Current"); dump_list(list_delta, "Delta"); } if(before == g_list_length(list_delta)) { /* aborted during start phase, print the contents of list_delta */ fprintf(stdout, "Could not complete restart of %s, %d resources remaining\n", rsc_id, g_list_length(list_delta)); display_list(list_delta, " * "); rc = -ETIME; goto failure; } } rc = pcmk_ok; goto done; failure: if (stop_via_ban) { cli_resource_clear(rsc_id, host, NULL, cib, TRUE); } else if (orig_target_role) { cli_resource_update_attribute(rsc, rsc_id, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, orig_target_role, FALSE, cib, data_set); free(orig_target_role); } else { cli_resource_delete_attribute(rsc, rsc_id, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, cib, data_set); } done: if (list_delta) { g_list_free(list_delta); } if (current_active) { g_list_free_full(current_active, free); } if (target_active && (target_active != restart_target_active)) { g_list_free_full(target_active, free); } if (restart_target_active) { g_list_free_full(restart_target_active, free); } free(rsc_id); pe_free_working_set(data_set); return rc; } static inline int action_is_pending(action_t *action) { if(is_set(action->flags, pe_action_optional)) { return FALSE; } else if(is_set(action->flags, pe_action_runnable) == FALSE) { return FALSE; } else if(is_set(action->flags, pe_action_pseudo)) { return FALSE; } else if(safe_str_eq("notify", action->task)) { return FALSE; } return TRUE; } /*! * \internal * \brief Return TRUE if any actions in a list are pending * * \param[in] actions List of actions to check * * \return TRUE if any actions in the list are pending, FALSE otherwise */ static bool actions_are_pending(GListPtr actions) { GListPtr action; for (action = actions; action != NULL; action = action->next) { action_t *a = (action_t *)action->data; if (action_is_pending(a)) { crm_notice("Waiting for %s (flags=0x%.8x)", a->uuid, a->flags); return TRUE; } } return FALSE; } /*! * \internal * \brief Print pending actions to stderr * * \param[in] actions List of actions to check * * \return void */ static void print_pending_actions(GListPtr actions) { GListPtr action; fprintf(stderr, "Pending actions:\n"); for (action = actions; action != NULL; action = action->next) { action_t *a = (action_t *) action->data; if (action_is_pending(a)) { fprintf(stderr, "\tAction %d: %s", a->id, a->uuid); if (a->node) { fprintf(stderr, "\ton %s", a->node->details->uname); } fprintf(stderr, "\n"); } } } /* For --wait, timeout (in seconds) to use if caller doesn't specify one */ #define WAIT_DEFAULT_TIMEOUT_S (60 * 60) /* For --wait, how long to sleep between cluster state checks */ #define WAIT_SLEEP_S (2) /*! * \internal * \brief Wait until all pending cluster actions are complete * * This waits until either the CIB's transition graph is idle or a timeout is * reached. * * \param[in] timeout_ms Consider failed if actions do not complete in this time * (specified in milliseconds, but one-second granularity * is actually used; if 0, a default will be used) * \param[in] cib Connection to the CIB manager * * \return pcmk_ok on success, -errno on failure */ int wait_till_stable(int timeout_ms, cib_t * cib) { pe_working_set_t *data_set = NULL; int rc = -1; int timeout_s = timeout_ms? ((timeout_ms + 999) / 1000) : WAIT_DEFAULT_TIMEOUT_S; time_t expire_time = time(NULL) + timeout_s; time_t time_diff; bool printed_version_warning = BE_QUIET; // i.e. don't print if quiet data_set = pe_new_working_set(); if (data_set == NULL) { return -ENOMEM; } set_bit(data_set->flags, pe_flag_no_counts); set_bit(data_set->flags, pe_flag_no_compat); do { /* Abort if timeout is reached */ time_diff = expire_time - time(NULL); if (time_diff > 0) { crm_info("Waiting up to %ld seconds for cluster actions to complete", time_diff); } else { print_pending_actions(data_set->actions); pe_free_working_set(data_set); return -ETIME; } if (rc == pcmk_ok) { /* this avoids sleep on first loop iteration */ sleep(WAIT_SLEEP_S); } /* Get latest transition graph */ pe_reset_working_set(data_set); rc = update_working_set_from_cib(data_set, cib); if (rc != pcmk_ok) { pe_free_working_set(data_set); return rc; } pcmk__schedule_actions(data_set, data_set->input, NULL); if (!printed_version_warning) { /* If the DC has a different version than the local node, the two * could come to different conclusions about what actions need to be * done. Warn the user in this case. * * @TODO A possible long-term solution would be to reimplement the * wait as a new controller operation that would be forwarded to the * DC. However, that would have potential problems of its own. */ const char *dc_version = g_hash_table_lookup(data_set->config_hash, "dc-version"); if (safe_str_neq(dc_version, PACEMAKER_VERSION "-" BUILD_VERSION)) { printf("warning: wait option may not work properly in " "mixed-version cluster\n"); printed_version_warning = TRUE; } } } while (actions_are_pending(data_set->actions)); pe_free_working_set(data_set); return pcmk_ok; } int cli_resource_execute_from_params(const char *rsc_name, const char *rsc_class, const char *rsc_prov, const char *rsc_type, const char *action, GHashTable *params, GHashTable *override_hash, int timeout_ms) { GHashTable *params_copy = NULL; int rc = pcmk_ok; svc_action_t *op = NULL; if (safe_str_eq(rsc_class, PCMK_RESOURCE_CLASS_STONITH)) { CMD_ERR("Sorry, the %s option doesn't support %s resources yet", action, rsc_class); crm_exit(CRM_EX_UNIMPLEMENT_FEATURE); } /* If no timeout was provided, grab the default. */ if (timeout_ms == 0) { timeout_ms = crm_get_msec(CRM_DEFAULT_OP_TIMEOUT_S); } /* add meta_timeout env needed by some resource agents */ g_hash_table_insert(params, strdup("CRM_meta_timeout"), crm_strdup_printf("%d", timeout_ms)); /* add crm_feature_set env needed by some resource agents */ g_hash_table_insert(params, strdup(XML_ATTR_CRM_VERSION), strdup(CRM_FEATURE_SET)); /* resources_action_create frees the params hash table it's passed, but we * may need to reuse it in a second call to resources_action_create. Thus * we'll make a copy here so that gets freed and the original remains for * reuse. */ params_copy = crm_str_table_dup(params); op = resources_action_create(rsc_name, rsc_class, rsc_prov, rsc_type, action, 0, timeout_ms, params_copy, 0); if (op == NULL) { /* Re-run with stderr enabled so we can display a sane error message */ crm_enable_stderr(TRUE); params_copy = crm_str_table_dup(params); op = resources_action_create(rsc_name, rsc_class, rsc_prov, rsc_type, action, 0, timeout_ms, params_copy, 0); /* Callers of cli_resource_execute expect that the params hash table will * be freed. That function uses this one, so for that reason and for * making the two act the same, we should free the hash table here too. */ g_hash_table_destroy(params); /* We know op will be NULL, but this makes static analysis happy */ services_action_free(op); crm_exit(CRM_EX_DATAERR); } setenv("HA_debug", resource_verbose > 0 ? "1" : "0", 1); if(resource_verbose > 1) { setenv("OCF_TRACE_RA", "1", 1); } /* A resource agent using the standard ocf-shellfuncs library will not print * messages to stderr if it doesn't have a controlling terminal (e.g. if * crm_resource is called via script or ssh). This forces it to do so. */ setenv("OCF_TRACE_FILE", "/dev/stderr", 0); if (override_hash) { GHashTableIter iter; char *name = NULL; char *value = NULL; g_hash_table_iter_init(&iter, override_hash); while (g_hash_table_iter_next(&iter, (gpointer *) & name, (gpointer *) & value)) { printf("Overriding the cluster configuration for '%s' with '%s' = '%s'\n", rsc_name, name, value); g_hash_table_replace(op->params, strdup(name), strdup(value)); } } if (services_action_sync(op)) { int more, lpc, last; char *local_copy = NULL; rc = op->rc; if (op->status == PCMK_LRM_OP_DONE) { printf("Operation %s for %s (%s:%s:%s) returned: '%s' (%d)\n", action, rsc_name, rsc_class, rsc_prov ? rsc_prov : "", rsc_type, services_ocf_exitcode_str(op->rc), op->rc); } else { printf("Operation %s for %s (%s:%s:%s) failed: '%s' (%d)\n", action, rsc_name, rsc_class, rsc_prov ? rsc_prov : "", rsc_type, services_lrm_status_str(op->status), op->status); } /* hide output for validate-all if not in verbose */ if (resource_verbose == 0 && safe_str_eq(action, "validate-all")) goto done; if (op->stdout_data) { local_copy = strdup(op->stdout_data); more = strlen(local_copy); last = 0; for (lpc = 0; lpc < more; lpc++) { if (local_copy[lpc] == '\n' || local_copy[lpc] == 0) { local_copy[lpc] = 0; printf(" > stdout: %s\n", local_copy + last); last = lpc + 1; } } free(local_copy); } if (op->stderr_data) { local_copy = strdup(op->stderr_data); more = strlen(local_copy); last = 0; for (lpc = 0; lpc < more; lpc++) { if (local_copy[lpc] == '\n' || local_copy[lpc] == 0) { local_copy[lpc] = 0; printf(" > stderr: %s\n", local_copy + last); last = lpc + 1; } } free(local_copy); } } else { rc = op->rc == 0 ? pcmk_err_generic : op->rc; } done: services_action_free(op); /* See comment above about why we free params here. */ g_hash_table_destroy(params); return rc; } int cli_resource_execute(resource_t *rsc, const char *requested_name, const char *rsc_action, GHashTable *override_hash, int timeout_ms, cib_t * cib, pe_working_set_t *data_set) { int rc = pcmk_ok; const char *rid = NULL; const char *rtype = NULL; const char *rprov = NULL; const char *rclass = NULL; const char *action = NULL; GHashTable *params = NULL; if (safe_str_eq(rsc_action, "validate")) { action = "validate-all"; } else if (safe_str_eq(rsc_action, "force-check")) { action = "monitor"; } else if (safe_str_eq(rsc_action, "force-stop")) { action = rsc_action+6; } else if (safe_str_eq(rsc_action, "force-start") || safe_str_eq(rsc_action, "force-demote") || safe_str_eq(rsc_action, "force-promote")) { action = rsc_action+6; if(pe_rsc_is_clone(rsc)) { rc = cli_resource_search(rsc, requested_name, data_set); if(rc > 0 && do_force == FALSE) { CMD_ERR("It is not safe to %s %s here: the cluster claims it is already active", action, rsc->id); CMD_ERR("Try setting target-role=Stopped first or specifying " "the force option"); crm_exit(CRM_EX_UNSAFE); } } } else { action = rsc_action; } if(pe_rsc_is_clone(rsc)) { /* Grab the first child resource in the hope it's not a group */ rsc = rsc->children->data; } if(rsc->variant == pe_group) { CMD_ERR("Sorry, the %s option doesn't support group resources", rsc_action); crm_exit(CRM_EX_UNIMPLEMENT_FEATURE); } rclass = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS); rprov = crm_element_value(rsc->xml, XML_AGENT_ATTR_PROVIDER); rtype = crm_element_value(rsc->xml, XML_ATTR_TYPE); params = generate_resource_params(rsc, data_set); if (timeout_ms == 0) { timeout_ms = pe_get_configured_timeout(rsc, action, data_set); } rid = pe_rsc_is_anon_clone(rsc->parent)? requested_name : rsc->id; rc = cli_resource_execute_from_params(rid, rclass, rprov, rtype, action, params, override_hash, timeout_ms); return rc; } int cli_resource_move(resource_t *rsc, const char *rsc_id, const char *host_name, cib_t *cib, pe_working_set_t *data_set) { int rc = pcmk_ok; unsigned int count = 0; node_t *current = NULL; node_t *dest = pe_find_node(data_set->nodes, host_name); bool cur_is_dest = FALSE; if (dest == NULL) { return -pcmk_err_node_unknown; } if (scope_master && is_not_set(rsc->flags, pe_rsc_promotable)) { resource_t *p = uber_parent(rsc); if (is_set(p->flags, pe_rsc_promotable)) { CMD_ERR("Using parent '%s' for move instead of '%s'.", rsc->id, rsc_id); rsc_id = p->id; rsc = p; } else { CMD_ERR("Ignoring master option: %s is not promotable", rsc_id); scope_master = FALSE; } } current = pe__find_active_requires(rsc, &count); if (is_set(rsc->flags, pe_rsc_promotable)) { GListPtr iter = NULL; unsigned int master_count = 0; pe_node_t *master_node = NULL; for(iter = rsc->children; iter; iter = iter->next) { resource_t *child = (resource_t *)iter->data; enum rsc_role_e child_role = child->fns->state(child, TRUE); if(child_role == RSC_ROLE_MASTER) { rsc = child; master_node = pe__current_node(child); master_count++; } } if (scope_master || master_count) { count = master_count; current = master_node; } } if (count > 1) { if (pe_rsc_is_clone(rsc)) { current = NULL; } else { return -pcmk_err_multiple; } } if (current && (current->details == dest->details)) { cur_is_dest = TRUE; if (do_force) { crm_info("%s is already %s on %s, reinforcing placement with location constraint.", rsc_id, scope_master?"promoted":"active", dest->details->uname); } else { return -pcmk_err_already; } } /* Clear any previous prefer constraints across all nodes. */ cli_resource_clear(rsc_id, NULL, data_set->nodes, cib, FALSE); /* Clear any previous ban constraints on 'dest'. */ cli_resource_clear(rsc_id, dest->details->uname, data_set->nodes, cib, TRUE); /* Record an explicit preference for 'dest' */ rc = cli_resource_prefer(rsc_id, dest->details->uname, cib); crm_trace("%s%s now prefers node %s%s", rsc->id, scope_master?" (master)":"", dest->details->uname, do_force?"(forced)":""); /* only ban the previous location if current location != destination location. * it is possible to use -M to enforce a location without regard of where the * resource is currently located */ if(do_force && (cur_is_dest == FALSE)) { /* Ban the original location if possible */ if(current) { (void)cli_resource_ban(rsc_id, current->details->uname, NULL, cib); } else if(count > 1) { CMD_ERR("Resource '%s' is currently %s in %d locations. " "One may now move to %s", rsc_id, (scope_master? "promoted" : "active"), count, dest->details->uname); CMD_ERR("To prevent '%s' from being %s at a specific location, " "specify a node.", rsc_id, (scope_master? "promoted" : "active")); } else { crm_trace("Not banning %s from its current location: not active", rsc_id); } } return rc; } static void cli_resource_why_without_rsc_and_host(cib_t *cib_conn,GListPtr resources) { GListPtr lpc = NULL; GListPtr hosts = NULL; for (lpc = resources; lpc != NULL; lpc = lpc->next) { resource_t *rsc = (resource_t *) lpc->data; rsc->fns->location(rsc, &hosts, TRUE); if (hosts == NULL) { printf("Resource %s is not running\n", rsc->id); } else { printf("Resource %s is running\n", rsc->id); } cli_resource_check(cib_conn, rsc); g_list_free(hosts); hosts = NULL; } } static void cli_resource_why_with_rsc_and_host(cib_t *cib_conn, GListPtr resources, resource_t *rsc, const char *host_uname) { if (resource_is_running_on(rsc, host_uname)) { printf("Resource %s is running on host %s\n",rsc->id,host_uname); } else { printf("Resource %s is not running on host %s\n", rsc->id, host_uname); } cli_resource_check(cib_conn, rsc); } static void cli_resource_why_without_rsc_with_host(cib_t *cib_conn,GListPtr resources,node_t *node) { const char* host_uname = node->details->uname; GListPtr allResources = node->details->allocated_rsc; GListPtr activeResources = node->details->running_rsc; GListPtr unactiveResources = subtract_lists(allResources,activeResources,(GCompareFunc) strcmp); GListPtr lpc = NULL; for (lpc = activeResources; lpc != NULL; lpc = lpc->next) { resource_t *rsc = (resource_t *) lpc->data; printf("Resource %s is running on host %s\n",rsc->id,host_uname); cli_resource_check(cib_conn,rsc); } for(lpc = unactiveResources; lpc != NULL; lpc = lpc->next) { resource_t *rsc = (resource_t *) lpc->data; printf("Resource %s is assigned to host %s but not running\n", rsc->id, host_uname); cli_resource_check(cib_conn,rsc); } g_list_free(allResources); g_list_free(activeResources); g_list_free(unactiveResources); } static void cli_resource_why_with_rsc_without_host(cib_t *cib_conn, GListPtr resources, resource_t *rsc) { GListPtr hosts = NULL; rsc->fns->location(rsc, &hosts, TRUE); printf("Resource %s is %srunning\n", rsc->id, (hosts? "" : "not ")); cli_resource_check(cib_conn, rsc); g_list_free(hosts); } void cli_resource_why(cib_t *cib_conn, GListPtr resources, resource_t *rsc, node_t *node) { const char *host_uname = (node == NULL)? NULL : node->details->uname; if ((rsc == NULL) && (host_uname == NULL)) { cli_resource_why_without_rsc_and_host(cib_conn, resources); } else if ((rsc != NULL) && (host_uname != NULL)) { cli_resource_why_with_rsc_and_host(cib_conn, resources, rsc, host_uname); } else if ((rsc == NULL) && (host_uname != NULL)) { cli_resource_why_without_rsc_with_host(cib_conn, resources, node); } else if ((rsc != NULL) && (host_uname == NULL)) { cli_resource_why_with_rsc_without_host(cib_conn, resources, rsc); } } diff --git a/tools/crm_simulate.c b/tools/crm_simulate.c index b6267d5421..0c7ee271eb 100644 --- a/tools/crm_simulate.c +++ b/tools/crm_simulate.c @@ -1,1109 +1,1109 @@ /* * Copyright 2009-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include cib_t *global_cib = NULL; GListPtr op_fail = NULL; bool action_numbers = FALSE; gboolean quiet = FALSE; gboolean print_pending = TRUE; char *temp_shadow = NULL; extern gboolean bringing_nodes_online; #define quiet_log(fmt, args...) do { \ if(quiet == FALSE) { \ printf(fmt , ##args); \ } \ } while(0) char *use_date = NULL; static void get_date(pe_working_set_t *data_set, bool print_original) { time_t original_date = 0; crm_element_value_epoch(data_set->input, "execution-date", &original_date); if (use_date) { data_set->now = crm_time_new(use_date); quiet_log(" + Setting effective cluster time: %s", use_date); crm_time_log(LOG_NOTICE, "Pretending 'now' is", data_set->now, crm_time_log_date | crm_time_log_timeofday); } else if (original_date) { data_set->now = crm_time_new(NULL); crm_time_set_timet(data_set->now, &original_date); if (print_original) { char *when = crm_time_as_string(data_set->now, crm_time_log_date|crm_time_log_timeofday); printf("Using the original execution date of: %s\n", when); free(when); } } } static void print_cluster_status(pe_working_set_t * data_set, long options) { char *online_nodes = NULL; char *online_remote_nodes = NULL; char *online_guest_nodes = NULL; char *offline_nodes = NULL; char *offline_remote_nodes = NULL; GListPtr gIter = NULL; for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; const char *node_mode = NULL; char *node_name = NULL; if (pe__is_guest_node(node)) { node_name = crm_strdup_printf("%s:%s", node->details->uname, node->details->remote_rsc->container->id); } else { node_name = crm_strdup_printf("%s", node->details->uname); } if (node->details->unclean) { if (node->details->online && node->details->unclean) { node_mode = "UNCLEAN (online)"; } else if (node->details->pending) { node_mode = "UNCLEAN (pending)"; } else { node_mode = "UNCLEAN (offline)"; } } else if (node->details->pending) { node_mode = "pending"; } else if (node->details->standby_onfail && node->details->online) { node_mode = "standby (on-fail)"; } else if (node->details->standby) { if (node->details->online) { node_mode = "standby"; } else { node_mode = "OFFLINE (standby)"; } } else if (node->details->maintenance) { if (node->details->online) { node_mode = "maintenance"; } else { node_mode = "OFFLINE (maintenance)"; } } else if (node->details->online) { if (pe__is_guest_node(node)) { online_guest_nodes = pcmk__add_word(online_guest_nodes, node_name); } else if (pe__is_remote_node(node)) { online_remote_nodes = pcmk__add_word(online_remote_nodes, node_name); } else { online_nodes = pcmk__add_word(online_nodes, node_name); } free(node_name); continue; } else { if (pe__is_remote_node(node)) { offline_remote_nodes = pcmk__add_word(offline_remote_nodes, node_name); } else if (pe__is_guest_node(node)) { /* ignore offline container nodes */ } else { offline_nodes = pcmk__add_word(offline_nodes, node_name); } free(node_name); continue; } if (pe__is_guest_node(node)) { printf("GuestNode %s: %s\n", node_name, node_mode); } else if (pe__is_remote_node(node)) { printf("RemoteNode %s: %s\n", node_name, node_mode); } else if (safe_str_eq(node->details->uname, node->details->id)) { printf("Node %s: %s\n", node_name, node_mode); } else { printf("Node %s (%s): %s\n", node_name, node->details->id, node_mode); } free(node_name); } if (online_nodes) { printf("Online: [%s ]\n", online_nodes); free(online_nodes); } if (offline_nodes) { printf("OFFLINE: [%s ]\n", offline_nodes); free(offline_nodes); } if (online_remote_nodes) { printf("RemoteOnline: [%s ]\n", online_remote_nodes); free(online_remote_nodes); } if (offline_remote_nodes) { printf("RemoteOFFLINE: [%s ]\n", offline_remote_nodes); free(offline_remote_nodes); } if (online_guest_nodes) { printf("GuestOnline: [%s ]\n", online_guest_nodes); free(online_guest_nodes); } fprintf(stdout, "\n"); for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; if (is_set(rsc->flags, pe_rsc_orphan) && rsc->role == RSC_ROLE_STOPPED) { continue; } rsc->fns->print(rsc, NULL, pe_print_printf | options, stdout); } fprintf(stdout, "\n"); } static char * create_action_name(pe_action_t *action) { char *action_name = NULL; const char *prefix = ""; const char *action_host = NULL; const char *clone_name = NULL; const char *task = action->task; if (action->node) { action_host = action->node->details->uname; } else if (is_not_set(action->flags, pe_action_pseudo)) { action_host = ""; } if (safe_str_eq(action->task, RSC_CANCEL)) { prefix = "Cancel "; task = action->cancel_task; } if (action->rsc && action->rsc->clone_name) { clone_name = action->rsc->clone_name; } if (clone_name) { char *key = NULL; guint interval_ms = 0; if (pcmk__guint_from_hash(action->meta, XML_LRM_ATTR_INTERVAL_MS, 0, &interval_ms) != pcmk_rc_ok) { interval_ms = 0; } if (safe_str_eq(action->task, RSC_NOTIFY) || safe_str_eq(action->task, RSC_NOTIFIED)) { const char *n_type = g_hash_table_lookup(action->meta, "notify_key_type"); const char *n_task = g_hash_table_lookup(action->meta, "notify_key_operation"); CRM_ASSERT(n_type != NULL); CRM_ASSERT(n_task != NULL); key = pcmk__notify_key(clone_name, n_type, n_task); } else { key = pcmk__op_key(clone_name, task, interval_ms); } if (action_host) { action_name = crm_strdup_printf("%s%s %s", prefix, key, action_host); } else { action_name = crm_strdup_printf("%s%s", prefix, key); } free(key); } else if (safe_str_eq(action->task, CRM_OP_FENCE)) { const char *op = g_hash_table_lookup(action->meta, "stonith_action"); action_name = crm_strdup_printf("%s%s '%s' %s", prefix, action->task, op, action_host); } else if (action->rsc && action_host) { action_name = crm_strdup_printf("%s%s %s", prefix, action->uuid, action_host); } else if (action_host) { action_name = crm_strdup_printf("%s%s %s", prefix, action->task, action_host); } else { action_name = crm_strdup_printf("%s", action->uuid); } if (action_numbers) { // i.e. verbose char *with_id = crm_strdup_printf("%s (%d)", action_name, action->id); free(action_name); action_name = with_id; } return action_name; } static void create_dotfile(pe_working_set_t * data_set, const char *dot_file, gboolean all_actions) { GListPtr gIter = NULL; FILE *dot_strm = fopen(dot_file, "w"); if (dot_strm == NULL) { crm_perror(LOG_ERR, "Could not open %s for writing", dot_file); return; } fprintf(dot_strm, " digraph \"g\" {\n"); for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) { action_t *action = (action_t *) gIter->data; const char *style = "dashed"; const char *font = "black"; const char *color = "black"; char *action_name = create_action_name(action); crm_trace("Action %d: %s %s %p", action->id, action_name, action->uuid, action); if (is_set(action->flags, pe_action_pseudo)) { font = "orange"; } 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 = "red"; font = "purple"; if (all_actions == FALSE) { goto do_not_write; } } else if (is_set(action->flags, pe_action_optional)) { color = "blue"; if (all_actions == FALSE) { goto do_not_write; } } else { color = "red"; CRM_CHECK(is_set(action->flags, pe_action_runnable) == FALSE,; ); } set_bit(action->flags, pe_action_dumped); crm_trace("\"%s\" [ style=%s color=\"%s\" fontcolor=\"%s\"]", action_name, style, color, font); fprintf(dot_strm, "\"%s\" [ style=%s color=\"%s\" fontcolor=\"%s\"]\n", action_name, style, color, font); do_not_write: free(action_name); } for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) { action_t *action = (action_t *) gIter->data; GListPtr gIter2 = NULL; for (gIter2 = action->actions_before; gIter2 != NULL; gIter2 = gIter2->next) { action_wrapper_t *before = (action_wrapper_t *) gIter2->data; char *before_name = NULL; char *after_name = NULL; const char *style = "dashed"; gboolean optional = TRUE; if (before->state == pe_link_dumped) { optional = FALSE; style = "bold"; } else if (is_set(action->flags, pe_action_pseudo) && (before->type & pe_order_stonith_stop)) { continue; } else if (before->type == pe_order_none) { continue; } else if (is_set(before->action->flags, pe_action_dumped) && is_set(action->flags, pe_action_dumped) && before->type != pe_order_load) { optional = FALSE; } if (all_actions || optional == FALSE) { before_name = create_action_name(before->action); after_name = create_action_name(action); crm_trace("\"%s\" -> \"%s\" [ style = %s]", before_name, after_name, style); fprintf(dot_strm, "\"%s\" -> \"%s\" [ style = %s]\n", before_name, after_name, style); free(before_name); free(after_name); } } } fprintf(dot_strm, "}\n"); fflush(dot_strm); fclose(dot_strm); } static void setup_input(const char *input, const char *output) { int rc = pcmk_ok; cib_t *cib_conn = NULL; xmlNode *cib_object = NULL; char *local_output = NULL; if (input == NULL) { /* Use live CIB */ cib_conn = cib_new(); rc = cib_conn->cmds->signon(cib_conn, crm_system_name, cib_command); if (rc == pcmk_ok) { rc = cib_conn->cmds->query(cib_conn, NULL, &cib_object, cib_scope_local | cib_sync_call); } cib_conn->cmds->signoff(cib_conn); cib_delete(cib_conn); cib_conn = NULL; if (rc != pcmk_ok) { fprintf(stderr, "Live CIB query failed: %s (%d)\n", pcmk_strerror(rc), rc); crm_exit(crm_errno2exit(rc)); } else if (cib_object == NULL) { fprintf(stderr, "Live CIB query failed: empty result\n"); crm_exit(CRM_EX_NOINPUT); } } else if (safe_str_eq(input, "-")) { cib_object = filename2xml(NULL); } else { cib_object = filename2xml(input); } 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); crm_exit(CRM_EX_CONFIG); } if (validate_xml(cib_object, NULL, FALSE) != TRUE) { free_xml(cib_object); crm_exit(CRM_EX_CONFIG); } if (output == NULL) { - char *pid = crm_getpid_s(); + char *pid = pcmk__getpid_s(); local_output = get_shadow_file(pid); temp_shadow = strdup(local_output); output = local_output; free(pid); } rc = write_xml_file(cib_object, output, FALSE); free_xml(cib_object); cib_object = NULL; if (rc < 0) { fprintf(stderr, "Could not create '%s': %s\n", output, pcmk_strerror(rc)); crm_exit(CRM_EX_CANTCREAT); } setenv("CIB_file", output, 1); free(local_output); } static pcmk__cli_option_t long_options[] = { // long option, argument type, storage, short option, description, flags { "help", no_argument, NULL, '?', "\tThis text", pcmk__option_default }, { "version", no_argument, NULL, '$', "\tVersion information", pcmk__option_default }, { "quiet", no_argument, NULL, 'Q', "\tDisplay only essential output", pcmk__option_default }, { "verbose", no_argument, NULL, 'V', "\tIncrease debug output", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\nOperations:", pcmk__option_default }, { "run", no_argument, NULL, 'R', "\tDetermine cluster's response to the given configuration and status", pcmk__option_default }, { "simulate", no_argument, NULL, 'S', "Simulate transition's execution and display resulting cluster status", pcmk__option_default }, { "in-place", no_argument, NULL, 'X', "Simulate transition's execution and store result back to input file", pcmk__option_default }, { "show-scores", no_argument, NULL, 's', "Show allocation scores", pcmk__option_default }, { "show-utilization", no_argument, NULL, 'U', "Show utilization information", pcmk__option_default }, { "profile", required_argument, NULL, 'P', "Run all tests in the named directory to create profiling data", pcmk__option_default }, { "repeat", required_argument, NULL, 'N', "With --profile, repeat each test N times and print timings", pcmk__option_default }, { "pending", no_argument, NULL, 'j', "\tDisplay pending state if 'record-pending' is enabled", pcmk__option_hidden }, { "-spacer-", no_argument, NULL, '-', "\nSynthetic Cluster Events:", pcmk__option_default }, { "node-up", required_argument, NULL, 'u', "\tBring a node online", pcmk__option_default }, { "node-down", required_argument, NULL, 'd', "\tTake a node offline", pcmk__option_default }, { "node-fail", required_argument, NULL, 'f', "\tMark a node as failed", pcmk__option_default }, { "op-inject", required_argument, NULL, 'i', "\tGenerate a failure for the cluster to react to in the simulation", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\t\tValue is of the form " "${resource}_${task}_${interval_in_ms}@${node}=${rc}.", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\t\t(for example, memcached_monitor_20000@bart.example.com=7)", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\t\tFor more information on OCF return codes, refer to: " "https://clusterlabs.org/pacemaker/doc/en-US/Pacemaker/" "2.0/html/Pacemaker_Administration/s-ocf-return-codes.html", pcmk__option_default }, { "op-fail", required_argument, NULL, 'F', "\tIf the specified task occurs during the simulation, have it fail " "with return code ${rc}", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\t\tValue is of the form " "${resource}_${task}_${interval_in_ms}@${node}=${rc}.", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\t\t(for example, memcached_stop_0@bart.example.com=1)\n", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\t\tThe transition will normally stop at the failed action. Save " "the result with --save-output and re-run with --xml-file", pcmk__option_default }, { "set-datetime", required_argument, NULL, 't', "Set date/time (ISO 8601 format, see " "https://en.wikipedia.org/wiki/ISO_8601)", pcmk__option_default }, { "quorum", required_argument, NULL, 'q', "\tSpecify a value for quorum", pcmk__option_default }, { "watchdog", required_argument, NULL, 'w', "\tAssume a watchdog device is active", pcmk__option_default }, { "ticket-grant", required_argument, NULL, 'g', "Grant a ticket", pcmk__option_default }, { "ticket-revoke", required_argument, NULL, 'r', "Revoke a ticket", pcmk__option_default }, { "ticket-standby", required_argument, NULL, 'b', "Make a ticket standby", pcmk__option_default }, { "ticket-activate", required_argument, NULL, 'e', "Activate a ticket", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\nOutput Options:", pcmk__option_default }, { "save-input", required_argument, NULL, 'I', "\tSave the input configuration to the named file", pcmk__option_default }, { "save-output", required_argument, NULL, 'O', "Save the output configuration to the named file", pcmk__option_default }, { "save-graph", required_argument, NULL, 'G', "\tSave the transition graph (XML format) to the named file", pcmk__option_default }, { "save-dotfile", required_argument, NULL, 'D', "Save the transition graph (DOT format) to the named file", pcmk__option_default }, { "all-actions", no_argument, NULL, 'a', "\tDisplay all possible actions in DOT graph (even if not part " "of transition)", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\nData Source:", pcmk__option_default }, { "live-check", no_argument, NULL, 'L', "\tConnect to CIB mamager and use the current CIB contents as input", pcmk__option_default }, { "xml-file", required_argument, NULL, 'x', "\tRetrieve XML from the named file", pcmk__option_default }, { "xml-pipe", no_argument, NULL, 'p', "\tRetrieve XML from stdin", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\nExamples:\n", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "Pretend a recurring monitor action found memcached stopped on node " "fred.example.com and, during recovery, that the memcached stop " "action failed", pcmk__option_paragraph }, { "-spacer-", no_argument, NULL, '-', " crm_simulate -LS --op-inject " "memcached:0_monitor_20000@bart.example.com=7 " "--op-fail memcached:0_stop_0@fred.example.com=1 " "--save-output /tmp/memcached-test.xml", pcmk__option_example }, { "-spacer-", no_argument, NULL, '-', "Now see what the reaction to the stop failure would be", pcmk__option_paragraph }, { "-spacer-", no_argument, NULL, '-', " crm_simulate -S --xml-file /tmp/memcached-test.xml", pcmk__option_example }, { 0, 0, 0, 0 } }; static void profile_one(const char *xml_file, long long repeat, pe_working_set_t *data_set) { xmlNode *cib_object = NULL; clock_t start = 0; printf("* Testing %s ...", xml_file); fflush(stdout); cib_object = filename2xml(xml_file); start = clock(); 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; } if (validate_xml(cib_object, NULL, FALSE) != TRUE) { free_xml(cib_object); return; } for (int i = 0; i < repeat; ++i) { xmlNode *input = (repeat == 1)? cib_object : copy_xml(cib_object); data_set->input = input; get_date(data_set, false); pcmk__schedule_actions(data_set, input, NULL); pe_reset_working_set(data_set); } printf(" %.2f secs\n", (clock() - start) / (float) CLOCKS_PER_SEC); } #ifndef FILENAME_MAX # define FILENAME_MAX 512 #endif static void profile_all(const char *dir, long long repeat, pe_working_set_t *data_set) { struct dirent **namelist; int file_num = scandir(dir, &namelist, 0, alphasort); if (file_num > 0) { struct stat prop; char buffer[FILENAME_MAX]; while (file_num--) { if ('.' == namelist[file_num]->d_name[0]) { free(namelist[file_num]); continue; } else if (!pcmk__ends_with_ext(namelist[file_num]->d_name, ".xml")) { free(namelist[file_num]); continue; } snprintf(buffer, sizeof(buffer), "%s/%s", dir, namelist[file_num]->d_name); if (stat(buffer, &prop) == 0 && S_ISREG(prop.st_mode)) { profile_one(buffer, repeat, data_set); } free(namelist[file_num]); } free(namelist); } } int main(int argc, char **argv) { int rc = pcmk_ok; guint modified = 0; gboolean store = FALSE; gboolean process = FALSE; gboolean simulate = FALSE; gboolean all_actions = FALSE; gboolean have_stdout = FALSE; pe_working_set_t *data_set = NULL; const char *xml_file = "-"; const char *quorum = NULL; const char *watchdog = NULL; const char *test_dir = NULL; const char *dot_file = NULL; const char *graph_file = NULL; const char *input_file = NULL; const char *output_file = NULL; const char *repeat_s = NULL; int flag = 0; int index = 0; int argerr = 0; long long repeat = 1; GListPtr node_up = NULL; GListPtr node_down = NULL; GListPtr node_fail = NULL; GListPtr op_inject = NULL; GListPtr ticket_grant = NULL; GListPtr ticket_revoke = NULL; GListPtr ticket_standby = NULL; GListPtr ticket_activate = NULL; xmlNode *input = NULL; crm_log_cli_init("crm_simulate"); pcmk__set_cli_options(NULL, " [options]", long_options, "simulate a Pacemaker cluster's response to events"); if (argc < 2) { pcmk__cli_help('?', CRM_EX_USAGE); } while (1) { flag = pcmk__next_cli_option(argc, argv, &index, NULL); if (flag == -1) break; switch (flag) { case 'V': if (have_stdout == FALSE) { /* Redirect stderr to stdout so we can grep the output */ have_stdout = TRUE; close(STDERR_FILENO); dup2(STDOUT_FILENO, STDERR_FILENO); } crm_bump_log_level(argc, argv); action_numbers = TRUE; break; case '?': case '$': pcmk__cli_help(flag, CRM_EX_OK); break; case 'p': xml_file = "-"; break; case 'Q': quiet = TRUE; break; case 'L': xml_file = NULL; break; case 'x': xml_file = optarg; break; case 'u': modified++; bringing_nodes_online = TRUE; node_up = g_list_append(node_up, optarg); break; case 'd': modified++; node_down = g_list_append(node_down, optarg); break; case 'f': modified++; node_fail = g_list_append(node_fail, optarg); break; case 't': use_date = strdup(optarg); break; case 'i': modified++; op_inject = g_list_append(op_inject, optarg); break; case 'F': process = TRUE; simulate = TRUE; op_fail = g_list_append(op_fail, optarg); break; case 'w': modified++; watchdog = optarg; break; case 'q': modified++; quorum = optarg; break; case 'g': modified++; ticket_grant = g_list_append(ticket_grant, optarg); break; case 'r': modified++; ticket_revoke = g_list_append(ticket_revoke, optarg); break; case 'b': modified++; ticket_standby = g_list_append(ticket_standby, optarg); break; case 'e': modified++; ticket_activate = g_list_append(ticket_activate, optarg); break; case 'a': all_actions = TRUE; break; case 's': process = TRUE; show_scores = TRUE; break; case 'U': process = TRUE; show_utilization = TRUE; break; case 'j': print_pending = TRUE; break; case 'S': process = TRUE; simulate = TRUE; break; case 'X': store = TRUE; process = TRUE; simulate = TRUE; break; case 'R': process = TRUE; break; case 'D': process = TRUE; dot_file = optarg; break; case 'G': process = TRUE; graph_file = optarg; break; case 'I': input_file = optarg; break; case 'O': output_file = optarg; break; case 'P': test_dir = optarg; break; case 'N': repeat_s = optarg; break; default: ++argerr; break; } } if (optind > argc) { ++argerr; } if (argerr) { pcmk__cli_help('?', CRM_EX_USAGE); } data_set = pe_new_working_set(); if (data_set == NULL) { crm_perror(LOG_ERR, "Could not allocate working set"); rc = -ENOMEM; goto done; } set_bit(data_set->flags, pe_flag_no_compat); if (test_dir != NULL) { if (repeat_s != NULL) { repeat = crm_parse_ll(repeat_s, NULL); if (errno || (repeat < 1)) { fprintf(stderr, "--repeat must be positive integer, not '%s' -- using 1", repeat_s); repeat = 1; } } profile_all(test_dir, repeat, data_set); return CRM_EX_OK; } setup_input(xml_file, store ? xml_file : output_file); global_cib = cib_new(); rc = global_cib->cmds->signon(global_cib, crm_system_name, cib_command); if (rc != pcmk_ok) { fprintf(stderr, "Could not connect to the CIB manager: %s\n", pcmk_strerror(rc)); goto done; } rc = global_cib->cmds->query(global_cib, NULL, &input, cib_sync_call | cib_scope_local); if (rc != pcmk_ok) { fprintf(stderr, "Could not get local CIB: %s\n", pcmk_strerror(rc)); goto done; } data_set->input = input; get_date(data_set, true); if(xml_file) { set_bit(data_set->flags, pe_flag_sanitized); } set_bit(data_set->flags, pe_flag_stdout); cluster_status(data_set); if (quiet == FALSE) { int options = print_pending ? pe_print_pending : 0; if (is_set(data_set->flags, pe_flag_maintenance_mode)) { quiet_log("\n *** Resource management is DISABLED ***"); quiet_log("\n The cluster will not attempt to start, stop or recover services"); quiet_log("\n"); } if (data_set->disabled_resources || data_set->blocked_resources) { quiet_log("%d of %d resource instances DISABLED and %d BLOCKED " "from further action due to failure\n", data_set->disabled_resources, data_set->ninstances, data_set->blocked_resources); } quiet_log("\nCurrent cluster status:\n"); print_cluster_status(data_set, options); } if (modified) { quiet_log("Performing requested modifications\n"); modify_configuration(data_set, global_cib, quorum, watchdog, node_up, node_down, node_fail, op_inject, ticket_grant, ticket_revoke, ticket_standby, ticket_activate); rc = global_cib->cmds->query(global_cib, NULL, &input, cib_sync_call); if (rc != pcmk_ok) { fprintf(stderr, "Could not get modified CIB: %s\n", pcmk_strerror(rc)); goto done; } cleanup_calculations(data_set); data_set->input = input; get_date(data_set, true); if(xml_file) { set_bit(data_set->flags, pe_flag_sanitized); } set_bit(data_set->flags, pe_flag_stdout); cluster_status(data_set); } if (input_file != NULL) { rc = write_xml_file(input, input_file, FALSE); if (rc < 0) { fprintf(stderr, "Could not create '%s': %s\n", input_file, pcmk_strerror(rc)); goto done; } } if (process || simulate) { crm_time_t *local_date = NULL; if (show_scores && show_utilization) { printf("Allocation scores and utilization information:\n"); } else if (show_scores) { fprintf(stdout, "Allocation scores:\n"); } else if (show_utilization) { printf("Utilization information:\n"); } pcmk__schedule_actions(data_set, input, local_date); input = NULL; /* Don't try and free it twice */ if (graph_file != NULL) { write_xml_file(data_set->graph, graph_file, FALSE); } if (dot_file != NULL) { create_dotfile(data_set, dot_file, all_actions); } if (quiet == FALSE) { GListPtr gIter = NULL; quiet_log("%sTransition Summary:\n", show_scores || show_utilization || modified ? "\n" : ""); fflush(stdout); LogNodeActions(data_set, TRUE); for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; LogActions(rsc, data_set, TRUE); } } } rc = pcmk_ok; if (simulate) { if (run_simulation(data_set, global_cib, op_fail, quiet) != pcmk_ok) { rc = pcmk_err_generic; } if(quiet == FALSE) { get_date(data_set, true); quiet_log("\nRevised cluster status:\n"); set_bit(data_set->flags, pe_flag_stdout); cluster_status(data_set); print_cluster_status(data_set, 0); } } done: pe_free_working_set(data_set); global_cib->cmds->signoff(global_cib); cib_delete(global_cib); free(use_date); fflush(stderr); if (temp_shadow) { unlink(temp_shadow); free(temp_shadow); } crm_exit(crm_errno2exit(rc)); } diff --git a/tools/crmadmin.c b/tools/crmadmin.c index fe29d79913..b2c48413ca 100644 --- a/tools/crmadmin.c +++ b/tools/crmadmin.c @@ -1,530 +1,530 @@ /* * Copyright 2004-2020 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include static int message_timer_id = -1; static int message_timeout_ms = 30 * 1000; static GMainLoop *mainloop = NULL; static crm_ipc_t *crmd_channel = NULL; static char *admin_uuid = NULL; gboolean do_init(void); int do_work(void); void crmadmin_ipc_connection_destroy(gpointer user_data); int admin_msg_callback(const char *buffer, ssize_t length, gpointer userdata); int do_find_node_list(xmlNode * xml_node); gboolean admin_message_timeout(gpointer data); static gboolean BE_VERBOSE = FALSE; static int expected_responses = 1; static gboolean BASH_EXPORT = FALSE; static gboolean DO_HEALTH = FALSE; static gboolean DO_RESET = FALSE; static gboolean DO_RESOURCE = FALSE; static gboolean DO_ELECT_DC = FALSE; static gboolean DO_WHOIS_DC = FALSE; static gboolean DO_NODE_LIST = FALSE; static gboolean BE_SILENT = FALSE; static gboolean DO_RESOURCE_LIST = FALSE; static const char *crmd_operation = NULL; static char *dest_node = NULL; static crm_exit_t exit_code = CRM_EX_OK; static const char *sys_to = NULL; static pcmk__cli_option_t long_options[] = { // long option, argument type, storage, short option, description, flags { "help", no_argument, NULL, '?', "\tThis text", pcmk__option_default }, { "version", no_argument, NULL, '$', "\tVersion information", pcmk__option_default }, { "quiet", no_argument, NULL, 'q', "\tDisplay only the essential query information", pcmk__option_default }, { "verbose", no_argument, NULL, 'V', "\tIncrease debug output", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\nCommands:", pcmk__option_default }, /* daemon options */ { "status", required_argument, NULL, 'S', "Display the status of the specified node.", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\n\tResult is state of node's internal finite state machine, which " "can be useful for debugging\n", pcmk__option_default }, { "dc_lookup", no_argument, NULL, 'D', "Display the uname of the node co-ordinating the cluster.", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "\n\tThis is an internal detail rarely useful to administrators " "except when deciding on which node to examine the logs.\n", pcmk__option_default }, { "nodes", no_argument, NULL, 'N', "\tDisplay the uname of all member nodes", pcmk__option_default }, { "election", no_argument, NULL, 'E', "(Advanced) Start an election for the cluster co-ordinator", pcmk__option_default }, { "kill", required_argument, NULL, 'K', "(Advanced) Stop controller (not rest of cluster stack) on " "specified node", pcmk__option_default }, { "health", no_argument, NULL, 'H', NULL, pcmk__option_hidden }, { "-spacer-", no_argument, NULL, '-', "\nAdditional Options:", pcmk__option_default }, { XML_ATTR_TIMEOUT, required_argument, NULL, 't', "Time (in milliseconds) to wait before declaring the operation failed", pcmk__option_default }, { "bash-export", no_argument, NULL, 'B', "Create Bash export entries of the form 'export uname=uuid'\n", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', "Notes:", pcmk__option_default }, { "-spacer-", no_argument, NULL, '-', " The -K and -E commands are rarely used and may be removed in " "future versions.", pcmk__option_default }, { 0, 0, 0, 0 } }; int main(int argc, char **argv) { int option_index = 0; int argerr = 0; int flag; crm_log_cli_init("crmadmin"); pcmk__set_cli_options(NULL, " [options]", long_options, "query and manage the Pacemaker controller"); if (argc < 2) { pcmk__cli_help('?', CRM_EX_USAGE); } while (1) { flag = pcmk__next_cli_option(argc, argv, &option_index, NULL); if (flag == -1) break; switch (flag) { case 'V': BE_VERBOSE = TRUE; crm_bump_log_level(argc, argv); break; case 't': message_timeout_ms = atoi(optarg); if (message_timeout_ms < 1) { message_timeout_ms = 30 * 1000; } break; case '$': case '?': pcmk__cli_help(flag, CRM_EX_OK); break; case 'D': DO_WHOIS_DC = TRUE; break; case 'B': BASH_EXPORT = TRUE; break; case 'K': DO_RESET = TRUE; crm_trace("Option %c => %s", flag, optarg); dest_node = strdup(optarg); crmd_operation = CRM_OP_LOCAL_SHUTDOWN; break; case 'q': BE_SILENT = TRUE; break; case 'S': DO_HEALTH = TRUE; crm_trace("Option %c => %s", flag, optarg); dest_node = strdup(optarg); break; case 'E': DO_ELECT_DC = TRUE; break; case 'N': DO_NODE_LIST = TRUE; break; case 'H': DO_HEALTH = TRUE; break; default: printf("Argument code 0%o (%c) is not (?yet?) supported\n", flag, 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) { pcmk__cli_help('?', CRM_EX_USAGE); } if (do_init()) { int res = 0; res = do_work(); if (res > 0) { /* wait for the reply by creating a mainloop and running it until * the callbacks are invoked... */ mainloop = g_main_loop_new(NULL, FALSE); crm_trace("Waiting for %d replies from the local CRM", expected_responses); message_timer_id = g_timeout_add(message_timeout_ms, admin_message_timeout, NULL); g_main_loop_run(mainloop); } else if (res < 0) { crm_err("No message to send"); exit_code = CRM_EX_ERROR; } } else { crm_warn("Init failed, could not perform requested operations"); exit_code = CRM_EX_UNAVAILABLE; } crm_trace("%s exiting normally", crm_system_name); return exit_code; } int do_work(void) { int ret = 1; /* construct the request */ xmlNode *msg_data = NULL; gboolean all_is_good = TRUE; if (DO_HEALTH == TRUE) { crm_trace("Querying the system"); sys_to = CRM_SYSTEM_DC; if (dest_node != NULL) { sys_to = CRM_SYSTEM_CRMD; crmd_operation = CRM_OP_PING; if (BE_VERBOSE) { expected_responses = 1; } } else { crm_info("Cluster-wide health not available yet"); all_is_good = FALSE; } } else if (DO_ELECT_DC) { /* tell the local node to initiate an election */ dest_node = NULL; sys_to = CRM_SYSTEM_CRMD; crmd_operation = CRM_OP_VOTE; ret = 0; /* no return message */ } else if (DO_WHOIS_DC) { dest_node = NULL; sys_to = CRM_SYSTEM_DC; crmd_operation = CRM_OP_PING; } else if (DO_NODE_LIST) { cib_t *the_cib = cib_new(); xmlNode *output = NULL; int rc = the_cib->cmds->signon(the_cib, crm_system_name, cib_command); if (rc != pcmk_ok) { return -1; } rc = the_cib->cmds->query(the_cib, NULL, &output, cib_scope_local | cib_sync_call); if(rc == pcmk_ok) { do_find_node_list(output); free_xml(output); } the_cib->cmds->signoff(the_cib); crm_exit(crm_errno2exit(rc)); } else if (DO_RESET) { /* tell dest_node to initiate the shutdown procedure * * if dest_node is NULL, the request will be sent to the * local node */ sys_to = CRM_SYSTEM_CRMD; ret = 0; /* no return message */ } else { crm_err("Unknown options"); all_is_good = FALSE; } if (all_is_good == FALSE) { crm_err("Creation of request failed. No message to send"); return -1; } /* send it */ if (crmd_channel == NULL) { crm_err("The IPC connection is not valid, cannot send anything"); return -1; } if (sys_to == NULL) { if (dest_node != NULL) { sys_to = CRM_SYSTEM_CRMD; } else { sys_to = CRM_SYSTEM_DC; } } { xmlNode *cmd = create_request(crmd_operation, msg_data, dest_node, sys_to, crm_system_name, admin_uuid); crm_ipc_send(crmd_channel, cmd, 0, 0, NULL); free_xml(cmd); } return ret; } void crmadmin_ipc_connection_destroy(gpointer user_data) { crm_err("Connection to controller was terminated"); if (mainloop) { g_main_loop_quit(mainloop); } else { crm_exit(CRM_EX_DISCONNECT); } } struct ipc_client_callbacks crm_callbacks = { .dispatch = admin_msg_callback, .destroy = crmadmin_ipc_connection_destroy }; gboolean do_init(void) { mainloop_io_t *source = mainloop_add_ipc_client(CRM_SYSTEM_CRMD, G_PRIORITY_DEFAULT, 0, NULL, &crm_callbacks); - admin_uuid = crm_getpid_s(); + admin_uuid = pcmk__getpid_s(); crmd_channel = mainloop_get_ipc_client(source); if (DO_RESOURCE || DO_RESOURCE_LIST || DO_NODE_LIST) { return TRUE; } else if (crmd_channel != NULL) { xmlNode *xml = create_hello_message(admin_uuid, crm_system_name, "0", "1"); crm_ipc_send(crmd_channel, xml, 0, 0, NULL); return TRUE; } return FALSE; } static bool validate_crm_message(xmlNode * msg, const char *sys, const char *uuid, const char *msg_type) { const char *type = NULL; const char *crm_msg_reference = NULL; if (msg == NULL) { return FALSE; } type = crm_element_value(msg, F_CRM_MSG_TYPE); crm_msg_reference = crm_element_value(msg, XML_ATTR_REFERENCE); if (type == NULL) { crm_info("No message type defined."); return FALSE; } else if (msg_type != NULL && strcasecmp(msg_type, type) != 0) { crm_info("Expecting a (%s) message but received a (%s).", msg_type, type); return FALSE; } if (crm_msg_reference == NULL) { crm_info("No message crm_msg_reference defined."); return FALSE; } return TRUE; } int admin_msg_callback(const char *buffer, ssize_t length, gpointer userdata) { static int received_responses = 0; xmlNode *xml = string2xml(buffer); received_responses++; g_source_remove(message_timer_id); crm_log_xml_trace(xml, "ipc"); if (xml == NULL) { crm_info("XML in IPC message was not valid... " "discarding."); } else if (validate_crm_message(xml, crm_system_name, admin_uuid, XML_ATTR_RESPONSE) == FALSE) { crm_trace("Message was not a CRM response. Discarding."); } else if (DO_HEALTH) { xmlNode *data = get_message_xml(xml, F_CRM_DATA); const char *state = crm_element_value(data, XML_PING_ATTR_CRMDSTATE); printf("Status of %s@%s: %s (%s)\n", crm_element_value(data, XML_PING_ATTR_SYSFROM), crm_element_value(xml, F_CRM_HOST_FROM), state, crm_element_value(data, XML_PING_ATTR_STATUS)); if (BE_SILENT && state != NULL) { fprintf(stderr, "%s\n", state); } } else if (DO_WHOIS_DC) { const char *dc = crm_element_value(xml, F_CRM_HOST_FROM); printf("Designated Controller is: %s\n", dc); if (BE_SILENT && dc != NULL) { fprintf(stderr, "%s\n", dc); } crm_exit(CRM_EX_OK); } free_xml(xml); if (received_responses >= expected_responses) { crm_trace("Received expected number (%d) of replies, exiting normally", expected_responses); crm_exit(CRM_EX_OK); } message_timer_id = g_timeout_add(message_timeout_ms, admin_message_timeout, NULL); return 0; } gboolean admin_message_timeout(gpointer data) { fprintf(stderr, "No messages received in %d seconds.. aborting\n", (int)message_timeout_ms / 1000); crm_err("No messages received in %d seconds", (int)message_timeout_ms / 1000); exit_code = CRM_EX_TIMEOUT; g_main_loop_quit(mainloop); return FALSE; } int do_find_node_list(xmlNode * xml_node) { int found = 0; xmlNode *node = NULL; xmlNode *nodes = get_object_root(XML_CIB_TAG_NODES, xml_node); for (node = __xml_first_child_element(nodes); node != NULL; node = __xml_next_element(node)) { if (crm_str_eq((const char *)node->name, XML_CIB_TAG_NODE, TRUE)) { if (BASH_EXPORT) { printf("export %s=%s\n", crm_element_value(node, XML_ATTR_UNAME), crm_element_value(node, XML_ATTR_ID)); } else { printf("%s node: %s (%s)\n", crm_element_value(node, XML_ATTR_TYPE), crm_element_value(node, XML_ATTR_UNAME), crm_element_value(node, XML_ATTR_ID)); } found++; } } if (found == 0) { printf("NO nodes configured\n"); } return found; }