Exemplo n.º 1
0
void
abort_transition_graph(int abort_priority, enum transition_action abort_action,
                       const char *abort_text, xmlNode * reason, const char *fn, int line)
{
    int add[] = { 0, 0, 0 };
    int del[] = { 0, 0, 0 };
    int level = LOG_INFO;
    xmlNode *diff = NULL;
    xmlNode *change = NULL;

    CRM_CHECK(transition_graph != NULL, return);

    switch (fsa_state) {
        case S_STARTING:
        case S_PENDING:
        case S_NOT_DC:
        case S_HALT:
        case S_ILLEGAL:
        case S_STOPPING:
        case S_TERMINATE:
            crm_info("Abort %s suppressed: state=%s (complete=%d)",
                     abort_text, fsa_state2string(fsa_state), transition_graph->complete);
            return;
        default:
            break;
    }

    /* Make sure any queued calculations are discarded ASAP */
    free(fsa_pe_ref);
    fsa_pe_ref = NULL;

    if (transition_graph->complete == FALSE) {
        if(update_abort_priority(transition_graph, abort_priority, abort_action, abort_text)) {
            level = LOG_NOTICE;
        }
    }

    if(reason) {
        xmlNode *search = NULL;

        for(search = reason; search; search = search->parent) {
            if (safe_str_eq(XML_TAG_DIFF, TYPE(search))) {
                diff = search;
                break;
            }
        }

        if(diff) {
            xml_patch_versions(diff, add, del);
            for(search = reason; search; search = search->parent) {
                if (safe_str_eq(XML_DIFF_CHANGE, TYPE(search))) {
                    change = search;
                    break;
                }
            }
        }
    }

    if(reason == NULL) {
        do_crm_log(level, "Transition aborted: %s (source=%s:%d, %d)",
                   abort_text, fn, line, transition_graph->complete);

    } else if(change == NULL) {
        char *local_path = xml_get_path(reason);

        do_crm_log(level, "Transition aborted by %s.%s: %s (cib=%d.%d.%d, source=%s:%d, path=%s, %d)",
                   TYPE(reason), ID(reason), abort_text, add[0], add[1], add[2], fn, line, local_path, transition_graph->complete);
        free(local_path);

    } else {
        const char *kind = NULL;
        const char *op = crm_element_value(change, XML_DIFF_OP);
        const char *path = crm_element_value(change, XML_DIFF_PATH);

        if(change == reason) {
            if(strcmp(op, "create") == 0) {
                reason = reason->children;

            } else if(strcmp(op, "modify") == 0) {
                reason = first_named_child(reason, XML_DIFF_RESULT);
                if(reason) {
                    reason = reason->children;
                }
            }
        }

        kind = TYPE(reason);
        if(strcmp(op, "delete") == 0) {
            const char *shortpath = strrchr(path, '/');

            do_crm_log(level, "Transition aborted by deletion of %s: %s (cib=%d.%d.%d, source=%s:%d, path=%s, %d)",
                       shortpath?shortpath+1:path, abort_text, add[0], add[1], add[2], fn, line, path, transition_graph->complete);

        } else if (safe_str_eq(XML_CIB_TAG_NVPAIR, kind)) { 
            do_crm_log(level, "Transition aborted by %s, %s=%s: %s (%s cib=%d.%d.%d, source=%s:%d, path=%s, %d)",
                       crm_element_value(reason, XML_ATTR_ID),
                       crm_element_value(reason, XML_NVPAIR_ATTR_NAME),
                       crm_element_value(reason, XML_NVPAIR_ATTR_VALUE),
                       abort_text, op, add[0], add[1], add[2], fn, line, path, transition_graph->complete);

        } else if (safe_str_eq(XML_LRM_TAG_RSC_OP, kind)) {
            const char *magic = crm_element_value(reason, XML_ATTR_TRANSITION_MAGIC);

            do_crm_log(level, "Transition aborted by %s '%s' on %s: %s (magic=%s, cib=%d.%d.%d, source=%s:%d, %d)",
                       crm_element_value(reason, XML_LRM_ATTR_TASK_KEY), op,
                       crm_element_value(reason, XML_LRM_ATTR_TARGET), abort_text,
                       magic, add[0], add[1], add[2], fn, line, transition_graph->complete);

        } else if (safe_str_eq(XML_CIB_TAG_STATE, kind)
                   || safe_str_eq(XML_CIB_TAG_NODE, kind)) {
            const char *uname = crm_peer_uname(ID(reason));

            do_crm_log(level, "Transition aborted by %s '%s' on %s: %s (cib=%d.%d.%d, source=%s:%d, %d)",
                       kind, op, uname ? uname : ID(reason), abort_text,
                       add[0], add[1], add[2], fn, line, transition_graph->complete);

        } else {
            do_crm_log(level, "Transition aborted by %s.%s '%s': %s (cib=%d.%d.%d, source=%s:%d, path=%s, %d)",
                       TYPE(reason), ID(reason), op?op:"change", abort_text, add[0], add[1], add[2], fn, line, path, transition_graph->complete);
        }
    }

    if (transition_graph->complete) {
        if (transition_timer->period_ms > 0) {
            crm_timer_stop(transition_timer);
            crm_timer_start(transition_timer);
        } else {
            register_fsa_input(C_FSA_INTERNAL, I_PE_CALC, NULL);
        }
        return;
    }

    mainloop_set_trigger(transition_trigger);
}
Exemplo n.º 2
0
/*!
 * \internal
 * \brief Update failure-related node attributes if warranted
 *
 * \param[in] event            XML describing operation that (maybe) failed
 * \param[in] event_node_uuid  Node that event occurred on
 * \param[in] rc               Actual operation return code
 * \param[in] target_rc        Expected operation return code
 * \param[in] do_update        If TRUE, do update regardless of operation type
 * \param[in] ignore_failures  If TRUE, update last failure but not fail count
 *
 * \return TRUE if this was not a direct nack, success or lrm status refresh
 */
static gboolean
update_failcount(xmlNode * event, const char *event_node_uuid, int rc,
                 int target_rc, gboolean do_update, gboolean ignore_failures)
{
    int interval = 0;

    char *task = NULL;
    char *rsc_id = NULL;

    const char *value = NULL;
    const char *id = crm_element_value(event, XML_LRM_ATTR_TASK_KEY);
    const char *on_uname = crm_peer_uname(event_node_uuid);
    const char *origin = crm_element_value(event, XML_ATTR_ORIGIN);

    /* Nothing needs to be done for success, lrm status refresh,
     * or direct nack (internal code for "busy, try again")
     */
    if ((rc == CRM_DIRECT_NACK_RC) || (rc == target_rc)) {
        return FALSE;
    } else if (safe_str_eq(origin, "build_active_RAs")) {
        crm_debug("No update for %s (rc=%d) on %s: Old failure from lrm status refresh",
                  id, rc, on_uname);
        return FALSE;
    }

    /* Sanity check */
    CRM_CHECK(on_uname != NULL, return TRUE);
    CRM_CHECK(parse_op_key(id, &rsc_id, &task, &interval),
              crm_err("Couldn't parse: %s", ID(event)); goto bail);
    CRM_CHECK(task != NULL, goto bail);
    CRM_CHECK(rsc_id != NULL, goto bail);

    /* Decide whether update is necessary and what value to use */
    if ((interval > 0) || safe_str_eq(task, CRMD_ACTION_PROMOTE)
        || safe_str_eq(task, CRMD_ACTION_DEMOTE)) {
        do_update = TRUE;

    } else if (safe_str_eq(task, CRMD_ACTION_START)) {
        do_update = TRUE;
        if (failed_start_offset == NULL) {
            failed_start_offset = strdup(INFINITY_S);
        }
        value = failed_start_offset;

    } else if (safe_str_eq(task, CRMD_ACTION_STOP)) {
        do_update = TRUE;
        if (failed_stop_offset == NULL) {
            failed_stop_offset = strdup(INFINITY_S);
        }
        value = failed_stop_offset;
    }

    /* Fail count will be either incremented or set to infinity */
    if (value == NULL || safe_str_neq(value, INFINITY_S)) {
        value = XML_NVPAIR_ATTR_VALUE "++";
    }

    if (do_update) {
        char *now = crm_itoa(time(NULL));
        char *attr_name = NULL;
        gboolean is_remote_node = FALSE;

        if (g_hash_table_lookup(crm_remote_peer_cache, event_node_uuid)) {
            is_remote_node = TRUE;
        }

        crm_info("Updating %s for %s on %s after failed %s: rc=%d (update=%s, time=%s)",
                 (ignore_failures? "last failure" : "failcount"),
                 rsc_id, on_uname, task, rc, value, now);

        /* Update the fail count, if we're not ignoring failures */
        if (!ignore_failures) {
            attr_name = crm_concat("fail-count", rsc_id, '-');
            update_attrd(on_uname, attr_name, value, NULL, is_remote_node);
            free(attr_name);
        }

        /* Update the last failure time (even if we're ignoring failures,
         * so that failure can still be detected and shown, e.g. by crm_mon)
         */
        attr_name = crm_concat("last-failure", rsc_id, '-');
        update_attrd(on_uname, attr_name, now, NULL, is_remote_node);
        free(attr_name);

        free(now);
    }

  bail:
    free(rsc_id);
    free(task);
    return TRUE;
}
Exemplo n.º 3
0
static gboolean
update_failcount(xmlNode * event, const char *event_node_uuid, int rc, int target_rc, gboolean do_update)
{
    int interval = 0;

    char *task = NULL;
    char *rsc_id = NULL;
    char *attr_name = NULL;

    const char *value = NULL;
    const char *id = crm_element_value(event, XML_LRM_ATTR_TASK_KEY);
    const char *on_uname = crm_peer_uname(event_node_uuid);
    const char *origin = crm_element_value(event, XML_ATTR_ORIGIN);

    if (rc == 99) {
        /* this is an internal code for "we're busy, try again" */
        return FALSE;

    } else if (rc == target_rc) {
        return FALSE;
    }

    if (safe_str_eq(origin, "build_active_RAs")) {
        crm_debug("No update for %s (rc=%d) on %s: Old failure from lrm status refresh",
                  id, rc, on_uname);
        return FALSE;
    }

    CRM_CHECK(on_uname != NULL, return TRUE);

    if (failed_stop_offset == NULL) {
        failed_stop_offset = strdup(INFINITY_S);
    }

    if (failed_start_offset == NULL) {
        failed_start_offset = strdup(INFINITY_S);
    }

    CRM_CHECK(parse_op_key(id, &rsc_id, &task, &interval), crm_err("Couldn't parse: %s", ID(event));
              goto bail);
    CRM_CHECK(task != NULL, goto bail);
    CRM_CHECK(rsc_id != NULL, goto bail);

    if (do_update || interval > 0) {
        do_update = TRUE;

    } else if (safe_str_eq(task, CRMD_ACTION_START)) {
        do_update = TRUE;
        value = failed_start_offset;

    } else if (safe_str_eq(task, CRMD_ACTION_STOP)) {
        do_update = TRUE;
        value = failed_stop_offset;

    } else if (safe_str_eq(task, CRMD_ACTION_STOP)) {
        do_update = TRUE;
        value = failed_stop_offset;

    } else if (safe_str_eq(task, CRMD_ACTION_PROMOTE)) {
        do_update = TRUE;

    } else if (safe_str_eq(task, CRMD_ACTION_DEMOTE)) {
        do_update = TRUE;
    }

    if (value == NULL || safe_str_neq(value, INFINITY_S)) {
        value = XML_NVPAIR_ATTR_VALUE "++";
    }

    if (do_update) {
        char *now = crm_itoa(time(NULL));
        gboolean is_remote_node = FALSE;

        if (g_hash_table_lookup(crm_remote_peer_cache, event_node_uuid)) {
            is_remote_node = TRUE;
        }

        crm_warn("Updating failcount for %s on %s after failed %s:"
                 " rc=%d (update=%s, time=%s)", rsc_id, on_uname, task, rc, value, now);

        attr_name = crm_concat("fail-count", rsc_id, '-');
        update_attrd(on_uname, attr_name, value, NULL, is_remote_node);
        free(attr_name);

        attr_name = crm_concat("last-failure", rsc_id, '-');
        update_attrd(on_uname, attr_name, now, NULL, is_remote_node);
        free(attr_name);

        free(now);
    }

  bail:
    free(rsc_id);
    free(task);
    return TRUE;
}
Exemplo n.º 4
0
void
abort_transition_graph(int abort_priority, enum transition_action abort_action,
                       const char *abort_text, xmlNode * reason, const char *fn, int line)
{
    const char *magic = NULL;

    CRM_CHECK(transition_graph != NULL, return);

    if (reason) {
        int diff_add_updates = 0;
        int diff_add_epoch = 0;
        int diff_add_admin_epoch = 0;

        int diff_del_updates = 0;
        int diff_del_epoch = 0;
        int diff_del_admin_epoch = 0;

        const char *uname = "";
        xmlNode *search = reason;
        xmlNode *diff = get_xpath_object("//" F_CIB_UPDATE_RESULT "//diff", reason, LOG_DEBUG_2);

        magic = crm_element_value(reason, XML_ATTR_TRANSITION_MAGIC);

        while(search) {
            const char *kind = TYPE(search);

            if (safe_str_eq(XML_CIB_TAG_STATE, kind)
               || safe_str_eq(XML_CIB_TAG_NODE, kind)) {

                uname = crm_peer_uname(ID(search));
                break;
            }
            search = search->parent;
        }

        if (diff) {
            cib_diff_version_details(diff,
                                     &diff_add_admin_epoch, &diff_add_epoch, &diff_add_updates,
                                     &diff_del_admin_epoch, &diff_del_epoch, &diff_del_updates);
            if (crm_str_eq(TYPE(reason), XML_CIB_TAG_NVPAIR, TRUE)) {
                crm_info
                    ("%s:%d - Triggered transition abort (complete=%d, node=%s, tag=%s, id=%s, name=%s, value=%s, magic=%s, cib=%d.%d.%d) : %s",
                     fn, line, transition_graph->complete, uname, TYPE(reason), ID(reason), NAME(reason),
                     VALUE(reason), magic ? magic : "NA", diff_add_admin_epoch, diff_add_epoch,
                     diff_add_updates, abort_text);
            } else {
                crm_info
                    ("%s:%d - Triggered transition abort (complete=%d, node=%s, tag=%s, id=%s, magic=%s, cib=%d.%d.%d) : %s",
                     fn, line, transition_graph->complete, uname, TYPE(reason), ID(reason),
                     magic ? magic : "NA", diff_add_admin_epoch, diff_add_epoch, diff_add_updates,
                     abort_text);
            }

        } else {
            crm_info
                ("%s:%d - Triggered transition abort (complete=%d, node=%s, tag=%s, id=%s, magic=%s) : %s",
                 fn, line, transition_graph->complete, uname, TYPE(reason), ID(reason),
                 magic ? magic : "NA", abort_text);
        }

    } else {
        crm_info("%s:%d - Triggered transition abort (complete=%d) : %s",
                 fn, line, transition_graph->complete, abort_text);
    }

    switch (fsa_state) {
        case S_STARTING:
        case S_PENDING:
        case S_NOT_DC:
        case S_HALT:
        case S_ILLEGAL:
        case S_STOPPING:
        case S_TERMINATE:
            crm_info("Abort suppressed: state=%s (complete=%d)",
                     fsa_state2string(fsa_state), transition_graph->complete);
            return;
        default:
            break;
    }

    if (magic == NULL && reason != NULL) {
        crm_log_xml_debug(reason, "Cause");
    }

    /* Make sure any queued calculations are discarded ASAP */
    free(fsa_pe_ref);
    fsa_pe_ref = NULL;

    if (transition_graph->complete) {
        if (transition_timer->period_ms > 0) {
            crm_timer_stop(transition_timer);
            crm_timer_start(transition_timer);
        } else {
            register_fsa_input(C_FSA_INTERNAL, I_PE_CALC, NULL);
        }
        return;
    }

    update_abort_priority(transition_graph, abort_priority, abort_action, abort_text);

    mainloop_set_trigger(transition_trigger);
}