2 monitoring links to all other nodes to detect dead nodes
5 Copyright (C) Ronnie Sahlberg 2007
7 This program is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 3 of the License, or
10 (at your option) any later version.
12 This program is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with this program; if not, see <http://www.gnu.org/licenses/>.
22 #include "lib/events/events.h"
23 #include "system/filesys.h"
24 #include "system/wait.h"
25 #include "../include/ctdb_private.h"
28 see if any nodes are dead
30 static void ctdb_check_for_dead_nodes(struct event_context *ev, struct timed_event *te,
31 struct timeval t, void *private_data)
33 struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
36 if (ctdb->monitoring_mode == CTDB_MONITORING_DISABLED) {
37 event_add_timed(ctdb->ev, ctdb->monitor_context,
38 timeval_current_ofs(ctdb->tunable.keepalive_interval, 0),
39 ctdb_check_for_dead_nodes, ctdb);
43 /* send a keepalive to all other nodes, unless */
44 for (i=0;i<ctdb->num_nodes;i++) {
45 struct ctdb_node *node = ctdb->nodes[i];
46 if (node->vnn == ctdb->vnn) {
50 if (node->flags & NODE_FLAGS_DISCONNECTED) {
51 /* it might have come alive again */
52 if (node->rx_cnt != 0) {
53 ctdb_node_connected(node);
59 if (node->rx_cnt == 0) {
67 if (node->dead_count >= ctdb->tunable.keepalive_limit) {
68 DEBUG(0,("dead count reached for node %u\n", node->vnn));
70 ctdb_send_keepalive(ctdb, node->vnn);
71 /* maybe tell the transport layer to kill the
77 if (node->tx_cnt == 0) {
78 DEBUG(5,("sending keepalive to %u\n", node->vnn));
79 ctdb_send_keepalive(ctdb, node->vnn);
85 event_add_timed(ctdb->ev, ctdb->monitor_context,
86 timeval_current_ofs(ctdb->tunable.keepalive_interval, 0),
87 ctdb_check_for_dead_nodes, ctdb);
90 static void ctdb_check_health(struct event_context *ev, struct timed_event *te,
91 struct timeval t, void *private_data);
94 called when a health monitoring event script finishes
96 static void ctdb_health_callback(struct ctdb_context *ctdb, int status, void *p)
98 struct ctdb_node *node = ctdb->nodes[ctdb->vnn];
100 struct ctdb_node_flag_change c;
102 event_add_timed(ctdb->ev, ctdb->monitor_context,
103 timeval_current_ofs(ctdb->tunable.monitor_interval, 0),
104 ctdb_check_health, ctdb);
106 if (status != 0 && !(node->flags & NODE_FLAGS_UNHEALTHY)) {
107 DEBUG(0,("monitor event failed - disabling node\n"));
108 node->flags |= NODE_FLAGS_UNHEALTHY;
109 } else if (status == 0 && (node->flags & NODE_FLAGS_UNHEALTHY)) {
110 DEBUG(0,("monitor event OK - node re-enabled\n"));
111 ctdb->nodes[ctdb->vnn]->flags &= ~NODE_FLAGS_UNHEALTHY;
118 c.flags = node->flags;
120 data.dptr = (uint8_t *)&c;
121 data.dsize = sizeof(c);
123 /* tell the other nodes that something has changed */
124 ctdb_daemon_send_message(ctdb, CTDB_BROADCAST_CONNECTED,
125 CTDB_SRVID_NODE_FLAGS_CHANGED, data);
131 see if the event scripts think we are healthy
133 static void ctdb_check_health(struct event_context *ev, struct timed_event *te,
134 struct timeval t, void *private_data)
136 struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
139 if (ctdb->monitoring_mode == CTDB_MONITORING_DISABLED) {
140 event_add_timed(ctdb->ev, ctdb->monitor_context,
141 timeval_current_ofs(ctdb->tunable.monitor_interval, 0),
142 ctdb_check_health, ctdb);
146 ret = ctdb_event_script_callback(ctdb,
147 timeval_current_ofs(ctdb->tunable.script_timeout, 0),
148 ctdb->monitor_context, ctdb_health_callback, ctdb, "monitor");
150 DEBUG(0,("Unable to launch monitor event script\n"));
151 event_add_timed(ctdb->ev, ctdb->monitor_context,
152 timeval_current_ofs(ctdb->tunable.monitor_interval, 0),
153 ctdb_check_health, ctdb);
157 /* stop any monitoring */
158 void ctdb_stop_monitoring(struct ctdb_context *ctdb)
160 talloc_free(ctdb->monitor_context);
161 ctdb->monitor_context = talloc_new(ctdb);
162 CTDB_NO_MEMORY_FATAL(ctdb, ctdb->monitor_context);
166 start watching for nodes that might be dead
168 void ctdb_start_monitoring(struct ctdb_context *ctdb)
170 struct timed_event *te;
172 ctdb_stop_monitoring(ctdb);
174 te = event_add_timed(ctdb->ev, ctdb->monitor_context,
175 timeval_current_ofs(ctdb->tunable.keepalive_interval, 0),
176 ctdb_check_for_dead_nodes, ctdb);
177 CTDB_NO_MEMORY_FATAL(ctdb, te);
179 te = event_add_timed(ctdb->ev, ctdb->monitor_context,
180 timeval_current_ofs(ctdb->tunable.monitor_interval, 0),
181 ctdb_check_health, ctdb);
182 CTDB_NO_MEMORY_FATAL(ctdb, te);
187 modify flags on a node
189 int32_t ctdb_control_modflags(struct ctdb_context *ctdb, TDB_DATA indata)
191 struct ctdb_node_modflags *m = (struct ctdb_node_modflags *)indata.dptr;
193 struct ctdb_node_flag_change c;
194 struct ctdb_node *node = ctdb->nodes[ctdb->vnn];
195 uint32_t old_flags = node->flags;
197 node->flags |= m->set;
198 node->flags &= ~m->clear;
200 if (node->flags == old_flags) {
205 DEBUG(0, ("Control modflags on node %u - flags now 0x%x\n", ctdb->vnn, node->flags));
207 /* if we have been banned, go into recovery mode */
209 c.flags = node->flags;
211 data.dptr = (uint8_t *)&c;
212 data.dsize = sizeof(c);
214 /* tell the other nodes that something has changed */
215 ctdb_daemon_send_message(ctdb, CTDB_BROADCAST_CONNECTED,
216 CTDB_SRVID_NODE_FLAGS_CHANGED, data);
218 if ((node->flags & NODE_FLAGS_BANNED) && !(old_flags & NODE_FLAGS_BANNED)) {
219 /* make sure we are frozen */
220 DEBUG(0,("This node has been banned - forcing freeze and recovery\n"));
221 ctdb_start_freeze(ctdb);
222 ctdb_release_all_ips(ctdb);
223 ctdb->recovery_mode = CTDB_RECOVERY_ACTIVE;