prevent a deadly embrace between smbd and ctdbd by moving the calling

of the startup event scripts after the point where recovery has
started and the node is in normal operation

This makes the 'startup' script just a special type of the 'monitor'
script which is called first

diff --git a/include/ctdb_private.h b/include/ctdb_private.h
index 7b98683e9ffe22244a2ae1f9e89a540cc88b52d7..cb76bb0074cb68a6a69dbf98e931859b0ebe5d6a 100644 (file)
--- a/include/ctdb_private.h
+++ b/include/ctdb_private.h
@@ -366,6 +366,7 @@ struct ctdb_context {
        const char *event_script_dir;
        const char *default_public_interface;
        pid_t recoverd_pid;
+       bool done_startup;
 };
 
 struct ctdb_db_context {

diff --git a/server/ctdb_daemon.c b/server/ctdb_daemon.c
index 671a7e8be2e40bd71c050432d8102ae671f63d61..8f66ade92767432452976051b89d37eec92762a5 100644 (file)
--- a/server/ctdb_daemon.c
+++ b/server/ctdb_daemon.c
@@ -68,13 +68,8 @@ static void print_exit_message(void)
 
 
 /* called when the "startup" event script has finished */
-static void ctdb_start_transport(struct ctdb_context *ctdb, int status, void *p)
+static void ctdb_start_transport(struct ctdb_context *ctdb)
 {
-       if (status != 0) {
-               DEBUG(0,("startup event failed!\n"));
-               ctdb_fatal(ctdb, "startup event script failed");                
-       }
-
        /* start the transport running */
        if (ctdb->methods->start(ctdb) != 0) {
                DEBUG(0,("transport failed to start!\n"));
@@ -664,12 +659,8 @@ int ctdb_start_daemon(struct ctdb_context *ctdb, bool do_fork)
        /* release any IPs we hold from previous runs of the daemon */
        ctdb_release_all_ips(ctdb);
 
-       ret = ctdb_event_script_callback(ctdb, timeval_zero(), ctdb, 
-                                        ctdb_start_transport, NULL, "startup");
-       if (ret != 0) {
-               DEBUG(0,("Failed startup event script\n"));
-               return -1;
-       }
+       /* start the transport going */
+       ctdb_start_transport(ctdb);
 
        /* go into a wait loop to allow other nodes to complete */
        event_loop_wait(ctdb->ev);

diff --git a/server/ctdb_monitor.c b/server/ctdb_monitor.c
index c96099e76c8e5ae116c03af23e187ccc9e4df549..52ecc7c7136eff33e8ffec7e315dd509922acf3f 100644 (file)
--- a/server/ctdb_monitor.c
+++ b/server/ctdb_monitor.c
@@ -137,6 +137,31 @@ static void ctdb_health_callback(struct ctdb_context *ctdb, int status, void *p)
 }
 
 
+/*
+  called when the startup event script finishes
+ */
+static void ctdb_startup_callback(struct ctdb_context *ctdb, int status, void *p)
+{
+       if (status != 0) {
+               DEBUG(0,("startup event failed\n"));
+       } else if (status == 0) {
+               DEBUG(0,("startup event OK - enabling monitoring\n"));
+               ctdb->done_startup = true;
+       }
+
+       if (ctdb->done_startup) {
+               event_add_timed(ctdb->ev, ctdb->monitor_context, 
+                               timeval_zero(),
+                               ctdb_check_health, ctdb);
+       } else {
+               event_add_timed(ctdb->ev, ctdb->monitor_context, 
+                               timeval_current_ofs(ctdb->tunable.monitor_interval, 0), 
+                               ctdb_check_health, ctdb);
+       }
+
+}
+
+
 /*
   see if the event scripts think we are healthy
  */
@@ -146,16 +171,25 @@ static void ctdb_check_health(struct event_context *ev, struct timed_event *te,
        struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
        int ret;
 
-       if (ctdb->monitoring_mode == CTDB_MONITORING_DISABLED) {
+       if (ctdb->monitoring_mode == CTDB_MONITORING_DISABLED && ctdb->done_startup) {
                event_add_timed(ctdb->ev, ctdb->monitor_context,
                                timeval_current_ofs(ctdb->tunable.monitor_interval, 0), 
                                ctdb_check_health, ctdb);
                return;
        }
        
-       ret = ctdb_event_script_callback(ctdb, 
-                                        timeval_current_ofs(ctdb->tunable.script_timeout, 0),
-                                        ctdb->monitor_context, ctdb_health_callback, ctdb, "monitor");
+       if (!ctdb->done_startup) {
+               ret = ctdb_event_script_callback(ctdb, 
+                                                timeval_current_ofs(ctdb->tunable.script_timeout, 0),
+                                                ctdb->monitor_context, ctdb_startup_callback, 
+                                                ctdb, "startup");
+       } else {
+               ret = ctdb_event_script_callback(ctdb, 
+                                                timeval_current_ofs(ctdb->tunable.script_timeout, 0),
+                                                ctdb->monitor_context, ctdb_health_callback, 
+                                                ctdb, "monitor");
+       }
+
        if (ret != 0) {
                DEBUG(0,("Unable to launch monitor event script\n"));
                event_add_timed(ctdb->ev, ctdb->monitor_context,