Restart bgworkers immediately after a crash-and-restart cycle.

Just as we would start bgworkers immediately after an initial startup
of the server, we should restart them immediately when reinitializing.

Petr Jelinek and Robert Haas

3 files changed, 33 insertions, 9 deletions

diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c
index 8078a38ed72..64c97229931 100644
--- a/src/backend/postmaster/bgworker.c
+++ b/src/backend/postmaster/bgworker.c
@@ -394,6 +394,27 @@ BackgroundWorkerStopNotifications(pid_t pid)
 	}
 }
 
+/*
+ * Reset background worker crash state.
+ *
+ * We assume that, after a crash-and-restart cycle, background workers should
+ * be restarted immediately, instead of waiting for bgw_restart_time to
+ * elapse.
+ */
+void
+ResetBackgroundWorkerCrashTimes(void)
+{
+	slist_mutable_iter	iter;
+
+	slist_foreach_modify(iter, &BackgroundWorkerList)
+	{
+		RegisteredBgWorker *rw;
+
+		rw = slist_container(RegisteredBgWorker, rw_lnode, iter.cur);
+		rw->rw_crashed_at = 0;
+	}
+}
+
 #ifdef EXEC_BACKEND
 /*
  * In EXEC_BACKEND mode, workers use this to retrieve their details from
@@ -478,13 +499,14 @@ bgworker_quickdie(SIGNAL_ARGS)
 	on_exit_reset();
 
 	/*
-	 * Note we do exit(0) here, not exit(2) like quickdie.  The reason is that
-	 * we don't want to be seen this worker as independently crashed, because
-	 * then postmaster would delay restarting it again afterwards.  If some
-	 * idiot DBA manually sends SIGQUIT to a random bgworker, the "dead man
-	 * switch" will ensure that postmaster sees this as a crash.
+	 * Note we do exit(2) not exit(0).  This is to force the postmaster into a
+	 * system reset cycle if some idiot DBA sends a manual SIGQUIT to a random
+	 * backend.  This is necessary precisely because we don't clean up our
+	 * shared memory state.  (The "dead man switch" mechanism in pmsignal.c
+	 * should ensure the postmaster sees this as a crash, too, but no harm in
+	 * being doubly sure.)
 	 */
-	exit(0);
+	exit(2);
 }
 
 /*
diff --git a/src/backend/postmaster/postmaster.c b/src/backend/postmaster/postmaster.c
index 0c6a4271a60..ec1a59d061a 100644
--- a/src/backend/postmaster/postmaster.c
+++ b/src/backend/postmaster/postmaster.c
@@ -2616,7 +2616,7 @@ reaper(SIGNAL_ARGS)
 			if (PgStatPID == 0)
 				PgStatPID = pgstat_start();
 
-			/* some workers may be scheduled to start now */
+			/* workers may be scheduled to start now */
 			maybe_start_bgworker();
 
 			/* at this point we are really open for business */
@@ -2860,7 +2860,6 @@ CleanupBackgroundWorker(int pid,
 		{
 			if (!EXIT_STATUS_0(exitstatus) && !EXIT_STATUS_1(exitstatus))
 			{
-				rw->rw_crashed_at = GetCurrentTimestamp();
 				HandleChildCrash(pid, exitstatus, namebuf);
 				return true;
 			}
@@ -2871,7 +2870,6 @@ CleanupBackgroundWorker(int pid,
 				 * Uh-oh, the child failed to clean itself up.  Treat as a
 				 * crash after all.
 				 */
-				rw->rw_crashed_at = GetCurrentTimestamp();
 				HandleChildCrash(pid, exitstatus, namebuf);
 				return true;
 			}
@@ -3546,6 +3544,9 @@ PostmasterStateMachine(void)
 		ereport(LOG,
 				(errmsg("all server processes terminated; reinitializing")));
 
+		/* allow background workers to immediately restart */
+		ResetBackgroundWorkerCrashTimes();
+
 		shmem_exit(1);
 		reset_shared(PostPortNumber);
 
diff --git a/src/include/postmaster/bgworker_internals.h b/src/include/postmaster/bgworker_internals.h
index 55401860d89..c2518a6c8d7 100644
--- a/src/include/postmaster/bgworker_internals.h
+++ b/src/include/postmaster/bgworker_internals.h
@@ -43,6 +43,7 @@ extern void BackgroundWorkerStateChange(void);
 extern void ForgetBackgroundWorker(slist_mutable_iter *cur);
 extern void ReportBackgroundWorkerPID(RegisteredBgWorker *);
 extern void BackgroundWorkerStopNotifications(pid_t pid);
+extern void ResetBackgroundWorkerCrashTimes(void);
 
 /* Function to start a background worker, called from postmaster.c */
 extern void StartBackgroundWorker(void);