Index: src/backend/postmaster/pgarch.c =================================================================== RCS file: /cvsroot/pgsql/src/backend/postmaster/pgarch.c,v retrieving revision 1.37 diff -c -r1.37 pgarch.c *** src/backend/postmaster/pgarch.c 1 Jan 2008 19:45:51 -0000 1.37 --- src/backend/postmaster/pgarch.c 9 Jan 2008 23:13:33 -0000 *************** *** 267,284 **** static void pgarch_exit(SIGNAL_ARGS) { ! /* ! * For now, we just nail the doors shut and get out of town. It might ! * seem cleaner to finish up any pending archive copies, but there's a ! * nontrivial risk that init will kill us partway through. ! */ ! exit(0); } ! /* SIGHUP: set flag to re-read config file at next convenient time */ static void ArchSigHupHandler(SIGNAL_ARGS) { got_SIGHUP = true; } --- 267,281 ---- static void pgarch_exit(SIGNAL_ARGS) { ! /* SIGQUIT means curl up and die ... */ ! exit(1); } ! /* SIGHUP signal handler for archiver process */ static void ArchSigHupHandler(SIGNAL_ARGS) { + /* set flag to re-read config file at next convenient time */ got_SIGHUP = true; } *************** *** 286,291 **** --- 283,289 ---- static void pgarch_waken(SIGNAL_ARGS) { + /* set flag that there is work to be done */ wakened = true; } *************** *** 298,303 **** --- 296,302 ---- pgarch_MainLoop(void) { time_t last_copy_time = 0; + bool postmaster_alive; /* * We run the copy loop immediately upon entry, in case there are *************** *** 309,314 **** --- 308,316 ---- do { + /* When the postmaster dies, we do one more archive cycle, then exit */ + postmaster_alive = PostmasterIsAlive(true); + /* Check for config update */ if (got_SIGHUP) { *************** *** 334,340 **** * sleep into 1-second increments, and check for interrupts after each * nap. */ ! while (!(wakened || got_SIGHUP)) { time_t curtime; --- 336,342 ---- * sleep into 1-second increments, and check for interrupts after each * nap. */ ! while (!(wakened || got_SIGHUP || !PostmasterIsAlive(true))) { time_t curtime; *************** *** 344,350 **** (unsigned int) PGARCH_AUTOWAKE_INTERVAL) wakened = true; } ! } while (PostmasterIsAlive(true)); } /* --- 346,352 ---- (unsigned int) PGARCH_AUTOWAKE_INTERVAL) wakened = true; } ! } while (postmaster_alive); } /* Index: src/backend/postmaster/postmaster.c =================================================================== RCS file: /cvsroot/pgsql/src/backend/postmaster/postmaster.c,v retrieving revision 1.550 diff -c -r1.550 postmaster.c *** src/backend/postmaster/postmaster.c 1 Jan 2008 19:45:51 -0000 1.550 --- src/backend/postmaster/postmaster.c 9 Jan 2008 23:13:33 -0000 *************** *** 1315,1327 **** /* * If we have lost the archiver, try to start a new one. We do this * even if we are shutting down, to allow archiver to take care of any ! * remaining WAL files. */ ! if (XLogArchivingActive() && PgArchPID == 0 && pmState >= PM_RUN) PgArchPID = pgarch_start(); /* If we have lost the stats collector, try to start a new one */ ! if (PgStatPID == 0 && pmState == PM_RUN) PgStatPID = pgstat_start(); /* --- 1315,1330 ---- /* * If we have lost the archiver, try to start a new one. We do this * even if we are shutting down, to allow archiver to take care of any ! * remaining WAL files. Once shutdown is done, though, don't restart ! * it anymore. */ ! if (XLogArchivingActive() && PgArchPID == 0 && ! pmState >= PM_RUN && pmState <= PM_SHUTDOWN) PgArchPID = pgarch_start(); /* If we have lost the stats collector, try to start a new one */ ! if (PgStatPID == 0 && ! pmState >= PM_RUN && pmState <= PM_SHUTDOWN) PgStatPID = pgstat_start(); /* *************** *** 2177,2186 **** * * At this point we should have no normal children left (else * we'd not be in PM_SHUTDOWN state) but we might have ! * dead_end children. */ Assert(Shutdown > NoShutdown); pmState = PM_WAIT_DEAD_END; } else { --- 2180,2197 ---- * * At this point we should have no normal children left (else * we'd not be in PM_SHUTDOWN state) but we might have ! * dead_end children. We can also tell the archiver and ! * stats collector processes to shut down. */ Assert(Shutdown > NoShutdown); pmState = PM_WAIT_DEAD_END; + + /* Waken archiver for the last time */ + if (PgArchPID != 0 && Shutdown <= SmartShutdown) + signal_child(PgArchPID, SIGUSR1); + /* Tell pgstat to shut down; nothing left for it to do */ + if (PgStatPID != 0) + signal_child(PgStatPID, SIGQUIT); } else { *************** *** 2235,2241 **** if (!EXIT_STATUS_0(exitstatus)) LogChildExit(LOG, _("archiver process"), pid, exitstatus); ! if (XLogArchivingActive() && pmState >= PM_RUN) PgArchPID = pgarch_start(); continue; } --- 2246,2253 ---- if (!EXIT_STATUS_0(exitstatus)) LogChildExit(LOG, _("archiver process"), pid, exitstatus); ! if (XLogArchivingActive() && ! pmState >= PM_RUN && pmState <= PM_SHUTDOWN) PgArchPID = pgarch_start(); continue; } *************** *** 2251,2257 **** if (!EXIT_STATUS_0(exitstatus)) LogChildExit(LOG, _("statistics collector process"), pid, exitstatus); ! if (pmState == PM_RUN) PgStatPID = pgstat_start(); continue; } --- 2263,2269 ---- if (!EXIT_STATUS_0(exitstatus)) LogChildExit(LOG, _("statistics collector process"), pid, exitstatus); ! if (pmState >= PM_RUN && pmState <= PM_SHUTDOWN) PgStatPID = pgstat_start(); continue; } *************** *** 2563,2568 **** --- 2575,2585 ---- * change causes ServerLoop to stop creating new ones. */ pmState = PM_WAIT_DEAD_END; + + /* + * We already SIGQUIT'd the archiver and stats processes, + * if any, when we entered FatalError state. + */ } else { *************** *** 2591,2603 **** */ FatalError = true; pmState = PM_WAIT_DEAD_END; } - /* Tell pgarch to shut down too; nothing left for it to do */ - if (PgArchPID != 0) - signal_child(PgArchPID, SIGQUIT); - /* Tell pgstat to shut down too; nothing left for it to do */ - if (PgStatPID != 0) - signal_child(PgStatPID, SIGQUIT); } } } --- 2608,2620 ---- */ FatalError = true; pmState = PM_WAIT_DEAD_END; + + /* Kill the archiver and stats collector too */ + if (PgArchPID != 0) + signal_child(PgArchPID, SIGQUIT); + if (PgStatPID != 0) + signal_child(PgStatPID, SIGQUIT); } } } } *************** *** 2629,2641 **** * before init decides to SIGKILL us.) * * Note: we do not wait around for exit of the archiver or stats ! * processes. They've been sent SIGQUIT by this point (either when we ! * entered PM_SHUTDOWN state, or when we set FatalError, and at least one ! * of those must have happened by now). In any case they contain logic to ! * commit hara-kiri if they notice the postmaster is gone. Since they ! * aren't connected to shared memory, they pose no problem for shutdown. ! * The syslogger is not considered either, since it's intended to survive ! * till the postmaster exits. */ if (Shutdown > NoShutdown && pmState == PM_NO_CHILDREN) { --- 2646,2658 ---- * before init decides to SIGKILL us.) * * Note: we do not wait around for exit of the archiver or stats ! * processes. They've been sent appropriate signals by this point (either ! * when we entered PM_WAIT_DEAD_END state, or when we set FatalError, and ! * at least one of those must have happened by now). In any case they ! * contain logic to commit hara-kiri if they notice the postmaster is ! * gone. Since they aren't connected to shared memory, they pose no ! * problem for shutdown. The syslogger is not considered either, since ! * it's intended to survive till the postmaster exits. */ if (Shutdown > NoShutdown && pmState == PM_NO_CHILDREN) {