From 7e2a18a9161fee7e67642863f72b51d77d3e996f Mon Sep 17 00:00:00 2001 From: Tom Lane Date: Tue, 6 Oct 2015 17:15:27 -0400 Subject: [PATCH] Perform an immediate shutdown if the postmaster.pid file is removed. The postmaster now checks every minute or so (worst case, at most two minutes) that postmaster.pid is still there and still contains its own PID. If not, it performs an immediate shutdown, as though it had received SIGQUIT. The original goal behind this change was to ensure that failed buildfarm runs would get fully cleaned up, even if the test scripts had left a postmaster running, which is not an infrequent occurrence. When the buildfarm script removes a test postmaster's $PGDATA directory, its next check on postmaster.pid will fail and cause it to exit. Previously, manual intervention was often needed to get rid of such orphaned postmasters, since they'd block new test postmasters from obtaining the expected socket address. However, by checking postmaster.pid and not something else, we can provide additional robustness: manual removal of postmaster.pid is a frequent DBA mistake, and now we can at least limit the damage that will ensue if a new postmaster is started while the old one is still alive. Back-patch to all supported branches, since we won't get the desired improvement in buildfarm reliability otherwise. --- src/backend/postmaster/postmaster.c | 58 ++++++++++++++++++------ src/backend/utils/init/miscinit.c | 70 +++++++++++++++++++++++++++++ src/include/miscadmin.h | 1 + 3 files changed, 115 insertions(+), 14 deletions(-) diff --git a/src/backend/postmaster/postmaster.c b/src/backend/postmaster/postmaster.c index baa43b203f..aab2f4ca70 100644 --- a/src/backend/postmaster/postmaster.c +++ b/src/backend/postmaster/postmaster.c @@ -1602,9 +1602,10 @@ ServerLoop(void) fd_set readmask; int nSockets; time_t now, + last_lockfile_recheck_time, last_touch_time; - last_touch_time = time(NULL); + last_lockfile_recheck_time = last_touch_time = time(NULL); nSockets = initMasks(&readmask); @@ -1754,19 +1755,6 @@ ServerLoop(void) if (StartWorkerNeeded || HaveCrashedWorker) maybe_start_bgworker(); - /* - * Touch Unix socket and lock files every 58 minutes, to ensure that - * they are not removed by overzealous /tmp-cleaning tasks. We assume - * no one runs cleaners with cutoff times of less than an hour ... - */ - now = time(NULL); - if (now - last_touch_time >= 58 * SECS_PER_MINUTE) - { - TouchSocketFiles(); - TouchSocketLockFiles(); - last_touch_time = now; - } - #ifdef HAVE_PTHREAD_IS_THREADED_NP /* @@ -1793,6 +1781,48 @@ ServerLoop(void) /* reset flag so we don't SIGKILL again */ AbortStartTime = 0; } + + /* + * Lastly, check to see if it's time to do some things that we don't + * want to do every single time through the loop, because they're a + * bit expensive. Note that there's up to a minute of slop in when + * these tasks will be performed, since DetermineSleepTime() will let + * us sleep at most that long. + */ + now = time(NULL); + + /* + * Once a minute, verify that postmaster.pid hasn't been removed or + * overwritten. If it has, we force a shutdown. This avoids having + * postmasters and child processes hanging around after their database + * is gone, and maybe causing problems if a new database cluster is + * created in the same place. It also provides some protection + * against a DBA foolishly removing postmaster.pid and manually + * starting a new postmaster. Data corruption is likely to ensue from + * that anyway, but we can minimize the damage by aborting ASAP. + */ + if (now - last_lockfile_recheck_time >= 1 * SECS_PER_MINUTE) + { + if (!RecheckDataDirLockFile()) + { + ereport(LOG, + (errmsg("performing immediate shutdown because data directory lock file is invalid"))); + kill(MyProcPid, SIGQUIT); + } + last_lockfile_recheck_time = now; + } + + /* + * Touch Unix socket and lock files every 58 minutes, to ensure that + * they are not removed by overzealous /tmp-cleaning tasks. We assume + * no one runs cleaners with cutoff times of less than an hour ... + */ + if (now - last_touch_time >= 58 * SECS_PER_MINUTE) + { + TouchSocketFiles(); + TouchSocketLockFiles(); + last_touch_time = now; + } } } diff --git a/src/backend/utils/init/miscinit.c b/src/backend/utils/init/miscinit.c index e871fef7fa..fb3cb6eb3d 100644 --- a/src/backend/utils/init/miscinit.c +++ b/src/backend/utils/init/miscinit.c @@ -1218,6 +1218,76 @@ AddToDataDirLockFile(int target_line, const char *str) } +/* + * Recheck that the data directory lock file still exists with expected + * content. Return TRUE if the lock file appears OK, FALSE if it isn't. + * + * We call this periodically in the postmaster. The idea is that if the + * lock file has been removed or replaced by another postmaster, we should + * do a panic database shutdown. Therefore, we should return TRUE if there + * is any doubt: we do not want to cause a panic shutdown unnecessarily. + * Transient failures like EINTR or ENFILE should not cause us to fail. + * (If there really is something wrong, we'll detect it on a future recheck.) + */ +bool +RecheckDataDirLockFile(void) +{ + int fd; + int len; + long file_pid; + char buffer[BLCKSZ]; + + fd = open(DIRECTORY_LOCK_FILE, O_RDWR | PG_BINARY, 0); + if (fd < 0) + { + /* + * There are many foreseeable false-positive error conditions. For + * safety, fail only on enumerated clearly-something-is-wrong + * conditions. + */ + switch (errno) + { + case ENOENT: + case ENOTDIR: + /* disaster */ + ereport(LOG, + (errcode_for_file_access(), + errmsg("could not open file \"%s\": %m", + DIRECTORY_LOCK_FILE))); + return false; + default: + /* non-fatal, at least for now */ + ereport(LOG, + (errcode_for_file_access(), + errmsg("could not open file \"%s\": %m; continuing anyway", + DIRECTORY_LOCK_FILE))); + return true; + } + } + len = read(fd, buffer, sizeof(buffer) - 1); + if (len < 0) + { + ereport(LOG, + (errcode_for_file_access(), + errmsg("could not read from file \"%s\": %m", + DIRECTORY_LOCK_FILE))); + close(fd); + return true; /* treat read failure as nonfatal */ + } + buffer[len] = '\0'; + close(fd); + file_pid = atol(buffer); + if (file_pid == getpid()) + return true; /* all is well */ + + /* Trouble: someone's overwritten the lock file */ + ereport(LOG, + (errmsg("lock file \"%s\" contains wrong PID: %ld instead of %ld", + DIRECTORY_LOCK_FILE, file_pid, (long) getpid()))); + return false; +} + + /*------------------------------------------------------------------------- * Version checking support *------------------------------------------------------------------------- diff --git a/src/include/miscadmin.h b/src/include/miscadmin.h index 77cf8d71de..e0d464e02f 100644 --- a/src/include/miscadmin.h +++ b/src/include/miscadmin.h @@ -453,6 +453,7 @@ extern void CreateSocketLockFile(const char *socketfile, bool amPostmaster, const char *socketDir); extern void TouchSocketLockFiles(void); extern void AddToDataDirLockFile(int target_line, const char *str); +extern bool RecheckDataDirLockFile(void); extern void ValidatePgVersion(const char *path); extern void process_shared_preload_libraries(void); extern void process_session_preload_libraries(void); -- 2.40.0