external
/
citus
mirror of https://github.com/citusdata/citus.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #3756 from citusdata/fix-maintenanced-error-restart 

maintenanced: use before_shmem_exit to clear workerPid
pull/3786/head
Philip Dubé 2020-04-20 14:57:30 +00:00 committed by GitHub
parent 1423433531 9093d51a22
commit 2e5b1bfa41
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 33 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

35
src/backend/distributed/utils/maintenanced.c
View File

@ -108,6 +108,7 @@ static void MaintenanceDaemonSigTermHandler(SIGNAL_ARGS);
static void MaintenanceDaemonSigHupHandler(SIGNAL_ARGS); static void MaintenanceDaemonSigHupHandler(SIGNAL_ARGS);
static size_t MaintenanceDaemonShmemSize(void); static size_t MaintenanceDaemonShmemSize(void);
static void MaintenanceDaemonShmemInit(void); static void MaintenanceDaemonShmemInit(void);
static void MaintenanceDaemonShmemExit(int code, Datum arg);
static void MaintenanceDaemonErrorContext(void *arg); static void MaintenanceDaemonErrorContext(void *arg);
static bool LockCitusExtension(void); static bool LockCitusExtension(void);
static bool MetadataSyncTriggeredCheckAndReset(MaintenanceDaemonDBData *dbData); static bool MetadataSyncTriggeredCheckAndReset(MaintenanceDaemonDBData *dbData);
@ -258,18 +259,22 @@ CitusMaintenanceDaemonMain(Datum main_arg)
MaintenanceDaemonDBData *myDbData = (MaintenanceDaemonDBData *) MaintenanceDaemonDBData *myDbData = (MaintenanceDaemonDBData *)
hash_search(MaintenanceDaemonDBHash, &databaseOid, hash_search(MaintenanceDaemonDBHash, &databaseOid,
HASH_FIND, NULL); HASH_FIND, NULL);
if (!myDbData || myDbData->workerPid != 0) if (!myDbData)
{ {
/* /*
* When the database crashes, background workers are restarted, but * When the database crashes, background workers are restarted, but
* the state in shared memory is lost. In that case, we exit and * the state in shared memory is lost. In that case, we exit and
* wait for a session to call InitializeMaintenanceDaemonBackend * wait for a session to call InitializeMaintenanceDaemonBackend
* to properly add it to the hash. * to properly add it to the hash.
* Alternatively, don't continue if another worker exists.
*/ */
proc_exit(0); proc_exit(0);
} }
before_shmem_exit(MaintenanceDaemonShmemExit, main_arg);
Assert(myDbData->workerPid == 0);
/* from this point, DROP DATABASE will attempt to kill the worker */ /* from this point, DROP DATABASE will attempt to kill the worker */
myDbData->workerPid = MyProcPid; myDbData->workerPid = MyProcPid;
@ -307,7 +312,6 @@ CitusMaintenanceDaemonMain(Datum main_arg)
/* enter main loop */ /* enter main loop */
while (!got_SIGTERM) while (!got_SIGTERM)
{ {
int rc;
int latchFlags = WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH; int latchFlags = WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH;
double timeout = 10000.0; /* use this if the deadlock detection is disabled */ double timeout = 10000.0; /* use this if the deadlock detection is disabled */
bool foundDeadlock = false; bool foundDeadlock = false;
@ -524,7 +528,7 @@ CitusMaintenanceDaemonMain(Datum main_arg)
* Wait until timeout, or until somebody wakes us up. Also cast the timeout to * Wait until timeout, or until somebody wakes us up. Also cast the timeout to
* integer where we've calculated it using double for not losing the precision. * integer where we've calculated it using double for not losing the precision.
*/ */
rc = WaitLatch(MyLatch, latchFlags, (long) timeout, PG_WAIT_EXTENSION); int rc = WaitLatch(MyLatch, latchFlags, (long) timeout, PG_WAIT_EXTENSION);
/* emergency bailout if postmaster has died */ /* emergency bailout if postmaster has died */
if (rc & WL_POSTMASTER_DEATH) if (rc & WL_POSTMASTER_DEATH)
@ -647,6 +651,29 @@ MaintenanceDaemonShmemInit(void)
} }
/*
* MaintenaceDaemonShmemExit is the before_shmem_exit handler for cleaning up MaintenanceDaemonDBHash
*/
static void
MaintenanceDaemonShmemExit(int code, Datum arg)
{
Oid databaseOid = DatumGetObjectId(arg);
LWLockAcquire(&MaintenanceDaemonControl->lock, LW_EXCLUSIVE);
MaintenanceDaemonDBData *myDbData = (MaintenanceDaemonDBData *)
hash_search(MaintenanceDaemonDBHash, &databaseOid,
HASH_FIND, NULL);
if (myDbData && myDbData->workerPid == MyProcPid)
{
myDbData->daemonStarted = false;
myDbData->workerPid = 0;
}
LWLockRelease(&MaintenanceDaemonControl->lock);
}
/* MaintenanceDaemonSigTermHandler calls proc_exit(0) */ /* MaintenanceDaemonSigTermHandler calls proc_exit(0) */
static void static void
MaintenanceDaemonSigTermHandler(SIGNAL_ARGS) MaintenanceDaemonSigTermHandler(SIGNAL_ARGS)