/*
* Report the current WAL insert location (same format as pg_start_backup etc)
*
* This function is mostly for debugging purposes.
*/
Datum
pg_current_xlog_insert_location(PG_FUNCTION_ARGS)
{
XLogRecPtr current_recptr;
if (RecoveryInProgress())
ereport(ERROR,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("recovery is in progress"),
errhint("WAL control functions cannot be executed during recovery.")));
current_recptr = GetXLogInsertRecPtr();
PG_RETURN_LSN(current_recptr);
}
/*
* Report the current WAL insert location (same format as pg_start_backup etc)
*
* This function is mostly for debugging purposes.
*/
Datum
pg_current_xlog_insert_location(PG_FUNCTION_ARGS)
{
XLogRecPtr current_recptr;
char location[MAXFNAMELEN];
if (RecoveryInProgress())
ereport(ERROR,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("recovery is in progress"),
errhint("WAL control functions cannot be executed during recovery.")));
current_recptr = GetXLogInsertRecPtr(true);
snprintf(location, sizeof(location), "%X/%X",
current_recptr.xlogid, current_recptr.xrecoff);
PG_RETURN_TEXT_P(cstring_to_text(location));
}
/*
* Reserve WAL for the currently active slot.
*
* Compute and set restart_lsn in a manner that's appropriate for the type of
* the slot and concurrency safe.
*/
void
ReplicationSlotReserveWal(void)
{
ReplicationSlot *slot = MyReplicationSlot;
Assert(slot != NULL);
Assert(slot->data.restart_lsn == InvalidXLogRecPtr);
/*
* The replication slot mechanism is used to prevent removal of required
* WAL. As there is no interlock between this routine and checkpoints, WAL
* segments could concurrently be removed when a now stale return value of
* ReplicationSlotsComputeRequiredLSN() is used. In the unlikely case that
* this happens we'll just retry.
*/
while (true)
{
XLogSegNo segno;
/*
* For logical slots log a standby snapshot and start logical decoding
* at exactly that position. That allows the slot to start up more
* quickly.
*
* That's not needed (or indeed helpful) for physical slots as they'll
* start replay at the last logged checkpoint anyway. Instead return
* the location of the last redo LSN. While that slightly increases
* the chance that we have to retry, it's where a base backup has to
* start replay at.
*/
if (!RecoveryInProgress() && SlotIsLogical(slot))
{
XLogRecPtr flushptr;
/* start at current insert position */
slot->data.restart_lsn = GetXLogInsertRecPtr();
/* make sure we have enough information to start */
flushptr = LogStandbySnapshot();
/* and make sure it's fsynced to disk */
XLogFlush(flushptr);
}
else
{
slot->data.restart_lsn = GetRedoRecPtr();
}
/* prevent WAL removal as fast as possible */
ReplicationSlotsComputeRequiredLSN();
/*
* If all required WAL is still there, great, otherwise retry. The
* slot should prevent further removal of WAL, unless there's a
* concurrent ReplicationSlotsComputeRequiredLSN() after we've written
* the new restart_lsn above, so normally we should never need to loop
* more than twice.
*/
XLByteToSeg(slot->data.restart_lsn, segno, wal_segment_size);
if (XLogGetLastRemovedSegno() < segno)
break;
}
}
/*
* Write COMMIT to the output stream.
*/
static void
pglogical_write_commit(StringInfo out, PGLogicalOutputData *data,
ReorderBufferTXN *txn, XLogRecPtr commit_lsn)
{
uint8 flags = 0;
if (txn->xact_action == XLOG_XACT_COMMIT)
flags = PGLOGICAL_COMMIT;
else if (txn->xact_action == XLOG_XACT_PREPARE)
flags = PGLOGICAL_PREPARE;
else if (txn->xact_action == XLOG_XACT_COMMIT_PREPARED)
flags = PGLOGICAL_COMMIT_PREPARED;
else if (txn->xact_action == XLOG_XACT_ABORT_PREPARED)
flags = PGLOGICAL_ABORT_PREPARED;
else
Assert(false);
if (flags == PGLOGICAL_COMMIT || flags == PGLOGICAL_PREPARE) {
if (MtmIsFilteredTxn) {
return;
}
} else {
csn_t csn = MtmTransactionSnapshot(txn->xid);
bool isRecovery = MtmIsRecoveredNode(MtmReplicationNodeId);
/*
* INVALID_CSN means replicated transaction (transaction initiated by some other nodes).
* We do not need to send such transactions unless we perform recovery
*/
if (csn == INVALID_CSN && !isRecovery) {
return;
}
if (MtmRecoveryCaughtUp(MtmReplicationNodeId, txn->end_lsn)) {
MTM_LOG1("wal-sender complete recovery of node %d at LSN(commit %lx, end %lx, log %lx) in transaction %s event %d", MtmReplicationNodeId, commit_lsn, txn->end_lsn, GetXLogInsertRecPtr(), txn->gid, flags);
flags |= PGLOGICAL_CAUGHT_UP;
}
}
pq_sendbyte(out, 'C'); /* sending COMMIT */
MTM_LOG2("PGLOGICAL_SEND commit: event=%d, gid=%s, commit_lsn=%lx, txn->end_lsn=%lx, xlog=%lx", flags, txn->gid, commit_lsn, txn->end_lsn, GetXLogInsertRecPtr());
/* send the flags field */
pq_sendbyte(out, flags);
pq_sendbyte(out, MtmNodeId);
/* send fixed fields */
pq_sendint64(out, commit_lsn);
pq_sendint64(out, txn->end_lsn);
pq_sendint64(out, txn->commit_time);
if (txn->xact_action == XLOG_XACT_COMMIT_PREPARED) {
pq_sendint64(out, MtmGetTransactionCSN(txn->xid));
}
if (txn->xact_action != XLOG_XACT_COMMIT) {
pq_sendstring(out, txn->gid);
}
}
/*
* Main entry point for bgwriter process
*
* This is invoked from AuxiliaryProcessMain, which has already created the
* basic execution environment, but not enabled signals yet.
*/
void
BackgroundWriterMain(void)
{
sigjmp_buf local_sigjmp_buf;
MemoryContext bgwriter_context;
bool prev_hibernate;
/*
* Properly accept or ignore signals the postmaster might send us.
*
* bgwriter doesn't participate in ProcSignal signalling, but a SIGUSR1
* handler is still needed for latch wakeups.
*/
pqsignal(SIGHUP, BgSigHupHandler); /* set flag to read config file */
pqsignal(SIGINT, SIG_IGN);
pqsignal(SIGTERM, ReqShutdownHandler); /* shutdown */
pqsignal(SIGQUIT, bg_quickdie); /* hard crash time */
pqsignal(SIGALRM, SIG_IGN);
pqsignal(SIGPIPE, SIG_IGN);
pqsignal(SIGUSR1, bgwriter_sigusr1_handler);
pqsignal(SIGUSR2, SIG_IGN);
/*
* Reset some signals that are accepted by postmaster but not here
*/
pqsignal(SIGCHLD, SIG_DFL);
pqsignal(SIGTTIN, SIG_DFL);
pqsignal(SIGTTOU, SIG_DFL);
pqsignal(SIGCONT, SIG_DFL);
pqsignal(SIGWINCH, SIG_DFL);
/* We allow SIGQUIT (quickdie) at all times */
sigdelset(&BlockSig, SIGQUIT);
/*
* Create a resource owner to keep track of our resources (currently only
* buffer pins).
*/
CurrentResourceOwner = ResourceOwnerCreate(NULL, "Background Writer");
/*
* We just started, assume there has been either a shutdown or
* end-of-recovery snapshot.
*/
last_snapshot_ts = GetCurrentTimestamp();
/*
* Create a memory context that we will do all our work in. We do this so
* that we can reset the context during error recovery and thereby avoid
* possible memory leaks. Formerly this code just ran in
* TopMemoryContext, but resetting that would be a really bad idea.
*/
bgwriter_context = AllocSetContextCreate(TopMemoryContext,
"Background Writer",
ALLOCSET_DEFAULT_MINSIZE,
ALLOCSET_DEFAULT_INITSIZE,
ALLOCSET_DEFAULT_MAXSIZE);
MemoryContextSwitchTo(bgwriter_context);
/*
* If an exception is encountered, processing resumes here.
*
* See notes in postgres.c about the design of this coding.
*/
if (sigsetjmp(local_sigjmp_buf, 1) != 0)
{
/* Since not using PG_TRY, must reset error stack by hand */
error_context_stack = NULL;
/* Prevent interrupts while cleaning up */
HOLD_INTERRUPTS();
/* Report the error to the server log */
EmitErrorReport();
/*
* These operations are really just a minimal subset of
* AbortTransaction(). We don't have very many resources to worry
* about in bgwriter, but we do have LWLocks, buffers, and temp files.
*/
LWLockReleaseAll();
AbortBufferIO();
UnlockBuffers();
/* buffer pins are released here: */
ResourceOwnerRelease(CurrentResourceOwner,
RESOURCE_RELEASE_BEFORE_LOCKS,
false, true);
/* we needn't bother with the other ResourceOwnerRelease phases */
AtEOXact_Buffers(false);
AtEOXact_SMgr();
AtEOXact_Files();
AtEOXact_HashTables(false);
/*
* Now return to normal top-level context and clear ErrorContext for
* next time.
*/
MemoryContextSwitchTo(bgwriter_context);
FlushErrorState();
/* Flush any leaked data in the top-level context */
MemoryContextResetAndDeleteChildren(bgwriter_context);
/* Now we can allow interrupts again */
RESUME_INTERRUPTS();
/*
* Sleep at least 1 second after any error. A write error is likely
* to be repeated, and we don't want to be filling the error logs as
* fast as we can.
*/
pg_usleep(1000000L);
/*
* Close all open files after any error. This is helpful on Windows,
* where holding deleted files open causes various strange errors.
* It's not clear we need it elsewhere, but shouldn't hurt.
*/
smgrcloseall();
/* Report wait end here, when there is no further possibility of wait */
pgstat_report_wait_end();
}
/* We can now handle ereport(ERROR) */
PG_exception_stack = &local_sigjmp_buf;
/*
* Unblock signals (they were blocked when the postmaster forked us)
*/
PG_SETMASK(&UnBlockSig);
/*
* Reset hibernation state after any error.
*/
prev_hibernate = false;
/*
* Loop forever
*/
for (;;)
{
bool can_hibernate;
int rc;
/* Clear any already-pending wakeups */
ResetLatch(MyLatch);
if (got_SIGHUP)
{
got_SIGHUP = false;
ProcessConfigFile(PGC_SIGHUP);
}
if (shutdown_requested)
{
/*
* From here on, elog(ERROR) should end with exit(1), not send
* control back to the sigsetjmp block above
*/
ExitOnAnyError = true;
/* Normal exit from the bgwriter is here */
proc_exit(0); /* done */
}
/*
* Do one cycle of dirty-buffer writing.
*/
can_hibernate = BgBufferSync();
/*
* Send off activity statistics to the stats collector
*/
pgstat_send_bgwriter();
if (FirstCallSinceLastCheckpoint())
{
/*
* After any checkpoint, close all smgr files. This is so we
* won't hang onto smgr references to deleted files indefinitely.
*/
smgrcloseall();
}
/*
* Log a new xl_running_xacts every now and then so replication can
* get into a consistent state faster (think of suboverflowed
* snapshots) and clean up resources (locks, KnownXids*) more
* frequently. The costs of this are relatively low, so doing it 4
* times (LOG_SNAPSHOT_INTERVAL_MS) a minute seems fine.
*
* We assume the interval for writing xl_running_xacts is
* significantly bigger than BgWriterDelay, so we don't complicate the
* overall timeout handling but just assume we're going to get called
* often enough even if hibernation mode is active. It's not that
* important that log_snap_interval_ms is met strictly. To make sure
* we're not waking the disk up unnecessarily on an idle system we
* check whether there has been any WAL inserted since the last time
* we've logged a running xacts.
*
* We do this logging in the bgwriter as its the only process that is
* run regularly and returns to its mainloop all the time. E.g.
* Checkpointer, when active, is barely ever in its mainloop and thus
* makes it hard to log regularly.
*/
if (XLogStandbyInfoActive() && !RecoveryInProgress())
{
TimestampTz timeout = 0;
TimestampTz now = GetCurrentTimestamp();
timeout = TimestampTzPlusMilliseconds(last_snapshot_ts,
LOG_SNAPSHOT_INTERVAL_MS);
/*
* only log if enough time has passed and some xlog record has
* been inserted.
*/
if (now >= timeout &&
last_snapshot_lsn != GetXLogInsertRecPtr())
{
last_snapshot_lsn = LogStandbySnapshot();
last_snapshot_ts = now;
}
}
/*
* Sleep until we are signaled or BgWriterDelay has elapsed.
*
* Note: the feedback control loop in BgBufferSync() expects that we
* will call it every BgWriterDelay msec. While it's not critical for
* correctness that that be exact, the feedback loop might misbehave
* if we stray too far from that. Hence, avoid loading this process
* down with latch events that are likely to happen frequently during
* normal operation.
*/
rc = WaitLatch(MyLatch,
WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH,
BgWriterDelay /* ms */ );
/*
* If no latch event and BgBufferSync says nothing's happening, extend
* the sleep in "hibernation" mode, where we sleep for much longer
* than bgwriter_delay says. Fewer wakeups save electricity. When a
* backend starts using buffers again, it will wake us up by setting
* our latch. Because the extra sleep will persist only as long as no
* buffer allocations happen, this should not distort the behavior of
* BgBufferSync's control loop too badly; essentially, it will think
* that the system-wide idle interval didn't exist.
*
* There is a race condition here, in that a backend might allocate a
* buffer between the time BgBufferSync saw the alloc count as zero
* and the time we call StrategyNotifyBgWriter. While it's not
* critical that we not hibernate anyway, we try to reduce the odds of
* that by only hibernating when BgBufferSync says nothing's happening
* for two consecutive cycles. Also, we mitigate any possible
* consequences of a missed wakeup by not hibernating forever.
*/
if (rc == WL_TIMEOUT && can_hibernate && prev_hibernate)
{
/* Ask for notification at next buffer allocation */
StrategyNotifyBgWriter(MyProc->pgprocno);
/* Sleep ... */
rc = WaitLatch(MyLatch,
WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH,
BgWriterDelay * HIBERNATE_FACTOR);
/* Reset the notification request in case we timed out */
StrategyNotifyBgWriter(-1);
}
/*
* Emergency bailout if postmaster has died. This is to avoid the
* necessity for manual cleanup of all postmaster children.
*/
if (rc & WL_POSTMASTER_DEATH)
exit(1);
prev_hibernate = can_hibernate;
}
}
