Be clear about whether a recovery pause has taken effect.

Previously, the code and documentation seem to have essentially
assumed than a call to pg_wal_replay_pause() would take place
immediately, but that's not the case, because we only check for a
pause in certain places. This means that a tool that uses this
function and then wants to do something else afterward that is
dependent on the pause having taken effect doesn't know how long it
needs to wait to be sure that no more WAL is going to be replayed.

To avoid that, add a new function pg_get_wal_replay_pause_state()
which returns either 'not paused', 'paused requested', or 'paused'.
After calling pg_wal_replay_pause() the status will immediate change
from 'not paused' to 'pause requested'; when the startup process
has noticed this, the status will change to 'pause'.  For backward
compatibility, pg_is_wal_replay_paused() still exists and returns
the same thing as before: true if a pause has been requested,
whether or not it has taken effect yet; and false if not.
The documentation is updated to clarify.

To improve the changes that a pause request is quickly confirmed
effective, adjust things so that WaitForWALToBecomeAvailable will
swiftly reach a call to recoveryPausesHere() when a pause request
is made.

Dilip Kumar, reviewed by Simon Riggs, Kyotaro Horiguchi, Yugo Nagata,
Masahiko Sawada, and Bharath Rupireddy.

Discussion: http://postgr.es/m/CAFiTN-vcLLWEm8Zr%3DYK83rgYrT9pbC8VJCfa1kY9vL3AUPfu6g%40mail.gmail.com

1 files changed, 70 insertions, 14 deletions

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 18af3d41209..e04250f4e9e 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -722,8 +722,8 @@ typedef struct XLogCtlData
 	 * only relevant for replication or archive recovery
 	 */
 	TimestampTz currentChunkStartTime;
-	/* Are we requested to pause recovery? */
-	bool		recoveryPause;
+	/* Recovery pause state */
+	RecoveryPauseState	recoveryPauseState;
 
 	/*
 	 * lastFpwDisableRecPtr points to the start of the last replayed
@@ -895,6 +895,7 @@ static void validateRecoveryParameters(void);
 static void exitArchiveRecovery(TimeLineID endTLI, XLogRecPtr endOfLog);
 static bool recoveryStopsBefore(XLogReaderState *record);
 static bool recoveryStopsAfter(XLogReaderState *record);
+static void ConfirmRecoveryPaused(void);
 static void recoveryPausesHere(bool endOfRecovery);
 static bool recoveryApplyDelay(XLogReaderState *record);
 static void SetLatestXTime(TimestampTz xtime);
@@ -6034,7 +6035,7 @@ recoveryStopsAfter(XLogReaderState *record)
 }
 
 /*
- * Wait until shared recoveryPause flag is cleared.
+ * Wait until shared recoveryPauseState is set to RECOVERY_NOT_PAUSED.
  *
  * endOfRecovery is true if the recovery target is reached and
  * the paused state starts at the end of recovery because of
@@ -6064,34 +6065,72 @@ recoveryPausesHere(bool endOfRecovery)
 				(errmsg("recovery has paused"),
 				 errhint("Execute pg_wal_replay_resume() to continue.")));
 
-	while (RecoveryIsPaused())
+	/* loop until recoveryPauseState is set to RECOVERY_NOT_PAUSED */
+	while (GetRecoveryPauseState() != RECOVERY_NOT_PAUSED)
 	{
 		HandleStartupProcInterrupts();
 		if (CheckForStandbyTrigger())
 			return;
 		pgstat_report_wait_start(WAIT_EVENT_RECOVERY_PAUSE);
+
+		/*
+		 * If recovery pause is requested then set it paused.  While we are in
+		 * the loop, user might resume and pause again so set this every time.
+		 */
+		ConfirmRecoveryPaused();
+
 		pg_usleep(1000000L);	/* 1000 ms */
 		pgstat_report_wait_end();
 	}
 }
 
-bool
-RecoveryIsPaused(void)
+/*
+ * Get the current state of the recovery pause request.
+ */
+RecoveryPauseState
+GetRecoveryPauseState(void)
 {
-	bool		recoveryPause;
+	RecoveryPauseState	state;
 
 	SpinLockAcquire(&XLogCtl->info_lck);
-	recoveryPause = XLogCtl->recoveryPause;
+	state = XLogCtl->recoveryPauseState;
 	SpinLockRelease(&XLogCtl->info_lck);
 
-	return recoveryPause;
+	return state;
 }
 
+/*
+ * Set the recovery pause state.
+ *
+ * If recovery pause is requested then sets the recovery pause state to
+ * 'pause requested' if it is not already 'paused'.  Otherwise, sets it
+ * to 'not paused' to resume the recovery.  The recovery pause will be
+ * confirmed by the ConfirmRecoveryPaused.
+ */
 void
 SetRecoveryPause(bool recoveryPause)
 {
 	SpinLockAcquire(&XLogCtl->info_lck);
-	XLogCtl->recoveryPause = recoveryPause;
+
+	if (!recoveryPause)
+		XLogCtl->recoveryPauseState = RECOVERY_NOT_PAUSED;
+	else if (XLogCtl->recoveryPauseState == RECOVERY_NOT_PAUSED)
+		XLogCtl->recoveryPauseState = RECOVERY_PAUSE_REQUESTED;
+
+	SpinLockRelease(&XLogCtl->info_lck);
+}
+
+/*
+ * Confirm the recovery pause by setting the recovery pause state to
+ * RECOVERY_PAUSED.
+ */
+static void
+ConfirmRecoveryPaused(void)
+{
+	/* If recovery pause is requested then set it paused */
+	SpinLockAcquire(&XLogCtl->info_lck);
+	if (XLogCtl->recoveryPauseState == RECOVERY_PAUSE_REQUESTED)
+		XLogCtl->recoveryPauseState = RECOVERY_PAUSED;
 	SpinLockRelease(&XLogCtl->info_lck);
 }
 
@@ -6292,7 +6331,7 @@ RecoveryRequiresIntParameter(const char *param_name, int currValue, int minValue
 					 errdetail("If recovery is unpaused, the server will shut down."),
 					 errhint("You can then restart the server after making the necessary configuration changes.")));
 
-			while (RecoveryIsPaused())
+			while (GetRecoveryPauseState() != RECOVERY_NOT_PAUSED)
 			{
 				HandleStartupProcInterrupts();
 
@@ -6311,6 +6350,13 @@ RecoveryRequiresIntParameter(const char *param_name, int currValue, int minValue
 					warned_for_promote = true;
 				}
 
+				/*
+				 * If recovery pause is requested then set it paused.  While we
+				 * are in the loop, user might resume and pause again so set
+				 * this every time.
+				 */
+				ConfirmRecoveryPaused();
+
 				pgstat_report_wait_start(WAIT_EVENT_RECOVERY_PAUSE);
 				pg_usleep(1000000L);	/* 1000 ms */
 				pgstat_report_wait_end();
@@ -7205,7 +7251,7 @@ StartupXLOG(void)
 		XLogCtl->lastReplayedTLI = XLogCtl->replayEndTLI;
 		XLogCtl->recoveryLastXTime = 0;
 		XLogCtl->currentChunkStartTime = 0;
-		XLogCtl->recoveryPause = false;
+		XLogCtl->recoveryPauseState = RECOVERY_NOT_PAUSED;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* Also ensure XLogReceiptTime has a sane value */
@@ -7309,7 +7355,8 @@ StartupXLOG(void)
 				 * otherwise would is a minor issue, so it doesn't seem worth
 				 * adding another spinlock cycle to prevent that.
 				 */
-				if (((volatile XLogCtlData *) XLogCtl)->recoveryPause)
+				if (((volatile XLogCtlData *) XLogCtl)->recoveryPauseState !=
+					RECOVERY_NOT_PAUSED)
 					recoveryPausesHere(false);
 
 				/*
@@ -7334,7 +7381,8 @@ StartupXLOG(void)
 					 * here otherwise pausing during the delay-wait wouldn't
 					 * work.
 					 */
-					if (((volatile XLogCtlData *) XLogCtl)->recoveryPause)
+					if (((volatile XLogCtlData *) XLogCtl)->recoveryPauseState !=
+						RECOVERY_NOT_PAUSED)
 						recoveryPausesHere(false);
 				}
 
@@ -12657,6 +12705,14 @@ WaitForWALToBecomeAvailable(XLogRecPtr RecPtr, bool randAccess,
 		}
 
 		/*
+		 * Check for recovery pause here so that we can confirm more quickly
+		 * that a requested pause has actually taken effect.
+		 */
+		if (((volatile XLogCtlData *) XLogCtl)->recoveryPauseState !=
+			RECOVERY_NOT_PAUSED)
+			recoveryPausesHere(false);
+
+		/*
 		 * This possibly-long loop needs to handle interrupts of startup
 		 * process.
 		 */