In walsender, don't sleep if there's outstanding WAL waiting to be sent,

otherwise we effectively rate-limit the streaming as pointed out by
Simon Riggs. Also, send the WAL in smaller chunks, to respond to signals
more promptly.

src/backend/replication/walsender.c

@@ -30,7 +30,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/replication/walsender.c,v 1.20 2010/05/09 18:11:55 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/replication/walsender.c,v 1.21 2010/05/26 22:21:33 heikki Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -100,13 +100,19 @@ static void InitWalSnd(void);
 static void WalSndHandshake(void);
 static void WalSndKill(int code, Datum arg);
 static void XLogRead(char *buf, XLogRecPtr recptr, Size nbytes);
-static bool XLogSend(StringInfo outMsg);
+static bool XLogSend(StringInfo outMsg, bool *caughtup);
 static void CheckClosedConnection(void);
 
 /*
  * How much WAL to send in one message? Must be >= XLOG_BLCKSZ.
+ *
+ * We don't have a good idea of what a good value would be; there's some
+ * overhead per message in both walsender and walreceiver, but on the other
+ * hand sending large batches makes walsender less responsive to signals
+ * because signals are checked only between messages. 128kB seems like
+ * a reasonable guess for now.
  */
-#define MAX_SEND_SIZE (XLOG_SEG_SIZE / 2)
+#define MAX_SEND_SIZE (128 * 1024)
 
 /* Main entry point for walsender process */
 int
@@ -360,6 +366,7 @@ static int
 WalSndLoop(void)
 {
 	StringInfoData output_message;
+	bool		caughtup = false;
 
 	initStringInfo(&output_message);
 
@@ -387,7 +394,7 @@ WalSndLoop(void)
 		 */
 		if (ready_to_stop)
 		{
-			XLogSend(&output_message);
+			XLogSend(&output_message, &caughtup);
 			shutdown_requested = true;
 		}
 
@@ -402,31 +409,32 @@ WalSndLoop(void)
 		}
 
 		/*
-		 * Nap for the configured time or until a message arrives.
+		 * If we had sent all accumulated WAL in last round, nap for the
+		 * configured time before retrying.
 		 *
 		 * On some platforms, signals won't interrupt the sleep.  To ensure we
 		 * respond reasonably promptly when someone signals us, break down the
 		 * sleep into NAPTIME_PER_CYCLE increments, and check for
 		 * interrupts after each nap.
 		 */
+		if (caughtup)
+		{
 			remain = WalSndDelay * 1000L;
 			while (remain > 0)
 			{
+				/* Check for interrupts */
 				if (got_SIGHUP || shutdown_requested || ready_to_stop)
 					break;
 
-			/*
-			 * Check to see whether a message from the standby or an interrupt
-			 * from other processes has arrived.
-			 */
+				/* Sleep and check that the connection is still alive */
 				pg_usleep(remain > NAPTIME_PER_CYCLE ? NAPTIME_PER_CYCLE : remain);
 				CheckClosedConnection();
 
 				remain -= NAPTIME_PER_CYCLE;
 			}
-
+		}
 		/* Attempt to send the log once every loop */
-		if (!XLogSend(&output_message))
+		if (!XLogSend(&output_message, &caughtup))
 			goto eof;
 	}
 
@@ -623,15 +631,20 @@ XLogRead(char *buf, XLogRecPtr recptr, Size nbytes)
 }
 
 /*
- * Read all WAL that's been written (and flushed) since last cycle, and send
- * it to client.
+ * Read up to MAX_SEND_SIZE bytes of WAL that's been written (and flushed),
+ * but not yet sent to the client, and send it. If there is no unsent WAL,
+ * *caughtup is set to true and nothing is sent, otherwise *caughtup is set
+ * to false.
  *
  * Returns true if OK, false if trouble.
  */
 static bool
-XLogSend(StringInfo outMsg)
+XLogSend(StringInfo outMsg, bool *caughtup)
 {
 	XLogRecPtr	SendRqstPtr;
+	XLogRecPtr	startptr;
+	XLogRecPtr	endptr;
+	Size		nbytes;
 	char		activitymsg[50];
 
 	/* use volatile pointer to prevent code rearrangement */
@@ -642,18 +655,17 @@ XLogSend(StringInfo outMsg)
 
 	/* Quick exit if nothing to do */
 	if (!XLByteLT(sentPtr, SendRqstPtr))
-		return true;
-
-	/*
-	 * We gather multiple records together by issuing just one XLogRead() of a
-	 * suitable size, and send them as one CopyData message. Repeat until
-	 * we've sent everything we can.
-	 */
-	while (XLByteLT(sentPtr, SendRqstPtr))
 	{
-		XLogRecPtr	startptr;
-		XLogRecPtr	endptr;
-		Size		nbytes;
+		*caughtup = true;
+		return true;
+	}
+	/*
+	 * Otherwise let the caller know that we're not fully caught up. Unless
+	 * there's a huge backlog, we'll be caught up to the current WriteRecPtr
+	 * after we've sent everything below, but more WAL could accumulate while
+	 * we're busy sending.
+	 */
+	*caughtup = false;
 
 	/*
 	 * Figure out how much to send in one message. If there's less than
@@ -719,7 +731,6 @@ XLogSend(StringInfo outMsg)
 
 	pq_putmessage('d', outMsg->data, outMsg->len);
 	resetStringInfo(outMsg);
-	}
 
 	/* Update shared memory status */
 	SpinLockAcquire(&walsnd->mutex);