From cafcb6b2258a10746409135037baa9ac4ab6089e Mon Sep 17 00:00:00 2001
From: Aaron Schulz <aschulz@wikimedia.org>
Date: Thu, 31 Jul 2014 23:11:34 -0700
Subject: [PATCH] Vastly increase how long wfWaitForSlaves() can block in CLI
 mode

* All maintenance scripts pretty much assumed this behavior, but it
  was not actually the case. This can avoid some massive lag problems.

Change-Id: I5894409d6abade16afd8144f6689c2aff3fa7327
---
 includes/GlobalFunctions.php |  9 ++++++---
 includes/db/Database.php     |  2 +-
 includes/db/LoadBalancer.php | 20 ++++++++++++++------
 3 files changed, 21 insertions(+), 10 deletions(-)

diff --git a/includes/GlobalFunctions.php b/includes/GlobalFunctions.php
index 7d54dbcfa0..4c2b7725c6 100644
--- a/includes/GlobalFunctions.php
+++ b/includes/GlobalFunctions.php
@@ -3752,6 +3752,7 @@ function wfGetNull() {
  * @param float|null $ifWritesSince Only wait if writes were done since this UNIX timestamp
  * @param string|bool $wiki Wiki identifier accepted by wfGetLB
  * @param string|bool $cluster Cluster name accepted by LBFactory. Default: false.
+ * @return bool Success (able to connect and no timeouts reached)
  */
 function wfWaitForSlaves( $ifWritesSince = false, $wiki = false, $cluster = false ) {
 	// B/C: first argument used to be "max seconds of lag"; ignore such values
@@ -3767,19 +3768,21 @@ function wfWaitForSlaves( $ifWritesSince = false, $wiki = false, $cluster = fals
 	// Prevents permission error when getting master position
 	if ( $lb->getServerCount() > 1 ) {
 		if ( $ifWritesSince && !$lb->hasMasterConnection() ) {
-			return; // assume no writes done
+			return true; // assume no writes done
 		}
 		$dbw = $lb->getConnection( DB_MASTER, array(), $wiki );
 		if ( $ifWritesSince && $dbw->lastDoneWrites() < $ifWritesSince ) {
-			return; // no writes since the last wait
+			return true; // no writes since the last wait
 		}
 		$pos = $dbw->getMasterPos();
 		// The DBMS may not support getMasterPos() or the whole
 		// load balancer might be fake (e.g. $wgAllDBsAreLocalhost).
 		if ( $pos !== false ) {
-			$lb->waitForAll( $pos );
+			return $lb->waitForAll( $pos, PHP_SAPI === 'cli' ? 86400 : null );
 		}
 	}
+
+	return true;
 }
 
 /**
diff --git a/includes/db/Database.php b/includes/db/Database.php
index 85f042007f..df19a499c2 100644
--- a/includes/db/Database.php
+++ b/includes/db/Database.php
@@ -3211,7 +3211,7 @@ abstract class DatabaseBase implements IDatabase, DatabaseType {
 	 * @param DBMasterPos $pos
 	 * @param int $timeout The maximum number of seconds to wait for
 	 *   synchronisation
-	 * @return int Zzero if the slave was past that position already,
+	 * @return int Zero if the slave was past that position already,
 	 *   greater than zero if we waited for some period of time, less than
 	 *   zero if we timed out.
 	 */
diff --git a/includes/db/LoadBalancer.php b/includes/db/LoadBalancer.php
index 0379e8627b..db4ed602b6 100644
--- a/includes/db/LoadBalancer.php
+++ b/includes/db/LoadBalancer.php
@@ -331,17 +331,23 @@ class LoadBalancer {
 	/**
 	 * Set the master wait position and wait for ALL slaves to catch up to it
 	 * @param DBMasterPos $pos
+	 * @param int $timeout Max seconds to wait; default is mWaitTimeout
+	 * @return bool Success (able to connect and no timeouts reached)
 	 */
-	public function waitForAll( $pos ) {
+	public function waitForAll( $pos, $timeout = null ) {
 		wfProfileIn( __METHOD__ );
 		$this->mWaitForPos = $pos;
 		$serverCount = count( $this->mServers );
+
+		$ok = true;
 		for ( $i = 1; $i < $serverCount; $i++ ) {
 			if ( $this->mLoads[$i] > 0 ) {
-				$this->doWait( $i, true );
+				$ok = $this->doWait( $i, true, $timeout ) && $ok;
 			}
 		}
 		wfProfileOut( __METHOD__ );
+
+		return $ok;
 	}
 
 	/**
@@ -363,11 +369,12 @@ class LoadBalancer {
 
 	/**
 	 * Wait for a given slave to catch up to the master pos stored in $this
-	 * @param int $index
-	 * @param bool $open
+	 * @param int $index Server index
+	 * @param bool $open Check the server even if a new connection has to be made
+	 * @param int $timeout Max seconds to wait; default is mWaitTimeout
 	 * @return bool
 	 */
-	protected function doWait( $index, $open = false ) {
+	protected function doWait( $index, $open = false, $timeout = null ) {
 		# Find a connection to wait on
 		$conn = $this->getAnyOpenConnection( $index );
 		if ( !$conn ) {
@@ -386,7 +393,8 @@ class LoadBalancer {
 		}
 
 		wfDebug( __METHOD__ . ": Waiting for slave #$index to catch up...\n" );
-		$result = $conn->masterPosWait( $this->mWaitForPos, $this->mWaitTimeout );
+		$timeout = $timeout ?: $this->mWaitTimeout;
+		$result = $conn->masterPosWait( $this->mWaitForPos, $timeout );
 
 		if ( $result == -1 || is_null( $result ) ) {
 			# Timed out waiting for slave, use master instead
-- 
2.20.1