maintenance: Add waitForReplication and sleep in migrateRevisionActorTemp

 - waitForReplication is critical in such scripts, it has caused a major
   outage in commons before.
 - The extra sleep is not important for scripts that take a couple of
   hours to finish but it is important for ones that take more than a
   day. Because replication lag to secondary datacenter and non-core dbs
   (cloud, backup source, analytics) is not reflected in waitForReplication
   and running a maint script for an extended period of time will cause
   issues for those.

You can see similar work done on similar scripts such as deleteAutoPatrolLogs.php

Bug: T275246
Change-Id: If6aa0aaadc62949c151a1f675f8aa24041bf3738

1 files changed, 15 insertions, 0 deletions

diff --git a/maintenance/migrateRevisionActorTemp.php b/maintenance/migrateRevisionActorTemp.php
index 93175584b987..26cb4554a22c 100644
--- a/maintenance/migrateRevisionActorTemp.php
+++ b/maintenance/migrateRevisionActorTemp.php
@@ -1,5 +1,7 @@
 <?php
 
+use MediaWiki\MediaWikiServices;
+
 require_once __DIR__ . '/Maintenance.php';
 
 /**
@@ -15,6 +17,12 @@ class MigrateRevisionActorTemp extends LoggedUpdateMaintenance {
 		$this->addDescription(
 			'Copy the data from the revision_actor_temp into the revision table'
 		);
+		$this->addOption(
+			'sleep',
+			'Sleep time (in seconds) between every batch. Default: 0',
+			false,
+			true
+		);
 	}
 
 	protected function getUpdateKey() {
@@ -76,6 +84,13 @@ class MigrateRevisionActorTemp extends LoggedUpdateMaintenance {
 
 			$this->output( "... rev_id=$last, updated $updated\n" );
 			$conds = [ 'rev_id > ' . $dbw->addQuotes( $last ) ];
+
+			// Sleep between batches for replication to catch up
+			MediaWikiServices::getInstance()->getDBLoadBalancerFactory()->waitForReplication();
+			$sleep = (int)$this->getOption( 'sleep', 0 );
+			if ( $sleep > 0 ) {
+				sleep( $sleep );
+			}
 		}
 
 		$this->output(