The MW code for getting pt-heartbeat lag checks the row for Master_Server_Id (from the slave status), which for master => slave1 => slave2, would check for the slave1 id entry (which wouldn't even exist since only the true masters get table updates for their server IDs)
Description
Description
Details
Details
Subject | Repo | Branch | Lines +/- | |
---|---|---|---|---|
Make getLagFromPtHeartbeat() always use the LB cluster master entry | mediawiki/core | master | +164 -23 |
Status | Subtype | Assigned | Task | ||
---|---|---|---|---|---|
Resolved | aaron | T88445 MediaWiki active/active datacenter investigation and work (tracking) | |||
Resolved | aaron | T111266 Make LoadBalancer slave lag check and read-only mode more robust (for example, using pt-heartbeat) | |||
Resolved | aaron | T119648 Fix getLagFromPtHeartbeat() to use the pt entry for the root master, not the immediate one |
Event Timeline
Comment Actions
Change 256875 had a related patch set uploaded (by Aaron Schulz):
Make getLagFromPtHeartbeat() always use the LB cluster master entry
Comment Actions
Change 256875 merged by jenkins-bot:
Make getLagFromPtHeartbeat() always use the LB cluster master entry