Jump to content

Incidents/2024-09-23 pc1013 hardware issue

From Wikitech

document status: draft

Summary

Incident metadata (see Incident Scorecard)
Incident ID 2024-09-23 replication Start 2024-09-23 14:47:03
Task T375382 End 2024-09-23 15:23:23
People paged 1 Responder count 4
Coordinators jynus Affected metrics/SLOs No relevant SLOs exist
Impact reading any wiki was way slower than before

A memory error triggered an error which killed MariaDB

Timeline

All conversation logs are visible under this archive: https://wm-bot.wmcloud.org/logs/%23wikimedia-operations/20240923.txt

  • 14:47:03: Initial alert from <icinga-wm> indicating that host pc1013 is down with 100% packet loss.
  • 14:47:21: Incident 5267 acknowledged by <sukhe>.
  • 14:48:39: <moritzm> reports a DIMM error in SEL.
  • 14:50:26: <moritzm> attempts to power cycle pc1013.
  • 14:54:13: pc1013 recovers, but concerns about stability remain.
  • 14:57:44: pc1013 crashes again.
  • 14:58:26: Decision to failover to pc1015 is made.
  • 14:59:14: <jynus> restarts pc1015 to apply Puppet changes.
  • 15:02:11: Failover to pc1015 is committed.
  • 15:03:09: Users start reconnecting to the new master.
  • 15:07:38: Recovery of MariaDB Replica Lag and IO on pc2013.
  • 15:09:12: Circular replication setup completed.
  • 15:24:31: <jynus> disables monitoring for pc1013.
  • 15:29:48: Acknowledgment that performance will be degraded for a few hours.

Detection

  • Root Cause: DIMM error in pc1013 causing repeated crashes. Caught by monitoring
  • Resolution: Failover to pc1015 and setup of circular replication.
  • Impact: Degraded performance due to empty cache and increased parser activity.

Conclusions

This is a minor incident which could have been way bigger if jynus was not around because we lack automation to handle that sort of fix which requires SQL coordination as parser cache instances are multi-source (see: Parser cache and Orchestrator). Fixing this issue required to move a spare server to the impacted section, which is still a non trivial operation as there is little documentation and automation around it.

Actionables

Scorecard

Incident Engagement ScoreCard
Question Answer

(yes/no)

Notes
People Were the people responding to this incident sufficiently different than the previous five incidents? yes
Were the people who responded prepared enough to respond effectively yes (?)
Were fewer than five people paged? yes
Were pages routed to the correct sub-team(s)? yes (?)
Were pages routed to online (business hours) engineers?  Answer “no” if engineers were paged after business hours. n/a
Process Was the "Incident status" section atop the Google Doc kept up-to-date during the incident? no A doc was not setup
Was a public wikimediastatus.net entry created? no A docstatus was not created
Is there a phabricator task for the incident? yes
Are the documented action items assigned? yes
Is this incident sufficiently different from earlier incidents so as not to be a repeat occurrence? yes (?) we have identified what to fix so this is actionable in several ways
Tooling To the best of your knowledge was the open task queue free of any tasks that would have prevented this incident? Answer “no” if there are

open tasks that would prevent this incident or make mitigation easier if implemented.

yes
Were the people responding able to communicate effectively during the incident with the existing tooling? yes
Did existing monitoring notify the initial responders? yes
Were the engineering tools that were to be used during the incident, available and in service? n/a
Were the steps taken to mitigate guided by an existing runbook? no
Total score (count of all “yes” answers above) 9
Retrieved from "https://wikitech.wikimedia.org/w/index.php?title=Incidents/2024-09-23_pc1013_hardware_issue&oldid=2232446"