Jump to content

Incidents/2023-05-25 eqiad/LVS

From Wikitech

document status: draft


Incident metadata (see Incident Scorecard)
Incident ID eqiad/LVS Start 2023-05-25 14:04
Task T337497 End 2023-05-25 14:25
People paged 2 Responder count 7
Coordinators Janis Affected metrics/SLOs edits per second, rps in general, 5xx responses from CDN, appserver latency
Impact For approximately 15-20 minutes logged in users connecting to to Wikimedia wikis through our Ashburn datacenter and editors in general may have received 503 errors

During scap deploys one of the LVS servers in eqiad was taken down which resulted in multiple servers being depooled making eqiad unable so serve traffic.


All times in UTC.

  • Scap runs at 13:52, 14:08, 14:10
  • 14:04 PyBal disabled on lvs1019 as part of maintenance but a deploy was ongoing OUTAGE BEGINS
  • 14:05 PROBLEM - PyBal backends health check on lvs1019 is CRITICAL: PYBAL CRITICAL - Bad Response from pybal: 500 Cant connect to localhost:9090 (Connection refused) https://wikitech.wikimedia.org/wiki/PyBal
  • 14:08 <icinga-wm> PROBLEM - PyBal backends health check on lvs1020 is CRITICAL: PYBAL CRITICAL - CRITICAL - appservers-https_443: Servers mw1418.eqiad.wmnet, mw1417.eqiad.wmnet, mw1416.eqiad.wmnet, mw1415.eqiad.wmnet, mw1414.eqiad.wmnet are marked down but pooled: parsoid-php_443: Servers parse1017.eqiad.wmnet, parse1011.eqiad.wmnet are marked down but pooled: api-https_443: Servers mw1447.eqiad.wmnet, mw1448.eqiad.wmnet, mw1449.eqiad.wmnet, mw1450.eqiad.wmnet marked down but pooled https://wikitech.wikimedia.org/wiki/PyBal
  • 14:09: PyBal re-started on lvs1019
  • 14:14 bblack started repooling parsoid in
  • 14:17  Incident opened.  Janis becomes IC.
  • 14:21 appservers and api_appservers repooled
  • 14:22 jobrunners and videoscalers repooled
  • 14:25 recoveries coming in OUTAGE ENDS
  • 14:42 still elevated latencies on api_appservers, might be unrelated as there is a spike of read traffic on s7
  • 14:58 s7 query throughput is high due to db maintenance, most likely unrelated
  • 14:58 api_appserver latency trending down
  • 15:08] <jinxer-wm> (MediaWikiLatencyExceeded) resolved: Average latency high: ...


The issue was detected immediately by SRE doing maintenance work. Multiple alerts as well as pages immediately followed.


What went well?

  • Automated alerting detected the issue quickly

What went poorly?

  • We did not manage to avoid this happening again although we already knew about LVS downtime during scap deploys being an issue

Where did we get lucky?

Links to relevant documentation



Incident Engagement ScoreCard
Question Answer


People Were the people responding to this incident sufficiently different than the previous five incidents? no
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)? no
Were pages routed to online (business hours) engineers?  Answer “no” if engineers were paged after business hours. yes
Process Was the "Incident status" section atop the Google Doc kept up-to-date during the incident? yes
Was a public wikimediastatus.net entry created? yes
Is there a phabricator task for the incident? yes task T337497
Are the documented action items assigned? yes task T334703
Is this incident sufficiently different from earlier incidents so as not to be a repeat occurrence? no Same issue as Incidents/2023-04-17 eqiad/LVS
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.

no task T334703
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? yes
Were the steps taken to mitigate guided by an existing runbook? no
Total score (count of all “yes” answers above) 10