You are browsing a read-only backup copy of Wikitech. The live site can be found at wikitech.wikimedia.org
document status: draft
This is a short (<= 1 paragraph) of what happened. While keeping it short, try to avoid assuming deep knowledge of the systems involved, and also try to differentiate between proximate causes and root causes. Please ensure to remove private information.
Roll out of wmf.8 to group1 broke the world.
Who was affected and how? For user-facing outages, estimate: How many queries were lost? How many users were affected, and which populations (editors only? all readers? just particular geographies?) were affected? etc. Be as specific as you can, and do not assume the reader already has a good idea of what your service is and who uses it.
Initial indicators of the issue were picked up in logstash and via logspam-watch on mwlog1001. A large number of Icinga alerts followed.
It seems likely that the primary issue was obscured during the initial deploy by a focus on Parsoid errors.
If automated, please add the relevant alerts to this section. Did the appropriate alert(s) fire? Was the alert volume manageable? Did they point to the problem with as much accuracy as possible?
This is a step by step outline of what happened to cause the incident and how it was remedied. Include the lead-up to the incident, as well as any epilogue, and clearly indicate when the user-visible outage began and ended.
All times in UTC.
- 20:12 brennen: Train wmf.8 roll fowards from group0 to group1 as well (try 1) 
- 20:12 Large amounts of logspam noticed, especially from Parsoid/PHP, and Icinga issues many alerts.
- 20:28 brennen: Train wmf.8 rolled back to just group0 
[Fixes to exclude Parsoid/PHP]
- 23:30 brennen: Train wmf.8 roll fowards from group0 to group1 as well (try 2) 
- 23:30 OUTAGE BEGINS
- 23:30 Large spike in database errors in logstash (T239877), shortly thereafter large amounts of Icinga alerts go off.
- 23:30+ Production group1 and group2 wikis become noticably sluggish, eventually stopping working entirely.
- 23:35 brennen: Attempted train wmf.8 roll back thwarted by canary failures 
- 23:38 brennen: Train wmf.8 rolled back to just group0, again 
- 23:38 OUTAGE ENDS
What weaknesses did we learn about and how can we address them?
The following sub-sections should have a couple brief bullet points each.
What went well?
- for example: automated monitoring detected the incident, outage was root-caused quickly, etc
What went poorly?
- for example: documentation on the affected service was unhelpful, communication difficulties, etc
Where did we get lucky?
- for example: user's error report was exceptionally detailed, incident occurred when the most people were online to assist, etc
How many people were involved in the remediation?
- for example: 2 SREs and 1 software engineer troubleshooting the issue plus 1 incident commander
Links to relevant documentation
Where is the documentation that someone responding to this alert should have (runbook, plus supporting docs). If that documentation does not exist, there should be an action item to create it.
Explicit next steps to prevent this from happening again as much as possible, with Phabricator tasks linked for every step.
NOTE: Please add the #wikimedia-incident Phabricator project to these follow-up tasks and move them to the "follow-up/actionable" column.
- To do #1 (TODO: Create task)
- To do #2 (TODO: Create task)