Project | Branch | Lines +/- | Subject | |
---|---|---|---|---|
mediawiki/core | master | +16 -2 | Use sync writes for replication of 'store' operations | |
mediawiki/core | wmf/1.27.0-wmf.14 | +16 -2 | Use sync writes for replication of 'store' operations |
Details
Related Objects
Event Timeline
Is there a backtrace? This could indicate files actually disappearing somewhere (thus SRE-swift-storage).
Logstash didn't provide a backtrace in this instance. Here's what I'm looking at, for reference: https://logstash.wikimedia.org/#dashboard/temp/AVMa6iSgO3D718AO1yx2
I wonder if this is related to 'store' ops with TempFSFile objects for async swift writes...
Change 273545 had a related patch set uploaded (by Aaron Schulz):
[WIP] Avoid async writes for store ops as the tmp file might not live long enough
Change 273553 had a related patch set uploaded (by Aaron Schulz):
Use sync writes for replication of 'store' operations
Change 273545 merged by jenkins-bot:
Use sync writes for replication of 'store' operations
Change 273553 merged by jenkins-bot:
Use sync writes for replication of 'store' operations