Author: metatron
Description:
Since start of upgrade process, I'm experiencing lots of Lock wait timeouts / deadlocks.
Usecase:
Xtools creates (temporary) tables in schema s51187__xtools_tmp (newly created in this new "combined" instance) to analyze data. Currently this breaks in ~ 9 out of 10 calls due to lock wait timeouts/deadlocks.
This only happens on the new "combined" instance. eg. accessing it via s[45].labsdb or dewiki.labsdb or similar.
While perfoming the identical analysis on an "old" shard eg. s1.labsdb/enwiki.labsdb everything works fine. There is also an s51187__xtools_tmp schema - but an "old" one.
Samples:
(enwiki ~25,000 edits)
https://tools.wmflabs.org/xtools/ec/index.php?user=TParis&lang=en&wiki=wikipedia&uselang=en&purge=1
(dewiki ~24,000 edits)
https://tools.wmflabs.org/xtools/ec/index.php?user=Anton-kurt&lang=de&wiki=wikipedia&uselang=de&purge=1
Questions:
- Is this behaviour transient? If not, please make it go away.
- Is new user-schema-storage slower than the older one? And is this realated?
- Is there some process that interferes with user schemas?
Version: unspecified
Severity: critical