Limit of 10 saves per 30 seconds was used in https://gerrit.wikimedia.org/r/#/c/298523/. We should evaluate whether it is good or should be changed.
Description
Details
Status | Subtype | Assigned | Task | ||
---|---|---|---|---|---|
Resolved | None | T140865 Follow up work after the CX downtime incident | |||
Resolved | Nikerabbit | T140615 Decide on cxsave ping limiter limits and monitor |
Event Timeline
For past 7 days we have had over 100k hits to this limiter: Logstash
Many at 10, some up to 20, few up to peaks of 500.
Based on chat, we think these limits are reasonable and rather than increasing them, we should change the code so that it happens less frequently.
Change 302879 had a related patch set uploaded (by Nikerabbit):
ext.cx.translation.storage: increase debounce delay to 3s
Change 302879 merged by Nikerabbit:
ext.cx.translation.storage: increase debounce delay to 3s
I am re-purposing this task to monitor that the changes we do makes it so that only in exceptional cases the current limits will be hit.
Now waiting for the mentioned patch to be deployed and be live for multiple days to observe change if any.
The patch doesn't seem to have helped much. We still have lots of logs for reaching the limits.
Change 304804 had a related patch set uploaded (by Nikerabbit):
TranslationStorage: improvements to save throttling