Description
We currently track the aggregate log append write latency, but this doesn't tell the whole story. We've encountered situations in production where kernel and disk issues can lead to a long tail of high latency appends, which are buried in the aggregate.
Attachments
Issue Links
- is blocked by
-
AURORA-118 Add percentiles to @Timed, or write a new decorator to add percentiles
- Open
- is duplicated by
-
AURORA-73 Break out storage latency and log latency into percentiles
- Resolved