Details
-
Improvement
-
Resolution: Unresolved
-
Critical
-
5.0.0
-
None
Description
Background
There can be a number of different causes for an operation to timeout, to name just a few:
- JVM garbage collection
- Network issues
- Disk performance problems.
Problem
A user needs the ability to quickly determine if the root cause is on the Server Side or not. Today that is done via looking at the logs and checking the timing stats via `cbstats` and `mctimings`.
Suggested Solution
It would be good if there were latency graphs in the UI showing the 50, 99, 99.9 percentiles for all operational (N1QL, Data (Reads/Writes), FTS, Views) . This will allow the user to quickly see if the issue is on the server side. Further more it would be good to include the disk latency too.
Attachments
Issue Links
- relates to
-
MB-30000 make histogram of KV operation metrics from cbstats timings available via REST
- Closed