Details
Description
memcached and ep-engine do not log crashes or aborts in memcached.log.
So critical runtime failures can sometimes get silently ignored.
This is made even worse if core files are not enabled on any machine and all crash information is lost.
It would be good to have the following..
1) Record a crash reason in memcached.log so the crash time is placed in context with other log messages
2) Dump backtraces for all the threads
3) Dump stats for all buckets - since state of the system at the point of the crash can be most relevant.
4) Allow the crash dump infrastructure to be invoked by an external process in case a hang is detected.
Attachments
Issue Links
- relates to
-
MB-12035 Rationalize and 'fix' memcached / ep_engine logging
- Resolved
Gerrit Reviews
For Gerrit Dashboard: MB-12053 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
43643,2 | MB-12053: Include the initial implementation of Breakpad for Windows | master | memcached | Status: ABANDONED | 0 | -1 |
44475,2 | MB-12053: Set Breakpad libraries variable to empty string if not found | master | tlm | Status: MERGED | +2 | +1 |