My next steps from here will be:
* To add USDT probes to the logs and back_ldbm to get better, fine detail about what is
causing the excessive latency.
* these probes are also needed to resolve what appears to be a symbol resolution issue
with systemtap when optimisations are enabled.
* To add probes in other parts of the code base to get better visualisation about where
and how long timings are taking through an operation.
* To run a lock contention profile (I was unable to do this today due to bugs in
* To document the setup proceedures
* Commit all these into a PR
* Document some actionable improvements we can make to improve the server performance.
mutrace will give you a good audit of lock contention, and it's simple to use, takes
-- Howard Chu
CTO, Symas Corp. http://www.symas.com
Director, Highland Sun http://highlandsun.com/hyc/
Chief Architect, OpenLDAP http://www.openldap.org/project/