contrail-api eating 100% CPU and not responding to requests

[Martin Gerhard Loschwitz]

The problem I have described in http://lists.opencontrail.org/pipermail/dev_lists.opencontrail.org/2015-January/001882.html happens several times a day. This renders opencontrail unusable, as neutron requests time out regularly. What would be a good way to debug this?

TLDR: Contrail will eat 100% cpu time after a while and not respond in time until it gets restarted, which resets the cycle.

[Prakash Bailkeri (prakashmb)]

Looked at strace of contrail-api from Martin. Saw GET on virtual-machine-interfaces quite often

Suggested two changes:
1. patch fix from https://review.opencontrail.org/#/c/6786/. Should optimize the list operation
2. heal_instance_info_cache_interval on nova compute to 0(or to higher value than default).

[Prakash Bailkeri (prakashmb)]

Several config fixes for optimization are committed.

https://review.opencontrail.org/#/c/6605/
https://review.opencontrail.org/#/c/7703/
https://review.opencontrail.org/#/c/7062/
https://review.opencontrail.org/#/c/7192/
https://github.com/Juniper/contrail-controller/commit/78aa4a45822db32e201ddbdff846194fbb9cca41
https://github.com/Juniper/contrail-generateDS/commit/28137d4a15cc921b3300bfb150617bff5cc7c0be
https://github.com/Juniper/contrail-generateDS/commit/a96a28280d1010f472a21b5f90e7aa7933b5d74f
https://github.com/Juniper/contrail-generateDS/commit/48550eff29ec91d3645ee678276dc712c340e7a3

[Andrew Woodward (xarses)]

What version is this fix in?