Description of problem: Add the ability to obtain a debug dump of the system state for an OpenStack service based process. eg upon sending SIGUSR1 trigger a dump of all native threads, green threads, live config and any other relevant info See this upstream proposal https://wiki.openstack.org/wiki/GuruMeditationReport (Filed against Nova, but applicable to all)
@Daniel Berrange: were you thinking something like this: http://pastebin.test.redhat.com/146078 (`git apply` to nova) (apply this: http://pastebin.test.redhat.com/146081 to python-novaclient to easily access without having to manually curl TENANT_ID/guru)
No, this is not at all what I was considering for design. Integration into the RPC service, nova client, etc is absolutely not something that is intended. This needs to be a capability that is ultra reliable in the face of broken arbitrarily OpenStack infrastructure. As such it must not depends on things like the RPC system, database, or REST interfaces. I wrote a proof of concept here https://github.com/berrange/nova/commits/gurumed2 however, I did not submit this upstream because I was intending to generalize it into a python package that is broadly independent of openstack code. Further alot of the same infrastructure is required for the related system tracing work in https://bugzilla.redhat.com/show_bug.cgi?id=913619
Oh, OK. That wasn't exactly clear from the description and wiki page. I'll take a look at the proof of concept and see if I can generalize it/integrate it into some of the framework that I wrote.
Patch submitted upstream to oslo-config to integrate oslo.report into Oslo
patch to Oslo merged upstream under 'openstack.common.report'. Now to push functionality into nova...
Patch for integrating functionality into nova posted upstream.
Since this is targeting 5.0 and the change made it into Icehouse, I'm closing this "UPSTREAM"