[openstack-dev] Collecting openstack exceptions/errors, blueprint RFC

Pitucha, Stanislaw Izaak stanislaw.pitucha at hp.com
Wed Apr 10 14:27:01 UTC 2013

Hi all,
I created blueprint at https://blueprints.launchpad.net/oslo/+spec/consistent-exception-notifications and intend to work on getting it implemented to collect some consistent information from the message queue.
But I hope I'm not the only one interested in getting a collection like this to work. I could find some references to nova-sentry existing around 2010 and there's a general interest in reviewing exceptions (tempest? https://blueprints.launchpad.net/nova/+spec/no-stacktraces-in-logs). Maybe someone is working on something similar now?

I'm really interested in a proper structured, rich information rather than just remote log collection over syslog or similar solutions.

So the main questions are:
- What are the current systems people are using for error collection? Stacktach has some overlap here, but with a much bigger scope.
- Are any specific stacktrace formats desirable? Are you using any sentry-like system? I'd prefer just sending over the mq what's already described in https://sentry.readthedocs.org/en/latest/developer/interfaces/index.html#sentry.interfaces.Stacktrace but if other system is more popular, maybe it's worth using their format instead.
- Is anyone working on unifying the format of notifications or solves message processing in a good way? Currently it looks like run_instance, terminate_instance, api.fault and others in nova have attributes all over the place - sometimes inconsistent, sometimes duplicated.

Any other comments / proposals are welcome.
(I'll be at the summit, so if anyone wants to discuss this over a pint, I'll be happy to do that)

Stanisław Pitucha
Cloud Services 
Hewlett Packard

More information about the OpenStack-dev mailing list