Open Stack

On 30/05/16 17:59, Eugen Block wrote:
> Hi *,
> 
> we're running Openstack on top of Leap 42.1. Since switching to Mitaka,
> we're experiencing severe delays when accessing the dashboard - i.e.
> switching between "Compute - Overview" and "Compute - Instances" takes
> 15+ seconds, even after multiple invocations.
> 
> I've had a look at the requests, the Apache logs and our control node's
> stats and noticed that it's a single call that's taking all the time...
> I see no indications of any error, it seems that once WSGI is invoked,
> that call simply takes its time. Intermediate curl requests are logged,
> so I see it's doing its work. Looking at "vmstat" I can see that it's
> user space taking all the load (Apache / mod_wsgi drives its CPU to
> 100%, while other CPUs are idle - and no i/o wait, no system space etc.).
> 
> Current test, switching to "Overview" took 13.09 secs loading
> http://control1/project/. Switching back to "Instances" took 14.7 secs
> for http://control1/project/instances/. Ain't no fun using the dashboard
> that way... and we did not see this with previous releases.
> 
> Do others see the same slowness or is it likely some (system or
> Openstack) configuration issue on our side?
> 
I believe, I saw something comparable, but thought it was due to my setup.

You mentioned, you already identified a single call taking all the time:
which one is it? Could you please file a bug?

https://bugs.launchpad.net/horizon

It could be either Horizon or nova API (or both) being slow.
-- 
Matthias Runge <mrunge at redhat.com>

Red Hat GmbH, http://www.de.redhat.com/, Registered seat: Grasbrunn,
Commercial register: Amtsgericht Muenchen, HRB 153243,
Managing Directors: Paul Argiry, Charles Cachera, Michael Cunningham,
                    Michael O'Neill