On Thu, 7 May 2015 19:57:35 -0400
Ralph Bean <rbean(a)redhat.com> wrote:
On Thu, May 07, 2015 at 04:17:29PM -0600, Tim Flink wrote:
> For at least the second time in the last couple weeks, the queue in
> production was way too long (~120 at one point).
>
> To help with this in the future, I added 4 new x86_64 clients to
> production. Hopefully this'll help keep the queue length lower and
> wait times down to something reasonable.
>
> Tim
Is there a graph of the queue length anywhere? If not, would you like
to get it added to collectd?
Our monitoring is pretty primitive, at best. collectd wasn't even
working for qa hosts for a while but that issue was resolved.
I'd love to have monitoring of queue length in production among other
things but hadn't gotten to actually writing/finding and adding them to
collectd/nagios.
We have graphs of the backlog over time on various fedmsg services
that help with debugging from time to time:
http://threebean.org/fedmsg-health-day.html#FMN
That's cool, we'll have to look into something similar for Taskotron at
some point. Thanks for the pointer.
Tim