[Openstack-operators] Uptime and SLA's

Matt Fischer matt at mattfischer.com
Thu Jun 2 22:29:23 UTC 2016


We do this a few different ways, some of which may meet your needs.

For API calls we measure a simple, quick, and impactless call for each
service (like heat stack-list) and we monitor East from West and vice
versa. The goal here is nothing added to the DBs, so nothing like neutron
net-create. The downside here is that some of these calls work even when
the service isn't 100% healthy so keep that in mind.

Then we also have a set of "what would a user do" calls like "spin up a VM
and attach a FIP and ssh in" or "create and delete a volume". These run
less often.

Finally we have a reference cloud application that uses our LBaaS, GSLB, HA
routers, and multiple front-end/back-end nodes. This has the highest
expectation of uptime and is used as an example for our customers of how
you can run an app with "more nines" than the underlying infra.

On any of these, especially the first two I mentioned, time series data is
super useful. It's good to know that your create volume times (for example)
are 40% slower after your deploy. We use Monasca and Grafana for that.


On Thu, Jun 2, 2016 at 2:37 PM, Kingshott, Daniel <
Daniel.Kingshott at bestbuy.com> wrote:

> We¹re currently in the process of writing up an internal SLA for our
> openstack cloud, I¹d be interested to hear what others have done and what
> metrics folks are capturing.
>
> My initial thoughts are success / fail spawning instances, creating and
> attaching volumes, API availability and so on.
>
> Can anyone on the list share their insights?
>
> Thanks,
>
> Dan
>
>
> Daniel Kingshott
> Cloud Dude
> (425) 623 4359 - Cell
>
> Best Buy Co. Inc.
> Technology Development Center
> 1000 Denny Way | 8th Floor | Seattle, WA | 98109 | USA
>
>
> _______________________________________________
> OpenStack-operators mailing list
> OpenStack-operators at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20160602/2364f90b/attachment.html>


More information about the OpenStack-operators mailing list