[openstack-dev] [ceilometer][aodh][vitrage] Raising custom alarms in AODH

ROSENSWEIG, ELISHA (ELISHA) elisha.rosensweig at alcatel-lucent.com
Wed Dec 2 12:49:19 UTC 2015

Thanks for your response. It definitely helped clarify things.

Regarding the second point: Say we have 30 different types of alarms we might want to raise on an OpenStack instance (VM). What I understand from your explanation is that when we create a new instance, we need to create 30 new alarms in Aodh that can be triggered some time in the future. If we have 100 instances, we will effectively have 3,000 alarms created in Aodh, and so on with more instances. 

Is this a correct depiction of the situation? If so, do you think it will scale?

A different approach might be to create a new alarm in Aodh on-the-fly. However, we are under the impression that the creation time can be up to one minute, which will cause a large delay. Is there any way to shorten this?



> -----Original Message-----
> From: Julien Danjou [mailto:julien at danjou.info]
> Sent: Wednesday, December 02, 2015 12:26 PM
> Cc: OpenStack Development Mailing List (not for usage questions); AFEK,
> Ifat (Ifat)
> Subject: Re: [openstack-dev] [ceilometer][aodh][vitrage] Raising custom
> alarms in AODH
> On Tue, Dec 01 2015, ROSENSWEIG, ELISHA (ELISHA) wrote:
> > 1. Does AODH currently support raising alarms on resources not
> modeled
> > in OpenStack? For example, raising an alarm on a Switch? Or does each
> > alarm have to relate to a resource ID (or IDs)(
> Yes, Aodh does not really care, especially with the Gnocchi backend. It
> can evaluate any metric on any resource type and just trigger the
> alarm.
> > 2. What we feel is missing is some way to raise an alarm on-the-fly.
> > In Vitrage, we have this concept of "deduced alarms", where based on
> > some analysis Vitrage determines we need to raise an alarm on some
> > resource. As we understand it, currently to raise an alarm in AODH we
> > need to register the alarm in advance, wait for it to be registered
> and only then we can trigger the event.
> > This will delay our response time to events.
> What you need to do, is create the alarm that you want to trigger in
> Aodh. Let's say Vitrage knows that if switch A is going down, it needs
> to send an email to an admin.
> You create an alarm in Aodh that says on event "switch A is down ->
> send a mail to admin". Then Vitrage runs, and just have to emit an
> event "switch A is down". You can do that via oslo.messaging or I guess
> via the REST API (not sure the mechanism is here but it could be I
> guess).
> Then Aodh will trigger the alarm actions for you.
> Next step could be give more job to Aodh, such as determining that
> "switch A is down" by doing some evaluation – unless the switch sends
> an event when it's going down, but I imagine it's not the point. ;-)
> Does that help?
> --
> Julien Danjou
> # Free Software hacker
> # https://julien.danjou.info

More information about the OpenStack-dev mailing list