Open Stack

Hi Yuanjing,

some pointers inline.


2014-04-03 4:33 GMT+02:00 Yuanjing (D) <yj.yuan at huawei.com>:

>  Hi
>
>  I have a requirement of monitoring VMs, if a VM's meter like cpu_util
> become too high, then system generate an alarm for this VM with meter
> information.
>
>  I have tested alarm function of ceilometer, below are commands I used to
> create alarm object with meter and resource id or not:
> ceilometer alarm-threshold-create  --name alarm1 --meter-name cpu_util
> --period 60 --evaluation-periods 1 --statistic avg --comparison-operator gt
> --threshold 1 -q resource_id=757dadaa-0707-4fad-808d-81edc11438aa
>  ceilometer alarm-threshold-create  --name alarm1 --meter-name cpu_util
> --period 60 --evaluation-periods 1 --statistic avg --comparison-operator gt
> --threshold 1
>
>  I have the following question:
> If I have to define alarm object for every VM and every meter?
>
>From my comprehension, yes it's the good and the unique way


> Take 100 VM and 2 meter cpu_util, memory_util as an example, I will have
> to define 100*2 alarm objects for them.
> I think if I just define alarm object with meter not but VM(resource_id),
> then alarm evaluator will count all VM's meter.
>
your're right, Here your alarm will be trigered on the average of all
samples in the period of all VMs .. which is not what you want I'm sure.


>  Another question produced by question above: I know that alarm evaluator
> will process alarm object one by one, so too many alarm object may result
> in performance problems too.
>
On which component have you observed a performance issue ?

This should be better/mitigate if you deploy correctly multiple services
involved in the alarm evaluation: at least the evaluator and the API.
About the backend, I read there is some performance issue with SQL [1] but
seems not about alarming but rather notifications handling.
The prefered backends for performance should be MongoDB or Hbase (on multi
nodes).


>  I am not a ceilometer programmer and I apologize if I am missing
> something very obvious.
> Can you give me some help to make me clear about them and how to implement
> my requirement?
>

>  Thanks
>
>
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
> [1]
http://lists.openstack.org/pipermail/openstack-dev/2014-March/030288.html
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20140404/b5998d91/attachment.html>