[Openstack] Ceilometer Alarming Service

Geoffrey Tran gtran at isi.edu
Wed Nov 4 03:11:32 UTC 2015


Hello,

I am trying to utilize ceilometer to have alarms triggered based on user-defined data.  Each
instance will write data to the same meter.  

An example is:
root at hp3:~# ceilometer sample-list -m testMeter
+--------------------------------------+-----------------+-------+--------+------+----------------------------+
| Resource ID                          | Name            | Type  | Volume | Unit | Timestamp                  |
+--------------------------------------+-----------------+-------+--------+------+----------------------------+
| 3e87382a-8d09-4647-b3ba-1769bbadb4a3 | testMeter | gauge | 2.0    | %    | 2015-11-04T03:01:37.034000 |
| 3e87382a-8d09-4647-b3ba-1769bbadb4a3 | testMeter | gauge | 2.0    | %    | 2015-11-04T03:01:36.592000 |
| 3e87382a-8d09-4647-b3ba-1769bbadb4a3 | testMeter | gauge | 2.0    | %    | 2015-11-04T03:01:31.810000 |
| 3e87382a-8d09-4647-b3ba-1769bbadb4a3 | testMeter | gauge | 2.0    | %    | 2015-11-04T02:50:17.831000 |
| 3e87382a-8d09-4647-b3ba-1769bbadb4a3 | testMeter | gauge | 1.0    | %    | 2015-11-04T02:49:09.872000 |
+--------------------------------------+-----------------+-------+--------+------+----------------------------+

This works so far, and does trigger the alarm that I created with the following data POSTed to a server:
{"severity": "low", "alarm_name": "testMeter", "current": "alarm", "alarm_id": "1d733207-7e8e-4471-9004-351206a2bb8d", "reason": "Transition to alarm due to 1 samples outside threshold, most recent: 2.0", "reason_data": {"count": 1, "most_recent": 2.0, "type": "threshold", "disposition": "outside"}, "previous": "insufficient data"}

However, when running multiple instances, it is difficult to determine which instance and which sample
was the one that triggered the alarm.  For example, if the above case had two instances write the same
value of "2" to the meter, the service processing the alarm would not know which instance triggered
it. 

Is there a better way to do this besides separating each instance to a different meter? That would involve 
also creating multiple alarms. 

Thanks,
Geoffrey




More information about the Openstack mailing list