| opendevreview | Merged openstack/ceilometer master: Remove references to Ocata and Newton https://review.opendev.org/c/openstack/ceilometer/+/797070 | 09:27 |
|---|---|---|
| indi | Hello | 16:34 |
| indi | Since cpu_util metric was deprecated in recent openstack versions, does anyone have a working heat template for auto scale group for cpu that works on Victoria or Wallaby? | 16:34 |
| indi | I have tried to use the OS::Aodh::GnocchiAggregationByResourcesAlarm, but it seems it does not work properly. After the first time it scale up, the logs start showing wrong numbers for cpu metric | 16:35 |
| mrunge | indi: hi, what kind of archive policy do you use? | 17:17 |
| mrunge | indi: the idea was, to let gnocchi do the calculation for cpu utilitization | 17:18 |
| mrunge | there is https://review.opendev.org/c/openstack/aodh/+/786880 , which is a kind of hack in my eyes | 17:18 |
| indi | Hi mrunge I use the ceilometer-low-rate policy. It is type instance and unit ns | 17:46 |
| indi | aggregation_methods | rate:mean, mean | 17:47 |
| indi | definition | - timespan: 30 days, 0:00:00, granularity: 0:05:00, points: 8640 | 17:47 |
| indi | When I perform the query using gnocchi command it seems correct, but the aodh alarm reports wrong numbers for cpu | 17:48 |
| mrunge | uhm, what do you get? | 18:28 |
| indi | http://paste.openstack.org/show/806839/ | 18:31 |
| mrunge | the values are *a lot* bigger than expected, right? | 18:37 |
| mrunge | other than that, not much should have changed | 18:38 |
| mrunge | indi: and you may be interested in https://berndbausch.medium.com/how-i-learned-to-stop-worrying-and-love-gnocchi-aggregation-c98dfa2e20fe | 18:39 |
| servagem | yes. I thought it both should return the same values, not? | 18:43 |
| indi | mrunge, not sure if I got you | 18:43 |
| indi | I have started the stack with asg, after the first 5 minutes, the alarm is triggered to scale up. A second instance is created. | 18:44 |
| indi | after that more 5 minutes the alarm trigger to scale down. The fist instance get deleted | 18:46 |
| indi | even after several minutes the alarm to scale down keeps in alarm state | 18:46 |
| indi | what is strange is that I have configured cpu stress to runs in cloud-init. So all instances are always 100% cpu used | 18:48 |
| indi | but after the first scale up, the alarm reports numbers a lot lower than expected | 18:48 |
| mrunge | I remember heading a comparable bug report | 19:00 |
| mrunge | ... hmmmm | 19:01 |
| mrunge | strange | 19:01 |
| indi | yep, I haven't found any example on internet that works for recent openstack versions | 19:18 |
| indi | most of examples still use cpu_util metric | 19:19 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!