| *** cooper6581 has quit IRC | 00:30 | |
| *** sapd1 has joined #openstack-telemetry | 02:04 | |
| sapd1 | Hi everyone. | 02:04 |
|---|---|---|
| sapd1 | I'm trying to setup autoscale using senlin with aodh/ceilometer. | 02:05 |
| sapd1 | follow this guide: https://docs.openstack.org/senlin/latest/scenarios/autoscaling_ceilometer.html | 02:05 |
| sapd1 | I have a question about aodh. | 02:05 |
| sapd1 | How aodh evaluate a cluster in senlin? When it triggers to scale out/in a cluster? | 02:05 |
| sapd1 | for above example, Does it calculate average cpu util of all instances in cluster. | 02:06 |
| *** gouthamr has quit IRC | 02:10 | |
| *** gouthamr has joined #openstack-telemetry | 02:12 | |
| *** psachin has joined #openstack-telemetry | 02:52 | |
| *** ivve has joined #openstack-telemetry | 04:01 | |
| *** ivve has quit IRC | 04:24 | |
| *** rcernin has quit IRC | 04:47 | |
| *** rcernin has joined #openstack-telemetry | 05:01 | |
| lxkong | sapd1: when aodh is evaluating alarm, it doesn't talk directly with senlin cluster, but with gnocchi. | 05:11 |
| lxkong | https://www.irccloud.com/pastebin/EUDhd9yq/ | 05:12 |
| sapd1 | lxkong, So how gnocchi calculate metric for senlin cluster? I know aodh will send query parameters about cluster_id metadata to gnocchi. how about then? | 05:14 |
| lxkong | ceilometer collects samples, stored in Gnocchi, Aodh query Gnocchi for metrics. If the threshold in Aodh alarm definition is met, then Aodh trigger the actions defined. | 05:16 |
| sapd1 | lxkong, I know this flow, But in the cluster, we have many instances, we only have a threshold for all instances in cluster. | 05:17 |
| sapd1 | assume we have 3 instances in a cluster, with 80, 70, 90% cpu_util. So when threshold is 75. Does it trigger? | 05:18 |
| sapd1 | if we calculate average of cpu_util, the result is 80%. If we compare all instances, maybe AODH will not trigger. | 05:18 |
| lxkong | sapd1: i think that's calculated in Gnocchi, Aodh only get result from Gnocchi | 05:19 |
| lxkong | IIUC, in this case, Aodh will get the result 80% | 05:20 |
| sapd1 | lxkong, I'm looking for this function in source code but I can't find. | 05:22 |
| lxkong | in which source code? Aodh? Gnocchi? | 05:22 |
| sapd1 | both of them. | 05:22 |
| lxkong | sapd1: I'm not familiar with Gnocchi, but for Aodh, it's here https://github.com/openstack/aodh/blob/1b552dd82565157c123925adecd12fc279cbbee0/aodh/evaluator/gnocchi.py#L60 | 05:26 |
| lxkong | in this class, aodh get result from Gnocchi, and do evaluation (defined in the parent class) | 05:27 |
| dangtrinhnt | sapd1, and in gnocchi: https://github.com/gnocchixyz/gnocchi/tree/c960f2934fe63f3bc67b78d7514c877ad17375f2/gnocchi/rest/aggregates | 05:29 |
| dangtrinhnt | That's how the metrics will be processed. | 05:30 |
| sapd1 | dangtrinhnt, lxkong So maybe when use query string `metadata.user_metadata.cluster_id=$MYCLUSTER_ID`, gnocchi will found all metric of this query (It will don't care about instances). Then it will aggregate all series return data point from storage. | 05:33 |
| sapd1 | Thanks, lxkong dangtrinhnt | 05:33 |
| lxkong | np | 05:33 |
| dangtrinhnt | btw, lxkong, since gnocchi has been moved out of OpenStack, do you think it should be a problem if we continue ceilometer? | 05:34 |
| lxkong | dangtrinhnt: what do you mean by 'continue ceilometer'? | 05:37 |
| dangtrinhnt | I means while maintaining ceilometer, would it be any blockages because of gnocchi dependence? | 05:38 |
| lxkong | i don't think so. | 05:38 |
| dangtrinhnt | okie | 05:38 |
| lxkong | they has different project scope, ceilometer is only responsible for sample collection, but don't care the storage | 05:39 |
| lxkong | dangtrinhnt: actually, we are not using Gnocchi | 05:40 |
| lxkong | because we are using a very old version Ceilometer (still has ceilometer-api), so we move the Aodh ceilometer evaluator back | 05:40 |
| dangtrinhnt | ok. I'm asking because I see many Ceilometer features have been deprecated and replaced by gnocchi. I think after we can setup a conference call with you, me and any other contributors and figure out what we gonna do next. | 05:43 |
| lxkong | sure | 05:44 |
| *** rcernin has quit IRC | 06:03 | |
| *** rcernin has joined #openstack-telemetry | 06:47 | |
| *** rcernin has quit IRC | 06:47 | |
| *** rcernin has joined #openstack-telemetry | 06:48 | |
| *** phasespace has quit IRC | 07:02 | |
| *** pcaruana has joined #openstack-telemetry | 07:20 | |
| *** amorin has quit IRC | 07:28 | |
| *** amorin has joined #openstack-telemetry | 07:28 | |
| *** phasespace has joined #openstack-telemetry | 07:46 | |
| *** rcernin has quit IRC | 07:48 | |
| *** rcernin has joined #openstack-telemetry | 08:04 | |
| *** JqckB has joined #openstack-telemetry | 08:36 | |
| *** JqckB has quit IRC | 09:35 | |
| *** JqckB has joined #openstack-telemetry | 09:37 | |
| *** tobias-urdin has quit IRC | 10:06 | |
| *** hoonetorg has quit IRC | 10:39 | |
| *** hoonetorg has joined #openstack-telemetry | 10:52 | |
| *** willksm has joined #openstack-telemetry | 11:51 | |
| *** JqckB has quit IRC | 12:40 | |
| *** JqckB has joined #openstack-telemetry | 12:53 | |
| *** phasespace has quit IRC | 14:36 | |
| *** JqckB has quit IRC | 16:23 | |
| *** JqckB has joined #openstack-telemetry | 16:24 | |
| *** psachin has quit IRC | 17:05 | |
| *** rcernin has quit IRC | 19:36 | |
| *** pcaruana has quit IRC | 20:08 | |
| *** JqckB has quit IRC | 20:09 | |
| *** openstackgerrit has quit IRC | 21:07 | |
| *** supreeth90 has joined #openstack-telemetry | 21:29 | |
| supreeth90 | hey guys, Can somebody review https://review.openstack.org/#/c/642902/ ? This fixes an issue where many instances of inspector are created instead of keeping it singleton. | 21:29 |
| *** JqckB has joined #openstack-telemetry | 21:32 | |
| *** rcernin has joined #openstack-telemetry | 21:43 | |
| lxkong | supreeth90: hi, that patch looks good to me, but neither me nor dangtrinhnt are not in the reviewer team atm :-( | 21:46 |
| supreeth90 | lxkong: thanks for updating, do you know who is in the reviewer team atm ? :-) | 21:47 |
| lxkong | https://review.openstack.org/#/admin/groups/107,members | 21:53 |
| lxkong | but i don't believe most of them are active nowadays | 21:53 |
| lxkong | at least for ceilometer | 21:53 |
| *** rcernin has quit IRC | 21:55 | |
| *** rcernin has joined #openstack-telemetry | 22:04 | |
| supreeth90 | thanks! will try to mail them directly | 22:05 |
| dangtrinhnt | I'll add the core reviewers to your patch. | 22:22 |
| supreeth90 | thank you dangtrinhnt ! | 22:24 |
| dangtrinhnt | np | 22:25 |
| *** JqckB has quit IRC | 23:03 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!