| rsarto | @mrunge: can you give more details on how that depends on my setup? | 12:10 |
|---|---|---|
| mrunge | rsarto: sorry, I do not remember your question | 12:11 |
| rsarto | mrunge: no problem, this is my question: Now that Panko has been deprecated, where are the events stored? | 12:13 |
| mrunge | ah! | 12:13 |
| mrunge | They are not. | 12:13 |
| mrunge | some are still stored in gnocchi | 12:13 |
| mrunge | inserted as metric in gnocchi | 12:13 |
| mrunge | that does not really solve any problem | 12:13 |
| mrunge | rsarto: first of all, you are the first one to ask for panko in a couple of years | 12:14 |
| mrunge | my proposal would be to store events e.g in a elasticsearch instance | 12:15 |
| mrunge | or in loki, or else | 12:15 |
| rsarto | mrunge: so we're just removing panko from the stack and/or trying to use gnocchi for some concepts... | 12:15 |
| mrunge | so, panko was not really used for years | 12:15 |
| rsarto | hmmm... i did not know that. | 12:15 |
| mrunge | before events were both inserted into panko and into gnocchi | 12:15 |
| mrunge | the issue with panko is, it does not scale | 12:16 |
| mrunge | if it worked for you: great! | 12:16 |
| mrunge | however, you may run into load issues sooner or later | 12:16 |
| rsarto | so, let me ask you another question. We're currently using openstack Ussuri, in that case, can I just stop using panko? | 12:17 |
| mrunge | yes, in almost all cases: yes | 12:18 |
| rsarto | understand that, in our case we just want to use default openstack telemetry metrics. Is that the case? | 12:19 |
| mrunge | yes | 12:19 |
| mrunge | just metrics: you don't need panko | 12:19 |
| mrunge | just remove the panko line from event_pipeline.yaml | 12:19 |
| rsarto | another question if allow me... what is the evolution strategy for openstack telemetry? | 12:20 |
| mrunge | what do you mean by that? | 12:21 |
| rsarto | are we going to keep the other components (ceilometer, gnocchi and aodh) as part of it? | 12:22 |
| mrunge | I'd keep ceilometer and aodh | 12:22 |
| mrunge | we have seen issues with gnocchi in the past, but are also struggling getting a replacement | 12:23 |
| mrunge | gnocchi has many nice things you would want to have, and alternatives don't have | 12:23 |
| mrunge | like high availability and multi-tenancy | 12:23 |
| rsarto | I understand that. | 12:23 |
| rsarto | what is a current list of data generated by core openstack services (nova, neutron, keystone, glance, cinder, heat, etc.) to ceilometer ? | 12:24 |
| rsarto | I've only found a from ocata in https://docs.openstack.org/ocata/admin-guide/telemetry-data-collection.html | 12:25 |
| mrunge | that are event types | 12:26 |
| mrunge | the issue with events is: they are defined in the components, such as nova, neutron etc. | 12:26 |
| mrunge | you have to go through the code there. unfortunately, they do not use the same methods to publish events. | 12:26 |
| mrunge | are you interested in how that looks like? | 12:27 |
| mrunge | otherwise I do not look for code locations and move on with my life | 12:27 |
| rsarto | besides the code, Is there any reference/documentation about the events defined in the components? | 12:28 |
| mrunge | no, there is not. Or at least I am not aware of | 12:28 |
| mrunge | e.g here: | 12:29 |
| mrunge | https://review.opendev.org/c/openstack/cinder/+/796806/2/cinder/transfer/api.py | 12:29 |
| mrunge | this adds notifications in cinder when a volume snapshot gets transferred | 12:29 |
| rsarto | do you know if there is a way to customize in the components which event is generated? | 12:30 |
| mrunge | that is also incorporated in metrics (you want the owner getting updated in gnocchi) | 12:30 |
| mrunge | if you look at the review I linked, you see the event itself is hard-coded | 12:30 |
| mrunge | it doesn't make much sense to send out a different event instead of the one I linked, no? | 12:31 |
| rsarto | ok, I got it, they're always sent by the components. | 12:31 |
| mrunge | if you look at https://review.opendev.org/c/openstack/ceilometer/+/798940 | 12:32 |
| mrunge | you'll see how events are configured on the ceilometer side | 12:32 |
| mrunge | *that* is configurable | 12:33 |
| mrunge | e.g you can configure what payload is sent, on which events it is sent etc. | 12:33 |
| mrunge | or when the data is updated | 12:34 |
| rsarto | so, Does ceilometer use this configuration to accept/reject the events? | 12:34 |
| mrunge | if you remove them from the config, they'll get ignored | 12:35 |
| mrunge | that does not events from being emitted | 12:35 |
| rsarto | cool. Can you advise me any documentation/tutorial about the telemetry? I would like to get more knowledge about this stuff. | 12:36 |
| mrunge | what do you want to learn? | 12:36 |
| mrunge | otherwise, there are a few talks on youtube from previous OpenStack summits | 12:36 |
| rsarto | as user perspective, I would like to understand the concepts and how to check the measurements for the metrics. | 12:37 |
| mrunge | and if you are struggling with telemetry, I have recently written some notes on how to debug it | 12:37 |
| mrunge | http://matthias-runge.de/2021/12/06/debugging-openstack-metrics/ | 12:37 |
| mrunge | ^^ that one would be a first step | 12:37 |
| rsarto | super cool mrunge! | 12:38 |
| rsarto | do you know any reference on how to use the openstack metric cli? | 12:39 |
| rsarto | for really beginners... | 12:39 |
| mrunge | uhm, not really. that's why I had that in the document | 12:40 |
| mrunge | ... how to get metrics collected from a vm | 12:40 |
| mrunge | also gnocchi has some concepts not really covered in their upstream docs | 12:40 |
| mrunge | ... what is a archive policy, how does it play together with retrieving metrics, etc. | 12:41 |
| rsarto | we're currently using openstack-helm to run openstack in containers... | 12:42 |
| mrunge | and I'm pretty sure you've seen the component docs: https://docs.openstack.org/aodh/xena/admin/ | 12:42 |
| mrunge | I can not comment on that, that's another installer project | 12:43 |
| rsarto | I'm supposing here that the ceilometer configuration could be addressed as helm overrides. Do you know something about this? | 12:43 |
| mrunge | I've never used openstack-helm | 12:43 |
| rsarto | no problem, thank you! | 12:44 |
| mrunge | out of curiosity, if you are using somethign like a container platform: what is missing there for you? | 12:44 |
| rsarto | It was really nice to talk to you, thank you for your time and support. | 12:44 |
| mrunge | ... what does OpenStack give you, what you don't already have? | 12:45 |
| rsarto | I think we're in a step before... We're trying to understand what openstack offers us. | 12:46 |
| mrunge | also: we (with a different hat) are working on something named service telemetry framework, which then sends out metrics and events from OpenStack to prometheus | 12:46 |
| mrunge | ah, okay. got you | 12:46 |
| rsarto | yes, can I have more information about the telemetry framework? | 12:47 |
| mrunge | https://infrawatch.github.io/documentation | 12:47 |
| rsarto | Is it going to replace all the legacy components (ceilometer, aodh, gnocchi)? | 12:48 |
| mrunge | it depends | 12:50 |
| mrunge | at the moment: no | 12:50 |
| mrunge | ceilometer is used in both | 12:50 |
| mrunge | (OpenStack Telemetry and Service Telemetry Framework) | 12:51 |
| rsarto | Does the service telemetry framework have a stable release? | 12:52 |
| mrunge | yes | 12:52 |
| rsarto | what openstack version is it compatible with? | 12:53 |
| mrunge | We use it with Queens and later | 12:53 |
| mrunge | newer features are compatible with train | 12:54 |
| mrunge | if you look at the documentation, the necessary steps are documented for the case you are using tripleo | 12:55 |
| rsarto | i'm understanding that the Service Telemetry Framework is another option for telemetry, it won't replace the Openstack Telemetry at least for now. Am I right? | 12:55 |
| rsarto | i'm understanding that the Service Telemetry Framework is another option for telemetry, it won't replace the Openstack Telemetry at least for now. Am I right? | 12:56 |
| *** outbrito_ is now known as outbrito | 13:02 | |
| mrunge | correct | 14:11 |
| rsarto | mrunge do you know if is there any panel to show telemetry metrics in horizon dashboard? | 18:15 |
| mrunge | rsarto: none. iirc. | 19:02 |
| mrunge | There was a panel inside the admin panel, but I believe we disabled it quite a while ago | 19:02 |
| *** tobias-urdin3 is now known as tobias-urdin | 20:10 | |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!