| jam | manadart_, do you have write access to Thumper's grafana? I do, but we should make sure several people can see the details. | 08:54 |
|---|---|---|
| manadart_ | jam: I assume I don't; none of the menus allow me to change anything. | 08:56 |
| jam | manadart_, you have to "sign in" from the little icon on the bottom left. | 08:56 |
| achilleasa | interestingly, the firewall worker has a per-machine watcher for the units, a per-application watcher etc. Any ideas why that is the case instead of watching all units, applications etc? | 10:06 |
| stickupkid | achilleasa, can only think of granularity | 10:07 |
| stickupkid | achilleasa, but even then it's a weak argument | 10:07 |
| achilleasa | but in a large model... | 10:07 |
| achilleasa | it's one watch loop (==goroutine) for each | 10:07 |
| stickupkid | achilleasa, we totally could optimise for that | 10:07 |
| stickupkid | achilleasa, yeah - hence my "weak argument" | 10:08 |
| achilleasa | yeap... that's my plan | 10:08 |
| achilleasa | because I need to also monitor spaces | 10:08 |
| achilleasa | manadart_: ^^^ this means getting a SpaceInfos API method | 10:08 |
| * stickupkid tries to make witty joke about manadart_ and a dentist | 10:10 | |
| achilleasa | stickupkid: we will have to drill the bits we need out of the API. Good enough? :D | 10:11 |
| stickupkid | hahaha | 10:11 |
| stickupkid | achilleasa, I'm unsure, it might be a painful extraction | 10:12 |
| achilleasa | It will be painful, but I am pretty sure it will be the _crown_ of the refactoring process | 10:12 |
| stickupkid | I think you're just _filling_ in the pieces here | 10:13 |
| stickupkid | We might need to polish some methods and scale the requests to make it whiter than white | 10:14 |
| achilleasa | But as you know, when you introduce new APIs you might get teething problems that need to be addressed, rough edges that you need to be extracted etc | 10:18 |
| achilleasa | however working with FLOSS is fun | 10:18 |
| achilleasa | stickupkid: there is much more room for optimization it seems.... e.g here(https://github.com/juju/juju/blob/develop/worker/firewaller/firewaller.go#L545-L564) we grab instance details for each machine in the model. However, the Instances method now (after the instancepoller changes) accepts a slice of instance.Id | 11:41 |
| achilleasa | soo 500 machines -> 500 API calls to the provider + 500 calls to grab the ingress rules (if we are using instance fw) | 11:42 |
| achilleasa | (although the latter calls might be already cached in the instance data depending on the provider) | 11:43 |
| stickupkid | nice and fast - lol | 11:43 |
| * achilleasa senses another long PR description coming up... | 11:44 | |
| hml | stickupkid: https://github.com/juju/juju/pull/11901 | 12:16 |
| hml | i have to look at static anaylsis and i think there are a couple of tests which need fixing. but can start on the review | 12:16 |
| hml | rebasing to your change was interesting. i think i got it right | 12:17 |
| hml | stickupkid: approved 313 | 14:13 |
| hml | fix for intermittent unit test failure, review please: https://github.com/juju/juju/pull/11903 | 15:00 |
| hml | achilleasa, stickupkid, manadart_ ^^ | 15:00 |
| hml | review please https://github.com/juju/juju/pull/11904 | 16:41 |
| achilleasa | hml: looking at 11904; ticked 11903 | 16:51 |
| hml | achilleasa: ty | 16:51 |
| achilleasa | btw, was the ports one caused by my refactoring changes? I wonder whether I missed a sort call somewhere | 16:52 |
| achilleasa | hml: left some comments | 16:56 |
| hml | achilleasa: reading, ty | 17:26 |
| hml | achilleasa: on the ports thing, is order important? i was wondering, but it’s also a very old test | 17:28 |
| hml | achilleasa: updated 11904 | 17:43 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!