[06:27:52] * arturo checks tools-k8s-worker-nfs-24 [06:34:36] I migrated it to another hypervisor and got the network right this time [06:57:18] * arturo back in a bit [08:50:23] FYI I have created T375259 [08:50:23] T375259: cloud: edge network suffers downtime if one cloudsw is down - https://phabricator.wikimedia.org/T375259 [08:51:06] cc topranks [08:53:13] I'm off. until Mon 30th guys... I'll review logs and such at that stage see if I can work out anything, but I fear the only proper way to do it is to take one cloudsw down and try to replicate which might not be so easy for us [08:54:13] enjoy your time off :-) [09:45:20] arturo: all the cloudsw are now being collected by gnmi and data is in Grafana [09:53:48] XioNoX: cool, really nice! [11:43:42] dhinus: wdyt? https://gitlab.wikimedia.org/repos/cloud/cloud-vps/tofu-infra/-/merge_requests/50 [11:44:05] TIL: https://artifacthub.io/ [11:47:03] also TIL: https://opentofu.org/blog/building-the-opentofu-registry/ [12:45:32] arturo: the MR looks good, though I'm starting to think of a possible refactor [12:45:41] but I think we can merge this one and discuss it separately [12:46:46] the idea (still thinking about the details) is to create a module for each "project" and have all the resources for that project depend from that single module [12:48:02] so e.g. "tools" would be a module (either in tofu-infra or in a separate repo), and it would handle all the resources for the project, including default security group rules, but also flavors, etc. [12:49:15] "cloudinfra" would also be a module (probably a folder in the tofu-infra repo), and everything including yaml files would be inside that folder, if it is about a resource that lives in that project [12:49:33] this is just brainstorming at the moment, but I wanted to share my thoughts with you [12:50:03] it is your idea to facilitate the view on the collection of resources a project has declared? [12:50:30] yep, I think it would keep things more tidy, and would allow to reason about 1 project at a time [12:51:22] I see [12:52:07] I guess we don't really have resources that are cloud-wide, because the cloud-wide ones are part of the `admin` tenant [12:52:19] for something like tools I think it makes a lot of sense, for things like cloudinfra maybe less, but it seems a good abstraction (to be verified!) [12:53:03] I think it is a good idea actually. I can see it working [12:53:04] I think I started thinking about "tools" and then my reasoning was "could we apply the same design to everything, maybe including 'admin'"? [12:54:08] that resembles the env / modules partition that I have seen elsewhere [12:54:24] the other benefit would be, if you're an admin of a project, let's say "foobar", that's not managed by WMCS, you could look at what we have for "tools" and create a similar tofu structure for your own project [12:54:31] maybe reusing some of our higher-level modules [12:54:52] nice [12:55:55] maybe we can create a phab ticket, under T370037 to define this new layout, and potentially track the work to implement it [12:55:55] T370037: Cloud VPS: extend tofu-infra coverage - https://phabricator.wikimedia.org/T370037 [12:56:27] good idea, I will create a subtask [12:57:22] regarding https://gitlab.wikimedia.org/repos/cloud/cloud-vps/tofu-infra/-/merge_requests/50 I'll let that rest for the weekend and merge on monday I think [12:57:47] I plan to go off a bit earlier today anyway, so wont have a lot of time to follow up in case `tofu apply` fails [12:58:07] sounds good [13:00:02] re: opentofu registry, I knew they were working on it, good to see they are making progress [13:00:23] I hope we can publish our own cloudvps provider there [13:00:46] that'd be nice, yes [13:02:12] the blog post is very nice about how they worked around github's rate limits :) [13:04:05] should have keep that secret :-P [13:05:41] :D [13:05:51] it is also fun that to work around a ratelimit they were _punished_ with having to work with XML [13:06:22] * arturo bad joke, sorry [13:06:43] will go for lunch, tty later [13:10:47] RSS is the only XML I don't complain about :P [15:21:25] * arturo offline