[11:31:38] claime: I've changed the OS request in https://phabricator.wikimedia.org/T418919 and https://phabricator.wikimedia.org/T418925 to Trixie [11:31:54] I saw yes, thanks [14:08:54] working my way down https://wikitech.wikimedia.org/wiki/Kubernetes/Add_a_new_service to https://wikitech.wikimedia.org/wiki/Kubernetes/Add_a_new_service#Set_resource_requests_and_limits_for_your_service/containers, it seems like at this point you're supposed to have a chart and/or helmfile, what is the canonical doc for that? [14:09:43] or perhaps a better question is: what are next steps after https://wikitech.wikimedia.org/wiki/Kubernetes/Add_a_new_service#Setting_up_Ingress [15:09:31] urandom: I guess first question is, do you already have a docker image? [15:09:39] cdanis: yes [15:10:14] great [15:10:19] https://wikitech.wikimedia.org/wiki/Kubernetes/Deployment_Charts is a decent starting point [15:10:25] have you Helmed before? [15:11:25] if you mean have I created (or edited) a chart, then no [15:11:34] I've deployed existing services tho [15:11:57] I'm kind of wondering here about the (terribly named) aqs-http-gateway chart [15:12:15] it seems to be used by the data-gateway [15:13:29] according to the commit message, it is the combination of an (also terribly named) cassandra-http-gateway chart, and a druid-http-gateway chart...apparently created for the new-style aqs services [15:13:37] but data-gateway doesn't use druid [15:14:11] so maybe it would make sense to use it here as well? [15:15:51] urandom: I think it'd help to give some more context, what service is this? [15:15:59] urandom: also, if you'd prefer to talk sync, I was just about to announce serviceops/kubernetes office hours :D [15:16:21] Raine: oh, when is that happening? [15:16:57] I was thinking every 2 weeks on Tuesday, so in 2 weeks because of my other calendar constraints, for a start [15:17:07] but if you want, you can drop something in my calendar sooner [15:17:23] ok [15:17:31] btw, the context is: https://docs.google.com/document/d/15SLV1_mMI7IKyYi4uQxuPb4BmBMgDlo0SeUEfFUeAcM/edit?tab=t.0#heading=h.bl3xbxwaibcz [15:18:16] ok, so a new service, the details of which aren't done yet? [15:18:35] reading https://phabricator.wikimedia.org/T414112 ... [15:18:37] well, the details for the mvp are done [15:19:10] and I don't expect anything that would subsequently change to affect the service deployment [15:19:26] is it just an http app? [15:19:31] it is, yeah [15:19:37] we have the (terribly-named) python-webapp chart [15:19:44] which is probably what you want in that case [15:19:47] one that connects to a cassandra cluster [15:19:49] yes we should rename it : D [15:19:51] mhm [15:19:52] and needs to be connected to [15:20:06] so I used data-gateway for comparison above, it is very close to it [15:20:58] I'm not familiar with data-gateway, so I need a sec [15:23:25] but the general thing we have in progress is that we should have a generic "http app" chart [15:23:40] claime: I think you were working on turning the python-webapp chart into that? [15:24:04] and that would be apropos for something ingress and egress? [15:24:15] s/something/something needing/ [15:24:20] yeah, I believe so [15:24:23] k [15:25:49] I wasn't yet but yeah [15:26:19] federico.3 submitted a generic chart, I don't remember if it was a copy of python-webapp or not, I haven't had time to review it yet... [15:26:53] but basically yeah that would be what you want, use the python-webapp chart, it has nothing python specific and has all the bits and pieces needed for a standard webservice behind ingress [15:27:51] how would cassandra connectivity be configured? because we have more than one way now (among cassandra-using deployments), and some of them are awful [15:29:25] external services if it exists [15:32:04] I'm guessing that must be a proper noun... [15:33:54] urandom: let me check if it does exist [15:36:43] urandom: see how it's done in services/_aqs2-common_/global-{eqiad,codfw}.yaml [15:37:11] external_services: definition in there [15:37:17] that will add the right egress rules [15:38:20] auh, ok yes, that is the one way that isn't awful [15:39:16] yeah that's the last implemented one and the canonical way to add egress to defined services like kafka, cassandra, redis, etc. [15:44:22] so... there are probably many things then that could be migrated to such common chart, no? I mean, the aforementioned data-gateway, but also sessionstore and echostore (to use examples I care for). [15:44:48] those latter two are annoying when it comes to cassandra cluster changes, and could benefit from the external services approach [15:46:09] does it make sense to uses the aqs-http-gateway chart for this now, and migrated them all to The One True Chart later? [15:49:37] s/migrated/migrate/ [15:50:45] indeed, Somebody Should Do Something :D [15:51:10] external services support can also be added to existing charts it's just a sextant module [15:52:33] ok, but I presumably we want to use common charts when there isn't a need for bespoke ones? [15:52:43] s/but I/but/ [15:53:25] yes [15:53:44] so perhaps you could just start with the python-webapp chart for the new service [15:53:56] (sorry about the name :D but other than that it should be good to go, I think) [15:54:18] ok