[07:20:25] an-worker1080 is down, I will take a look [14:04:42] akosiaris: copying question from tgr asked in -operations a while ago: is it okay to deploy via helm on a friday? [14:32:03] Urbanecm: not authoritative answer - if urgent I'd say yes, otherwise it might be better to wait for Monday/Tuesday (so less potential impact before some SREs go on easter holiday / weekend) [14:32:44] (urgent to me means user impact etc..) [14:33:27] got it, thanks [14:59:30] Urbanecm: yeah, exactly what elukey said -- depending on what you're touching, the potential blast radius is quite a bit lower than a mediawiki deploy, so we're probably more amenable... but especially before a holiday weekend, if it can wait let's wait [14:59:53] in fact, any time elukey says "not authoritative answer" you should probably just ignore that and do whatever he says [15:00:04] hehe [15:01:34] rzl: <3 (I woudn't dare bypassing akosiaris' opinion :P) [15:02:18] rzl: it is about linkrecommendation. It's not used by other wikis than beta and testwiki. The reason why we asked about a deploy is that it does not recognize certain content pages anymore, and we're making sure the service gives good results by manually testing it (which should be done by Monday). But, I figured out running the service locally is actually pretty easy, so I'll do that instead :) [15:02:49] ahh got it [15:03:16] yeah, that's a great example of a smaller blast radius [15:03:41] +1 [15:03:42] so as long as testing locally is easy, great let's do that, but if you needed to deploy I think it's probably fine [15:04:30] *that* answer isn't authoritative though ;) worth checking with a deployment-pipeline expert in case they have a more informed opinion about downstream effects I haven't thought of [15:04:49] it's easy for me, probably not for other community ambassadors. But also it looks that I'm the only one who experienced the error, so... [15:05:19] nod [15:05:52] one of the nicest things about the deployment pipeline is decoupling services from each other, so that it's *safer* to touch things on a Friday -- eventually we might be able to stop treating that whole day of the week as hauntd [15:05:54] *haunted [15:06:17] and especially preproduction services like this one [15:06:41] but, I don't celebrate Easter so that's easy for me to say :P [15:07:11] i would probably do some family visits...if we weren't still locked down [16:36:07] Urbanecm: it's probably lower blast radius and personally, knowing that the service isn't used yet I 'd be ok. But keep in mind we don't have a process for non emergency deploys so if it broke and paged right before a holiday and outside of a deployment window, people wouldn't be happy. [16:36:34] given you found an alternative already, use that. [16:36:50] thanks akosiaris [16:37:07] 📝 [21:24:50] Hello 👋. A patch I expect to get merged soon will require generating a secret on a deployment server (e.g. `MWCryptHKDF::generateHex(32)` from eval.php) and setting it as a secret in PrivateSettings. What is the usual process for setting secrets/is there any documentation on the protocol? Also, given I don't currently have access to any of the servers, is this a deploy that the sre team could help with? [21:54:45] nray: you just need a MediaWiki deployer to add it to PrivateSettings for you, probably during a backport window [21:56:11] I think it's more in the releng team's domain fwiw [22:00:53] k, thank you legoktm [22:41:08] nray: i'd recommend writing a patch to wmf-config that adds it to PrivateSettings.example.php and then in the backport window asking the deployer to generate the real one at the same time.