[10:05:41] hello on-clalers [10:05:43] uff [10:05:46] retrying [10:05:50] hello on-callers! :D [10:06:28] I am merging https://gerrit.wikimedia.org/r/c/operations/puppet/+/1090450, that should disallow calls to the docker registry's catalog (/v2/_catalog) from the outside internet [10:23:09] of course I am totally stupid, nginx on registry nodes will always see a 10.x IP for the request, since from the outside il will come from cpxxxx hosts [10:23:17] * elukey cries in a corner [13:29:22] <_joe_> elukey: the edge forwards the original client ip in X-Client-IP IIRC [13:30:33] _joe_ I think so yes, either we add a requestctl rule (probably easier) or something in the nginx config that inspects xff or x-client-ip [13:30:58] <_joe_> I wouldn't use requestctl for this kind of stuff though :) [13:31:32] yep agreed but it is easier for sure :) context in https://phabricator.wikimedia.org/T378618 [17:26:26] jhathaway, moritzm, is the 'puppet-dev' project still useful? It is unclaimed on https://wikitech.wikimedia.org/wiki/News/Cloud_VPS_2024_Purge and has lots of VMs in it. [17:28:45] andrewbogott: I have not been actively using it, I'm not sure if moritzm has [18:54:55] when reimaging a server with os and data partitions, is it possible to retain the data partition? It's not critical, but it would save a couple hours of data transfer per host [18:55:54] generally yes, it's possible [18:58:23] hmm, is it also automatable? I'm trying to put together a plan for migrating ~100 hosts and reimaging seems cleanest. I probably have to work up some changes to our rolling restart to have a reimage task that i would integrate this into, maybe [18:59:48] short answer 'yes' with an 'if', long answer 'no' with a 'but' [19:00:25] lol, sounds about right :) [19:01:02] is it okay if retaining the data doesn't always work? [19:01:24] yea, this is a cluster that keeps copies of data, it will just have to bring in ~500GB over the network [19:01:52] oh not even that bad [19:05:20] ebernhardson: do you know the partman recipe being used on these hosts atm? [19:06:48] cdanis: hmm, not sure. Certainly it varies. cloudelastic* and relforge* servers have many hdds, elastic* servers have a couple ssds, checking puppet [19:09:33] looks like `partman/{standard,raid10-6dev}.cfg`, `partman/{standard,raid0}cfg`, `partman/{standard,raid0,raid0-2dev}.cfg`, `partman/{standard,raid10-4dev}.cfg` depending on server [19:10:25] heh well for some of those there exists a `reuse-` variant like `reuse-raid10-4dev.cfg` [19:10:56] there's documentation of sorts in https://gerrit.wikimedia.org/r/plugins/gitiles/operations/puppet/+/refs/heads/production/modules/install_server/files/autoinstall/scripts/reuse-parts.sh [19:11:07] cdanis: excellent, thanks [19:21:28] andrewbogott: welcome back! I haven't been actively using puppet-dev in 2024, this was the initial test bed for a lot of Puppet 7 work, but now that everything runs in prod I think it can go away [19:22:05] OK. can you mark the linked page with a note about keeping or deleting? thanks! https://wikitech.wikimedia.org/wiki/News/Cloud_VPS_2024_Purge [19:22:05] either removing all VMs in there and keeping it or nuking the project altogether works for me [19:22:13] will do