[13:23:04] \o [13:25:09] ebernhardson I added you to https://gerrit.wikimedia.org/r/c/operations/puppet/+/1287889 if you want to review. It's OK if not, but FYI I'll probably merge/reimage relforge today unless you need me to hold up [15:44:15] inflatador: so basically, now it installs and runs a docker image? [15:46:21] hmm, what did i do wrong in tests...cindy failed to lookup cirrustestwiki.mediawiki.mwdd when it should be cirrustestwiki.mediawiki.local.wmftest.net/ [15:49:14] ebernhardson yeah. This is just a starting point, we'll probably have to set some envvars similar to https://gerrit.wikimedia.org/r/plugins/gitiles/operations/puppet/+/refs/heads/production/modules/profile/manifests/wmcs/striker/docker.pp#27. [15:50:06] yea that should be ine [15:50:08] fine [15:54:06] Cool, thanks for taking a look [16:58:34] ebernhardson another thing we could do is turn the relforge hosts into dedicated k8s workers, similar to what we're doing with the new WDQS (T425653) . Basically they would be in the same control plane but be reserved for Search Platform workloads. Any preference (docker puppet role vs k8s)? [16:58:35] T425653: Add the wdqs::alternative nodes to the dse-k8s clusters - with a taint to avoid normal jobs being scheduled - https://phabricator.wikimedia.org/T425653 [17:02:11] inflatador: plausible? But i worry about yak shaving. It's re-using the existing opensearch stuff so probably shouldn't be too much trouble? [17:31:15] ebernhardson yeah, I hear ya. In a perfect world, WMCS would have 256 GB VMs or baremetal hosts we could just bring up and down when we needed them [17:35:42] anyway, it sounds like I'll have to punt till next week until my team has a bigger discussion on what to do with relforge. Will also talk to everyone in Search Platform stand up first to see what y'all prefer