[08:14:27] 10Traffic, 10Operations: Replace Varnish backends with ATS on cache text nodes - https://phabricator.wikimedia.org/T227432 (10ema) [08:14:35] 10Traffic, 10Operations: Replace Varnish backends with ATS on cache text nodes - https://phabricator.wikimedia.org/T227432 (10ema) p:05Triage→03Normal [08:17:29] 10Traffic, 10Operations: Replace Varnish backends with ATS on cache text nodes - https://phabricator.wikimedia.org/T227432 (10ema) [09:41:52] 10Traffic, 10Operations, 10Patch-For-Review: Rename role::cache::upload_ats to role::cache::upload - https://phabricator.wikimedia.org/T227328 (10ema) 05Open→03Resolved a:03ema Done. [09:41:54] 10Traffic, 10Operations: Replace Varnish backends with ATS on cache upload nodes - https://phabricator.wikimedia.org/T226589 (10ema) [09:42:34] 10Traffic, 10Operations: Replace Varnish backends with ATS on cache upload nodes - https://phabricator.wikimedia.org/T226589 (10ema) 05Open→03Resolved a:03ema All cache_upload nodes are now using ATS instead of Varnish for on-disk caching. Closing. [10:53:41] 10Traffic, 10Operations, 10Pybal, 10Patch-For-Review: Upgrade LVS servers to stretch - https://phabricator.wikimedia.org/T177961 (10MoritzMuehlenhoff) 05Open→03Resolved a:03MoritzMuehlenhoff With the recent replacement of lvs100*, all LVS servers are running Stretch. [15:36:35] hello! The nginx submodule is the only remaining one in ops/puppet, I am wondering if we could fold it in too [15:38:09] elukey: at the end there can be only one :-P [15:38:10] * volans hides [15:43:27] yeah we should [15:43:54] I think we've asked for it / talked about it in the distant past, but never made the effort since it has a low rate of change and there were others [15:48:03] the way I did the analytics ones was a trick suggested by Joe (using environments in ops/puppet) but it doesn't preserve history.. if we want something different for nginx, we'll need to find a different procedure :) [15:48:47] what are "environments" in this context? [15:49:09] AIUI, I don't think we have a functional need for the submodule split of nginx or anything similar [15:49:32] it was just done back in a different time when it seemed that things like an "nginx" module should be abstract enough for reuse/upstreaming, and that such modules should be submodules. [15:50:32] in ops/puppet there is a environments/production directory [15:50:40] but I don't think our nginx module is sufficiently-generic (and it may not be worth the time to make it so, especially given the whole nginx-- thing), was never upstreamed anywhere AFAIK, and it being a submodule generally hurts us when working on related things like tlsproxy [15:51:10] the procedure that I followed for zookeeper for example was [15:51:43] https://gerrit.wikimedia.org/r/#/c/520267/ [15:51:51] https://gerrit.wikimedia.org/r/#/c/operations/puppet/+/520271/ [15:52:14] so removing the submodule and adding "modules/zookeeper" under the environments/production dir [15:52:30] and then moving it back under /modules [15:52:56] ah you mean as a hack to prevent breakage of others' pulls? [15:54:23] in theory you can do a merge without a common base, too (as in, merge the history of modules/nginx back into ops/puppet, perhaps as part of the hack above) [15:54:24] I think that it is also safer for puppet-merge, but I am a bit out of my comfort-zone of puppet :) [15:54:29] elukey: don't kill me but I had to do a manual step when upgrading my local puppet recently, but me just me as I run a script to update branches [15:54:46] but the result is a git repo with multiple oldest-ancestor commits instead of the usual single origin commit [15:54:58] which probably has some fallout we don't anticipate somewhere down the line [15:55:23] the linux kernel has made several such merges over the years, but it's iffy [15:56:56] I doubt the nginx module has a ton of valuable history anyways, we can probably live with the loss, maybe record the history of the commitmsgs in the new squashed commitmsg or something for posterity [15:57:15] there are only 35 total commits in it [15:57:50] that is also an option yes [16:08:56] https://en.wikipedia.org/wiki/ATS_(programming_language) <- Source of future confusion re: "ATS" :) [19:23:08] 10Traffic, 10Varnish, 10Operations, 10Reading-Infrastructure-Team-Backlog, 10Maps (Tilerator): Tilerator should purge Varnish cache - https://phabricator.wikimedia.org/T109776 (10Mholloway) a:05Mholloway→03None