[08:08:54] Trey314159 (for later): Celtic Knot is back. Did you want to participate this year (see email). [10:28:27] lunch [11:22:03] errand [13:05:50] greetings [13:46:37] o/ [14:34:48] Wednesday meeting has started: https://meet.google.com/eki-rafx-cxi cc: dcausse, ryankemper, ebernhardson [15:09:23] in another meeting [15:09:36] might join in 1h30 [16:15:36] workout, back in ~40 [16:52:51] back [17:08:02] ebernhardson merged your patch on https://phabricator.wikimedia.org/T309648 , gonna build the new pkg shortly [17:08:25] also, thanks for the update, really clears that up [17:25:15] ohhhh boy, still not getting notification [17:25:59] Giving up on apple calendar, will try thunderbird [17:53:02] lunch, back in an hr [19:04:02] back [19:13:54] inflatador: see https://phabricator.wikimedia.org/T310216 please [19:16:13] we can stop writes if it's causing problems elsewhere [19:16:25] but if the only problem is generating log messages, i'd prefer to leave it [19:16:40] ebernhardson: you'd need to ask releng [19:16:41] RhinosF1 this is for cloudelastic, right? [19:16:53] inflatador: from a very quick glance ye [19:17:52] I can reach out in releng IRC if you think that would help [19:18:02] yes please [19:18:12] They are in charge of the train and that's holding it up [19:18:22] for turning off writes, we just need an extra line in InitialiseSettings.php, easy enough if needed [20:15:10] quick break, back in ~15 [20:48:18] sorry, been back [22:10:27] another PR for the new ES plugin package: https://gerrit.wikimedia.org/r/c/operations/software/elasticsearch/plugins/+/803996 [22:10:54] doh! this is why we need ci to verify things :( [22:11:14] * ebernhardson has a strong preference for writing a program that checks if i did a thing, over remembering do to a thing :P [22:12:03] hmm, maybe could make prepare_commit do it, checking [22:12:07] 100% agree maybe we can put that in jenkins or something? Otherwise, I'm writing playbook for the build process and can include a check there [22:16:41] hmm, not sure how i would check if it's been released before. I was thinking to check apt.wikimedia.org, but it's a bit tedious since the exact apt component will change over time and it doesn't look like we specify the apt component from inside the repo so nothing to reference. hmm [22:18:20] I think we just compare the BUILD_VERSION from the repo to the one in the commit [22:19:35] hmm, i wonder if there are reasons we patch without updating BUILD_VERSION, seems like we always should? checking the git history [22:20:47] I'm trying to think if of a use case for patching without updating BUILD_VERSION, but if the ultimate goal is to make a .deb, I don't think it's possible to build it without incrementing(?) [22:22:04] (well, you can build it, but you can't push it to the apt repo) [22:22:14] yea poking through it seems like we should be fine, there are a few commits that only update docs, but since this is a quick hack into the prepare_commit script and not a full ci integration that wont stop anything [22:23:44] I'd be OK with putting it in CI , we could always override if need be [22:24:21] the difficulty in CI is the actual integration, we don't just drop a .travis.yml in the repo :) [22:25:52] maybe it's a candidate to move to gitlab, with self-serve CI. but scope creep is real..i dunno :) [22:25:59] Point taken, I haven't challenged the CI dragon yet ;) [22:26:36] jenkins isn't terrible, but it's configuration is build through a complex series of templates...i dunno maybe SRE's enjoy that :P [22:27:05] (mostly a jab at how insane k8s config seems to be) [22:27:18] yeah, I have a love/hate relationship with templating [22:27:46] but hey, don't shoot the messenger! Poor YAML didn't ask to do that [22:28:42] lol. we have someone on staff that wrote a yaml library, fairly popular. And they are not much of a fan of yaml anymore :) [22:29:26] I'll take YAML any day over JSON, but that's not saying much [22:29:42] fun fact, json is valid yaml :) [22:29:55] (maybe less fun when you think about parser complexity :P) [22:30:07] yeah, and I'm sure no one ever (ab)uses that fact ;P [22:47:08] i put a patch up, https://gerrit.wikimedia.org/r/c/operations/software/elasticsearch/plugins/+/804004 . I ended up sticking with checking apt.wikimedia.org rather than `git log -1` because i wasn't sure how to differentiate if the user intends to ammend the existing patch, or they are preparing a new patch [22:52:47] ebernhardson: ooh I really like that. one question tho, what's the word `shell` doing here? `$(shell echo $(ELASTIC_VERSION) | cut -d '.' -f 1-2 | tr -d '.')` [23:01:35] ryankemper: it's a makefile, so everything is $(command arg arg arg) [23:01:55] maybe makefile lets you not say it though? I should test :P [23:02:22] ebernhardson: but is there a command called `shell`? that's what's confusing me [23:02:31] it seems like it should just be `$(echo $(ELASTIC_VERSION) | cut -d '.' -f 1-2 | tr -d '.')` [23:02:49] ryankemper: https://www.gnu.org/software/make/manual/html_node/Shell-Function.html [23:03:42] ah I see, so it's necessary to set the variable to the result [23:04:13] ryankemper: try this Makefile out: https://phabricator.wikimedia.org/P29562 [23:04:35] ryankemper: when i run `make works` it echos bar, when i run `make blank` it echos nothing [23:04:44] (i'm not actually sure what it's doing there :P) [23:05:44] * ebernhardson tries to not think about how gnu make isn't the only make :P [23:06:50] ebernhardson: neat example, and yeah I get the same behavior (ofc) [23:10:37] Okay new package is built, will try upgrading package on relforge and see if the snapshot takes now