[09:50:18] lunch [10:17:11] lunch 2 [12:35:51] ejoseph: 1:1? [12:36:13] Oh sorry, forgot your email of this morning. Please disregard! [13:00:28] greetings [13:06:56] dcausse ebernhardson are you OK with us reimaging relforge today? (We'll go one at a time of course) [13:07:22] inflatador: hey, no objections from me [13:09:13] dcausse thanks, we'll get started once we get approval from e-bernhardson [14:43:23] inflatador: should be good [14:44:26] cindy seems happy at last :) [14:44:29] ACK, we can do it at SRE pairing session if that works for everyone [14:44:33] also look at that, a happy cindy :) [14:45:11] now I need to remember how to make jenkins happy with circular dependencies :) [14:45:58] i suppose if anyone is curious, data engineering is hosting a 'how to test an airflow dag' session at top of the hour. I can only imagine whatever they do is very different from what I did :) [14:46:37] :) [14:49:28] I wonder if should allow 6.8 & 7.10 during the transition as some maint scripts will target different es versions than they expect [14:50:48] hmm, on the one hand we don't expect to run maintenance while transitioning, but i guess it's not particluarly hard and if we do need it would be nice if they worked [14:51:08] i guess we also have a multi-day transition if we are going to ride the train for 7.x like before [14:51:15] which means completion suggester has to accept both? [14:51:27] yes that's what I thought [14:51:27] errr, no if we ride the train then it doesn't matter [14:51:34] we can disable it on one cluster tho [14:51:39] but it's easy enough to accept both [14:52:44] yes... unsure what's best here, I'll write this down and see how it looks like [14:54:31] the update path must be working both ways too, current master -> es7, es710 branch -> 6.8 [14:54:50] hmm, yea i suppose we do want it able to go back [14:55:07] even without going back I think [14:55:38] oh right, the updates. I suppose i'm blindly assuming the updates stayed the same shape, at least i dont remember changing anything there, [14:56:11] sigh, i'm totally not thinking :P [14:56:12] me neither so I think it's fine, I will try to double check locally [14:56:20] the updates for the completion suggester go to both clusters, so it needs to work on both clusters [14:56:31] yes, page updates too [14:56:51] page updates i'm guessing are fine, i dont think anything there checks version numbers, and the format is still compatible? [14:57:36] yes I suppose so, I removed some include_type_name in some places tho [14:58:08] hmm, yea 6.8 still needs those, i guess we would have to retain them to keep compatability [14:58:18] ok will check that [14:58:30] switching office, back in ~1h [15:40:54] workout, back in ~30 [16:22:59] back [17:28:47] hmm, why would `./mvnw -pl war dependency:tree` say that org.wikidata.query.rdf.common requires logback-classic at runtime, but `./mvnw -pl common dependency:tree` says no such thing? [17:28:57] (i removed logback-classic from the pom) [17:31:12] so it must be a transitive dep, the tree should help identify the source perhaps? [17:31:26] it says it comes from common jar :) sec lemme copy/paste [17:31:34] oh [17:33:03] beware that -pl might sometimes acts weirdly, e.g. I wonder that if in some cases it pulls other module deps from the local .m2 repo instead of the actual pom in your dev folder [17:33:20] this is for the war, logback at bottom: https://phabricator.wikimedia.org/P27886 [17:33:43] and common: https://phabricator.wikimedia.org/P27885 [17:34:02] hmm, maybe a full run on dependency:tree without pl? checking [17:34:07] same if you run ./mvnw dependency:tree ? [17:34:33] i can add an exclusion to the war pom, but i can only imagine code review if i send a patch that exclude random jars from other jars :) [17:34:53] yes it makes no sense that it's still there [17:35:48] dcausse: doh, ok so it is probably -pl doing that. The full build only shows logback-core [17:36:32] ok I rarely -pl for this reason on this project eventhough it would be very helpful sometimes [17:41:10] yea, -pl is convenient if only to have less output to page through :) can search through the output easy enough though [17:42:04] i also wonder about excluding all the jetty stuff from the bigdata war, but getting far out of scope [17:54:53] lunch, back for SRE session [18:24:18] back [20:44:37] quick break, back in ~15 [21:11:53] back [22:27:37] time to cook some chicken korma! [22:34:24] hmm, i don't seem to have packagist.org in my password manager, log-in via github says invalid credentials, create account says account already exists, and the reset-password doesn't end up sending me any email. Meh :P