[07:00:18] <pfischer>	 o/
[07:04:18] <dcausse>	 o/
[09:53:50] <dcausse>	 lunch
[13:04:25] <inflatador_>	 <o/
[13:10:43] <dcausse>	 o/
[13:29:18] <ebernhardson>	 \o
[13:30:02] <dcausse>	 o/
[14:02:28] <inflatador_>	 dcausse stuck in mtg, will probably be at least 5-10m
[14:06:08] <dcausse>	 inflatador_: np
[14:12:34] <inflatador_>	 dcausse actually, let's go ahead and cancel if that's OK...gotta get some stuff done before Weds mtg
[14:12:43] <dcausse>	 sure
[14:17:23] <inflatador_>	 ACK, thanks
[14:34:52] <pfischer>	 I won’t be able to join the Wednesday meeting today.
[16:12:59] <inflatador_>	 re: cluster quorum...another possibly-silly thing we could do is update to the latest 1.x opensearch
[16:13:55] <inflatador_>	 anyway, workout..back in ~40
[17:06:57] <inflatador_>	 back
[17:34:07] <dcausse>	 dinner
[18:01:20] <inflatador_>	 lunch, back in ~45
[18:41:06] <ebernhardson>	 meh, writing dumps from spark has a few more annoyances that i first thought of :P  All solvable...but annoying
[18:42:18] <ebernhardson>	 like, we have to partition the output by wiki, can't mix them together in a file. But we can't have a single file for big wikis, because that would all go through a single executor
[18:42:57] <ebernhardson>	 so have to first loop through the data to determine partitioning, then a second pass to do the work.  But we stored the data as avro, so the first-pass still has to read the entire text content
[18:43:23] <ebernhardson>	 could use the index without content for the partitioning count...but then that feels odd to count one table and use it to partition another
[18:50:44] <inflatador_>	 back
[19:09:45] <ebernhardson>	 hmm, maybe this would be easier if it entirely skipped pyspark...what we really needs is a simple 1-to-1 mapping from the source .parquet files to a .txt file. Then we can reuse the partitioning that was done at ingestion time
[19:10:07] <ebernhardson>	 except we still need the distributed processing :P  will have to ponder more
[19:36:37] <ebernhardson>	 have to head out early today, but maybe found a solution to above.  might actually be easy and i was making it overcmplicated.
[20:38:05] <inflatador_>	 break, back in ~20
[21:02:09] <ryankemper>	 inflatador_: finishing up lunch, 7 mins
[21:07:33] <inflatador_>	 ACK, just joined
[21:50:15] <ryankemper>	 ebernhardson: I'm going down a phabricator related ticket infinite recursion loop and came across https://gerrit.wikimedia.org/r/c/operations/mediawiki-config/+/1154300 / https://phabricator.wikimedia.org/T391383#10891089. Is that patch still something we want to ship out?