[07:04:55] <dcausse>	 started a backfill of wikidata in codfw using: python3 cirrus_reindex.py codfw wikidatawiki backfill 2024-05-07T23:30:00 2024-05-14T16:15:00
[10:40:45] <dcausse>	 lunch
[12:52:11] <inflatador>	 <o/
[12:52:22] <dcausse>	 o/
[14:27:52] <dcausse>	 quick errand
[15:00:44] <dcausse>	 will be 5mins late for the wed meeting
[16:20:35] <inflatador>	 break, back in ~30
[16:31:16] <ebernhardson>	 dcausse: pfischer: https://gerrit.wikimedia.org/r/c/operations/deployment-charts/+/1031966
[16:56:37] <inflatador>	 back
[17:28:13] <inflatador>	 re: airflow build, hatch seems to be working much better than virtualenv 
[17:32:37] <ebernhardson>	 and the answer is ... $wgCirrusSearchWriteClusters is null, even though in ext-CirrusSearch.php it's set to []
[17:33:18] * ebernhardson wonders if some sort of default value behaviour is turning the empty array into the default null value
[18:31:33] * ebernhardson apparently hasn't dug through config loading in a few years...so many steps :P
[18:39:35] <ebernhardson>	 finally...indeed thats exactly whats happening:// Optimistic: If the global is not set, or is an empty array, replace it entirely.
[18:39:51] <ebernhardson>	 so our empty array gets replaced with the default, just needs to switch to provide_default
[21:27:19] <ebernhardson>	 random fun things, while spark understands hdfs:///foo to mean use the default hadoop cluster,  pyarrow doesn't understand it :P  Not a big deal but gotta add analytics-hadoop to a couple urls
[21:27:53] <ebernhardson>	 default hadoop utils seem to understand that too, seems like a pyarrow bug
[23:30:49] <ebernhardson>	 discolytics updates yesterday didn't 100% work, made some updates and verified it's now working as expected (i hope)