[10:37:32] <dcausse>	 lunch
[13:34:19] <ebernhardson>	 \o
[13:35:10] <dcausse>	 o/
[15:29:51] <inflatador>	 .o/
[15:31:34] <dcausse>	 o/
[15:55:05] <dcausse>	 heading out
[16:00:26] <ebernhardson>	 randomly curious, semantic search p50s jumped from 170ms to 230ms at 10:50 today, wonder what changed
[16:01:59] <ebernhardson>	 per SAL, at 10:58 one of the dse-k8s hosts went for reimaging. Timing lines up, but doesn't seem like it should be related
[16:02:35] <ebernhardson>	 not a big deal, 230ms is still fine, just curious
[16:56:56] <inflatador>	 ebernhardson interesting, I guess it behooves us to check and make sure the ureadahead was applied after the move. Let me check which pods are the newest
[16:59:28] <inflatador>	 `opensearch-semantic-search-masters-5` and `opensearch-semantic-search-masters-6` are the newest, running on `dse-k8s-worker1013.eqiad.wmnet` and `dse-k8s-worker1028.eqiad.wmnet`, respectively. Let me check
[16:59:44] <ebernhardson>	 oh i didn't think to check if a pod started at that time
[17:04:00] <inflatador>	 `opensearch-semantic-search-masters-6` has the correct read_ahead
[17:34:24] <inflatador>	 In a mtg for the next 90m, but I think we're good on that front
[18:24:48] <ebernhardson>	 Looked closer into the archive search problem, it's not an easy fix :(  The problem is we use the page id as the doc id in archive search.  But really old logs don't always have a log_page, so they are missing.  And pages can be delete->restore->move->delete, which generates two logs in two different namespaces for the same page id
[18:25:23] <ebernhardson>	 plausibly the solution is to accept some duplication and use log_id as the docid, but that's a larger process
[20:15:12] <inflatador>	 Looks like a PR landed for the x-request-id stuff: https://github.com/opensearch-project/OpenSearch/pull/21048