[15:41:31] Amir1: looking at T328872 for perf impact [15:41:32] T328872: Commons: UploadChunkFileException: Error storing file: backend-fail-internal; local-swift-codfw - https://phabricator.wikimedia.org/T328872 [15:41:38] https://grafana-rw.wikimedia.org/d/000000559/mediawiki-action-api-breakdown?orgId=1&from=now-60d&to=now&timezone=utc&var-module=upload&var-query=&var-Percentile=0.5 [15:41:59] Looks like Sep 11-12 is when a big improvement happened, since then, less clear. [15:42:14] the patches rolled this month though [15:43:08] https://sal.toolforge.org/production?p=0&q=s4&d=2025-09-12 [15:43:12] s4 primary switch perhaps? [15:45:04] The improvement starts sudden on 01:00 AM UTC sharp on Sept 11 [15:45:05] https://grafana-rw.wikimedia.org/d/000000559/mediawiki-action-api-breakdown?orgId=1&from=2025-09-10T20:13:52.850Z&to=2025-09-11T08:32:09.613Z&timezone=utc&var-module=upload&var-query=&var-Percentile=0.5 [15:45:32] https://sal.toolforge.org/production?p=5&q=&d=2025-09-11 [15:45:54] 00:47 … Repooling after maintenance db2240 (T402763)', diff saved to https://phabricator.wikimedia.org/P83223 [15:45:54] 00:45 … finished single-replica PHP 8.3 pilot on shellbox-constraints - T403284 [15:45:54] T402763: Drop rc_new from recentchanges table in wmf production - https://phabricator.wikimedia.org/T402763 [15:45:54] T403284: Migrate production Shellbox services to PHP 8.3 - https://phabricator.wikimedia.org/T403284 [15:46:16] I doubt they had any effect [15:46:26] for my patch, I saw it in flamegraph [15:46:52] upload rate raised at that time [15:47:04] I guess a persistnet upload bot started around then with relaivelyh small files that upload quickly [15:47:07] and has continued to this day [15:47:22] nearly trippled [15:47:36] so nothing got faster there, just more fast requests were added [15:48:14] https://performance.wikimedia.org/arclamp/svgs/daily/2025-10-23.excimer-wall.all.fn-Upload.reversed.svgz vs https://performance.wikimedia.org/arclamp/svgs/daily/2025-09-23.excimer-wall.all.fn-Upload.reversed.svgz [15:48:48] consistency check was 20% of the time in all upload operations [15:50:16] https://performance.wikimedia.org/arclamp/svgs/daily/2025-09-23.excimer-wall.all.fn-Upload.svgz?s=consist [15:50:18] ack, nice [15:50:42] I saw that pre-merge as well, you did an excimer profile in the commit message [15:50:58] yeah [15:51:00] but was hoping to see something absolute afterward, but harder to see [15:51:33] I was hoping to see less GET on swift side but there are so many read GETs that the hange got drowned [15:53:08] actually maybe it can show itself in HEAD? [15:59:30] nope, it doesn't do HEAD [16:04:22] duesen: hehe, you're welcome xD - RE: T404739 [16:04:22] T404739: "kube-env: command not found" when in GNU screen - https://phabricator.wikimedia.org/T404739