19 Apr 2024 |
corvus | so maybe when this is done, we just make a db dump, delete mariadb and do that | 21:26:53 |
corvus | * so maybe when this is done, we just make a db dump, delete mariadb and do that, then restore | 21:27:00 |
Clark | ya based on what I could tell the commit to really wide upgrade ranges too so you don't even really need to upgrade often. Made me feel more confident in the ability to do that | 21:27:43 |
corvus | i'm replacing those "alter/rename" lines with drop/create pairs. i only did the rename in the migration so that it would be fast on massive data sets. this data set is small, so it's no big deal to drop/create. | 21:28:03 |
Paladox none | ah ok | 21:28:46 |
corvus | 3 more index renames to muddle through here | 21:29:44 |
corvus | i see a duplicate index on the zuul_provides table; i want to check whether we have that in opendev real quick | 21:34:45 |
corvus | we do not. i'm going to delete the extra key there, because i don't think any migrations after this one will do it, and it's not supposed to be there. | 21:37:43 |
corvus | * we do not. i'm going to delete the extra key in gerrit's db, because i don't think any migrations after this one will do it, and it's not supposed to be there. | 21:37:55 |
corvus | okay, [manual] migration finished. i'll start up zuul again | 21:39:20 |
corvus | it's loading config from git | 21:40:48 |
corvus | 2024-04-19 21:40:12,041 ERROR zuul.GerritConnection: zuul.driver.gerrit.gerritconnection.HTTPConflictException
2024-04-19 21:40:12,044 INFO zuul.GerritConnection: Got default branch for jgit: None
| 21:42:05 |
corvus | something is weird about the jgit repo | 21:42:13 |
corvus | (that's not a fatal error, it's continuing -- just noting that we don't know what the default branch is because of that) | 21:42:33 |
corvus | startup is a little slower than normal since i deleted the git repo cache, so that's warming up at the same time | 21:44:48 |
Clark | jgit is a read only copy of the upstream repo iirc | 21:45:02 |
Clark | that may explain it if the state in gerrit is such that gerrit can't say what the default branch is? | 21:45:17 |
corvus | hrm maybe. we do a GET to projects/%s/HEAD to find the default branch | 21:45:55 |
corvus | curl -o - https://gerrit-review.googlesource.com/projects/jgit/HEAD
project state HIDDEN does not permit read
| 21:47:11 |
corvus | https://ci.gerritcodereview.com/components shiny new (dark mode for me!) zuul is up | 21:48:01 |
Clark | I'm guessing opendev's zuul has the same issue with jgit but we end up cehcking out the submodule sha so it doesn't matter too much for opendev | 21:48:05 |
Clark | not sure if it matters for this zuul as it may end up just always doing the same thing: check out an explicit sha and call it good | 21:48:27 |
corvus | i think we have a backup default default default of master, but, yeah, probably something to keep in mind if we see jgit repo-related problems | 21:48:48 |
corvus | let's merge that change now | 21:49:27 |
Paladox none | awesome! Congrats and thanks! | 21:49:36 |
corvus | anyone else want to +2/submit https://gerrit-review.googlesource.com/c/zuul/ops/+/421497 ? :) or i can self-approve | 21:50:36 |
Clark | done | 21:51:30 |
Clark | oh I just +2'd not submitted | 21:51:36 |
Clark | muscle memory. I'll submit now too | 21:51:42 |
Clark | now it is actually done | 21:51:59 |