| pranali | abhishekk, rosmaita, dansmith, croelandt, mrjoshi glance weekly meeting in 20 minutes at #openstack-meeting | 13:40 |
|---|---|---|
| pranali | Not sure who all are back from PTOs this week, kindly please let me know your availability so that we can go ahead with this meeting | 13:41 |
| rosmaita | pranali: i'll be there | 13:41 |
| rosmaita | (and thanks for the reminder) | 13:42 |
| pranali | rosmaita, ohh gr8 ! | 13:42 |
| pranali | I think Cyril is gonna be back next week | 13:43 |
| abhishekk | I will be thre | 14:00 |
| pranali | ack , starting the meeting | 14:00 |
| dansmith | abhishekk: you need ack from me on what? the ceph regression? | 15:05 |
| abhishekk | no, overall work | 15:05 |
| abhishekk | location patch | 15:05 |
| dansmith | okay I don't remember the state it was in, but I thought you had a number of issues that needed resolving.. are those all fixed now? | 15:11 |
| abhishekk | I think so, including the workaround for ceph issue | 15:14 |
| abhishekk | and with the latest change your DNM patches are green as well | 15:14 |
| * abhishekk will be back after dinner break | 15:15 | |
| dansmith | abhishekk: I haven't really even looked at the ceph issue, but we did merge this last year: https://review.opendev.org/c/openstack/glance_store/+/884524 | 15:44 |
| dansmith | although I thought it was much later in the year than June | 15:44 |
| dansmith | I was thinking it might be related, but maybe not given how early it was | 15:44 |
| dansmith | it seems like just moving ceph back to pacific is not a good solution if pacific doesn't match what we're actually going to run on the current openstack/distro/ceph releases | 15:45 |
| dansmith | gouthamr: thanks, abhi just dropped a few minutes ago but said he'd be back after dinner | 16:39 |
| gouthamr | dansmith following this from https://meetings.opendev.org/meetings/glance/2024/glance.2024-01-04-14.00.log.html#l-36 ; i misread the comment on the bug; the issue seems to occur on all supported releases of ceph: pacific (soon to be EOL), quincy and reef.. a prior version: octopus (EOL) doesn't seem to have this issue.. I wanted to check if there's a ceph bug reported | 17:52 |
| gouthamr | (https://tracker.ceph.com/projects/rbd/issues?set_filter=1&tracker_i | 17:52 |
| gouthamr | d=1).. there were some changes in the remove image codepath like this one that added a progress callback: https://github.com/ceph/ceph/commit/6dc8249c7ba2553acb686a2462de93239f1f07b4 ; i wonder if its related.. maaybe ImageIsBusy is raised by ceph, but the deletion continues in the background | 17:52 |
| * gouthamr grr; that got garbled | 17:52 | |
| gouthamr | ignore the last line.. | 17:52 |
| dansmith | okay but I don't know why this is happening only with the new location proposal and not everything else in the gate | 17:53 |
| dansmith | seems like that means that we're doing something different, which is not the intent here.. the mechanics between nova and glance are changing, but the ceph stuff on the backend should be the same | 17:53 |
| gouthamr | ack; was this workflow (create image, start download, delete image in parallel - see failure and re-attempt download) tested well before? | 18:01 |
| dansmith | yeah, what is being tested hasn't changed, just the API with which we do those steps | 18:06 |
| dansmith | that's my reason for being suspicious :) | 18:06 |
| gouthamr | ah.. that's a pickle | 18:08 |
| dansmith | yeah, so I don't want to just say "this new API is clearly all good, except for some weird reason we have to roll back to a nearly EOL ceph to make it work, oh well" :) | 18:10 |
| abhishekk | Only thing changed here is setting image active and then calculating the hash (which needs to download the image data from ceph) | 18:10 |
| abhishekk | also I don't think we have a test which deletes the image while download is in progress (practically it is not possible to have such test) | 18:12 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!