| opendevreview | Merged openstack/glance-specs master: Update new location APIs spec https://review.opendev.org/c/openstack/glance-specs/+/883491 | 04:13 |
|---|---|---|
| opendevreview | Andrew Bogott proposed openstack/glance_store master: rbd: compute appropriate resize amount before resizing image https://review.opendev.org/c/openstack/glance_store/+/885581 | 13:02 |
| opendevreview | Andrew Bogott proposed openstack/glance_store master: rbd: compute appropriate resize amount before resizing image https://review.opendev.org/c/openstack/glance_store/+/885581 | 13:06 |
| abhishekk | dansmith, croelandt, mrjoshi, rosmaita, glance weekly meeting at #openstack-meeting | 13:56 |
| opendevreview | Merged openstack/python-glanceclient master: Docs generation: mock the six module for autodoc https://review.opendev.org/c/openstack/python-glanceclient/+/885505 | 14:18 |
| abhishekk | that was quick | 14:19 |
| opendevreview | Dan Smith proposed openstack/glance_store master: WIP: Add per-store weight config element https://review.opendev.org/c/openstack/glance_store/+/885595 | 16:17 |
| dansmith | rosmaita: could you have a glance over my suggestion (and POC) here? https://review.opendev.org/c/openstack/glance-specs/+/881951/3/specs/2023.2/approved/glance/store-id-strategy.rst | 16:20 |
| dansmith | should be quick I think | 16:20 |
| dansmith | I'm just trying to avoid the need for the operators to define another list-of-store-names, which isn't very friendly for modular deployment stuff where you might put a config file per store in a directory in multiple deploy steps | 16:21 |
| abhishekk | dansmith, so on get call we will just reshuffle (or sort) the locations based on weight, right? | 16:28 |
| dansmith | abhishekk: I haven't seen your PoC, but you should be able to create a list of stores just like you do based on name, but sorted(..., key=weight) right? | 16:28 |
| abhishekk | yeah | 16:29 |
| abhishekk | we need changes in glance-store as well as glance | 16:30 |
| dansmith | abhishekk: you saw my glance_store poc right? | 16:30 |
| dansmith | that should be all we need AFAIK | 16:31 |
| abhishekk | yes | 16:31 |
| abhishekk | that + sort logic at glance download call | 16:32 |
| abhishekk | because the above code which I have written will be enabled only if location_strategy is enabled | 16:32 |
| dansmith | right, but the latter should be one ine | 16:33 |
| abhishekk | easier one | 16:33 |
| dansmith | but right now, we can just always do the sort, and if no weights are set, it's a no-op | 16:33 |
| abhishekk | ack | 16:33 |
| abhishekk | just gathering information from spec point of view | 16:33 |
| dansmith | cool | 16:34 |
| abhishekk | this will be much easier than configuring location strategies | 16:34 |
| abhishekk | thanks, I will make sure spec will be up before next meeting | 16:35 |
| dansmith | ack | 16:36 |
| dansmith | I think this will be easier too | 16:36 |
| dansmith | and certainly easier for deployment tools | 16:36 |
| rosmaita | dansmith: just got back from lunch, will take a look | 17:01 |
| dansmith | rosmaita: thanks, it was mostly because abhishekk had gone for the day (I thought), but definitely good to get you to sanity check the idea | 17:10 |
| rosmaita | dansmith: seems reasonable ... i'm sure you thought of this, we need to document some kind of deterministic behavior when there are equal weights on stores | 17:23 |
| dansmith | I'm not sure why we would need to do that.. today it's not determinisitic because they're all effectively unweighted right? | 17:24 |
| dansmith | or deterministic even | 17:24 |
| dansmith | this makes it behave the same today if they're not given any weights | 17:25 |
| rosmaita | well, you get location order | 17:25 |
| dansmith | order in which locations were added you mean? | 17:25 |
| rosmaita | well, we could document that equal weights behave nondeterministically, and if you care, make sure you use unique weights | 17:26 |
| dansmith | that's stable perhaps but not observable I imagine, and certainly not for a regular user that doesn't know about those things | 17:26 |
| dansmith | yeah, that'd be my preference, but it would also be trivial to sort by weight,name just so it's deterministic | 17:26 |
| rosmaita | we just need to make a choice and document it so we can tell bugs from intended functioning | 17:27 |
| dansmith | sure | 17:27 |
| opendevreview | Andrew Bogott proposed openstack/glance_store master: rbd: compute appropriate resize amount before resizing image https://review.opendev.org/c/openstack/glance_store/+/885581 | 19:01 |
| opendevreview | Merged openstack/glance master: Test glance RBAC old defaults https://review.opendev.org/c/openstack/glance/+/883602 | 23:33 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!