| *** mhen_ is now known as mhen | 02:22 | |
| opendevreview | Abhishek Kekane proposed openstack/glance stable/2025.2: Fix FileNotFoundError when registering task operations https://review.opendev.org/c/openstack/glance/+/969017 | 07:17 |
|---|---|---|
| opendevreview | Abhishek Kekane proposed openstack/glance master: Cache PropertyRules instances in Gateway https://review.opendev.org/c/openstack/glance/+/968766 | 07:34 |
| croelandt | rosmaita: Hey! Did you get a chance to look at the logs frmo Thursday meeting? We were trying to figure out what to call the parameters for https://review.opendev.org/c/openstack/glance-specs/+/963239 | 13:14 |
| croelandt | dansmith: ^ | 13:14 |
| rosmaita | croelandt: i kind of don't care about the parameters tbh, but need a bit more time to think about the cache situation | 13:27 |
| rosmaita | thanks to abhishekk for pointing it out in his comments, btw | 13:27 |
| rosmaita | i had seen it and thought it was fine, but on rereading, it would be good if we could have consistent return code behavior with the new stuff | 13:27 |
| rosmaita | though i am not sure that is possible | 13:27 |
| abhishekk | rosmaita: o/ I think not everyone uses caching though | 13:28 |
| rosmaita | interesting ... i would have thought everyone would use it! | 13:29 |
| abhishekk | I remembered to have few customers in past which were not using glance caching | 13:30 |
| rosmaita | well, maybe we handle it through documentation ... if an operator wants "require" to work correctly, they must not use caching | 13:30 |
| abhishekk | Yeah, I am going to add it in documentation about the same | 13:30 |
| abhishekk | other than this I am not sure how we are going to have consistent behavior with this | 13:32 |
| abhishekk | may be option to bypass caching? | 13:32 |
| rosmaita | don't know ... like you point out in the spec, the caching happens in the middleware, so the request wouldn't even hit the api | 13:37 |
| rosmaita | so we'd have to handle it in the middleware | 13:37 |
| rosmaita | which i guess is do-able? | 13:38 |
| abhishekk | Can be, need to have a look, but I think the 1st step is document this behavior and then later we can think about enhancing it to baypass caching | 13:39 |
| abhishekk | We definitely have query param access in middleware I guess, so can be doable | 13:39 |
| rosmaita | the issue is, a 200 response from GET images/{id}/file?stores=one&required=true could mean the image is coming from store one, or could mean it's coming from the cache ... you don't know | 13:43 |
| rosmaita | i guess maybe that doesn't matter? | 13:43 |
| abhishekk | yeah, it doesn't matter | 13:43 |
| abhishekk | because it's the same case now with download api call | 13:44 |
| rosmaita | ok, but what is the use case for having the "require" there at all? does the list of stores really need to be anything other than a suggestion? | 13:46 |
| abhishekk | I think that should be a additional support for deployer to mention that you need to have image in this store others may be disabled for them? | 13:50 |
| rosmaita | looks like the case would be "Operators cannot test if a specific store is working" (under the current scheme) | 13:50 |
| rosmaita | but if caching is enabled, they can't test this anyway | 13:50 |
| abhishekk | They can, by creating new image :D | 13:50 |
| rosmaita | that's true | 13:51 |
| rosmaita | i guess it depends on how you GET images/{id}/file ... if the idea is "give me that image data (unless it is not public or community or not shared with me)", then it would make sense that the store list is just suggestions for the order to try if the image isn't cached | 13:56 |
| abhishekk | yep | 13:59 |
| rosmaita | that makes me think that "required" isn't necessary for this ... we just want to allow the requestor to pick the store order that will be tried, but we figure if they request a file, they mean it | 14:01 |
| rosmaita | i mean, they could do an image-show to see what stores the image is in if they really care, before they make the download request | 14:02 |
| abhishekk | either works for me, my initial idea is was to give a store preference | 14:03 |
| rosmaita | ok, i will leave a comment on the spec ... thanks for the discussion! | 14:05 |
| abhishekk | NP, thank you for your time | 14:15 |
| rosmaita | croelandt: left a comment on the spec | 14:31 |
| abhishekk | rosmaita: thank you, does that means just stores query param is enough? | 14:35 |
| rosmaita | yes, from my perspective | 14:35 |
| abhishekk | ack, i will wait for dansmith and croelandt about there suggestions and then update the spec | 14:36 |
| dansmith | I'm fine with just the store list and fall back to any store | 14:38 |
| abhishekk | cool thank you!! | 14:38 |
| opendevreview | Abhishek Kekane proposed openstack/glance-specs master: [spec] Download image from specific store https://review.opendev.org/c/openstack/glance-specs/+/963239 | 14:50 |
| rosmaita | "AI Overview: "Glance-specs" can refer to several products or services, including the specifications for Mobile Pixels Glance portable monitors, the Mobile Pixels Glance Pro, or the design specifications for the OpenStack Glance project. " | 15:30 |
| dansmith | rosmaita: lol | 18:02 |
| croelandt | rosmaita: damn, now I don't know which one this channel is about! | 18:21 |
Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!