| *** Sravan has joined #openstack-glance | 00:01 | |
| *** Sravan has quit IRC | 00:05 | |
| *** _alastor_ has quit IRC | 00:06 | |
| *** Sravan has joined #openstack-glance | 00:13 | |
| *** Sravan has quit IRC | 00:14 | |
| *** Sravan has joined #openstack-glance | 00:20 | |
| *** Liang__ has joined #openstack-glance | 00:33 | |
| *** _alastor_ has joined #openstack-glance | 00:35 | |
| *** _alastor_ has quit IRC | 00:39 | |
| *** rmcallis has joined #openstack-glance | 00:42 | |
| *** rmcallis has quit IRC | 00:45 | |
| *** rmcallis has joined #openstack-glance | 00:49 | |
| *** rmcallis has quit IRC | 00:51 | |
| *** Liang__ is now known as LiangFang | 01:01 | |
| *** Sravan has quit IRC | 01:15 | |
| *** trident has quit IRC | 01:31 | |
| *** trident has joined #openstack-glance | 01:35 | |
| openstackgerrit | zhouxinyong proposed openstack/glance master: Replacing the link in HACKING.rst https://review.openstack.org/618368 | 01:55 |
|---|---|---|
| *** Sravan has joined #openstack-glance | 02:14 | |
| *** Sravan has quit IRC | 02:19 | |
| *** openstackgerrit has quit IRC | 02:22 | |
| *** _alastor_ has joined #openstack-glance | 03:00 | |
| *** _alastor_ has quit IRC | 03:05 | |
| *** openstackgerrit has joined #openstack-glance | 03:06 | |
| openstackgerrit | Liang Fang proposed openstack/glance master: Fix for FK constraint violation https://review.openstack.org/617889 | 03:06 |
| LiangFang | Abhishek, thank you that you had approved my review: show store info | 03:20 |
| *** udesale has joined #openstack-glance | 03:51 | |
| *** Sravan has joined #openstack-glance | 03:52 | |
| *** abhishekk has joined #openstack-glance | 04:25 | |
| openstackgerrit | Liang Fang proposed openstack/glance master: Fix for FK constraint violation https://review.openstack.org/617889 | 05:07 |
| *** ratailor has joined #openstack-glance | 05:22 | |
| *** pooja_jadhav has joined #openstack-glance | 05:25 | |
| openstackgerrit | Abhishek Kekane proposed openstack/glance master: Data remains in staging area if 'file' store is not enabled https://review.openstack.org/618468 | 05:35 |
| *** LiangFang has quit IRC | 06:01 | |
| *** Liang__ has joined #openstack-glance | 06:02 | |
| *** Sravan has quit IRC | 06:39 | |
| *** mosulica has joined #openstack-glance | 06:54 | |
| *** _alastor_ has joined #openstack-glance | 07:02 | |
| *** _alastor_ has quit IRC | 07:07 | |
| *** Sravan has joined #openstack-glance | 07:10 | |
| *** Sravan has quit IRC | 07:14 | |
| *** Luzi has joined #openstack-glance | 07:15 | |
| *** Sravan has joined #openstack-glance | 07:51 | |
| *** Sravan has quit IRC | 07:55 | |
| *** pcaruana has joined #openstack-glance | 08:39 | |
| *** udesale has quit IRC | 09:18 | |
| *** mosulica has quit IRC | 09:33 | |
| *** Liang__ has quit IRC | 09:43 | |
| *** openstackgerrit has quit IRC | 09:52 | |
| *** jmlowe has quit IRC | 09:53 | |
| *** w|zzy has quit IRC | 09:53 | |
| *** w|zzy has joined #openstack-glance | 09:54 | |
| *** mosulica has joined #openstack-glance | 10:00 | |
| *** udesale has joined #openstack-glance | 10:10 | |
| *** udesale has quit IRC | 10:21 | |
| *** abhishekk has quit IRC | 10:26 | |
| *** imacdonn has quit IRC | 11:01 | |
| *** imacdonn has joined #openstack-glance | 11:01 | |
| *** _alastor_ has joined #openstack-glance | 11:04 | |
| *** _alastor_ has quit IRC | 11:08 | |
| *** pooja_jadhav has quit IRC | 11:19 | |
| *** _alastor_ has joined #openstack-glance | 12:32 | |
| *** ratailor has quit IRC | 12:50 | |
| *** mordred has quit IRC | 12:55 | |
| *** rosmaita has left #openstack-glance | 13:26 | |
| *** rosmaita has joined #openstack-glance | 13:31 | |
| *** _alastor_ has quit IRC | 13:33 | |
| *** Sravan has joined #openstack-glance | 13:48 | |
| *** Sravan has quit IRC | 13:53 | |
| *** MattMan_ has quit IRC | 14:05 | |
| *** MattMan_ has joined #openstack-glance | 14:05 | |
| *** mosulica has quit IRC | 14:08 | |
| *** imacdonn has quit IRC | 14:52 | |
| *** Luzi has quit IRC | 14:56 | |
| *** _alastor_ has joined #openstack-glance | 14:57 | |
| *** ratailor has joined #openstack-glance | 15:02 | |
| *** jistr is now known as jistr|call | 15:29 | |
| *** ratailor has quit IRC | 15:39 | |
| *** pcaruana has quit IRC | 15:56 | |
| *** jistr|call is now known as jistr|biab | 15:59 | |
| *** gyee has joined #openstack-glance | 16:29 | |
| *** jistr|biab is now known as jistr | 16:43 | |
| *** imacdonn has joined #openstack-glance | 17:16 | |
| *** mordred has joined #openstack-glance | 17:21 | |
| *** Sravan has joined #openstack-glance | 17:33 | |
| *** Sravan has quit IRC | 17:34 | |
| *** Sravan has joined #openstack-glance | 17:34 | |
| imacdonn | rosmaita: ping? | 19:57 |
| *** Dinesh_Bhor has joined #openstack-glance | 20:00 | |
| rosmaita | imacdonn: pong | 20:08 |
| imacdonn | rosmaita: hi. I was having OpenStack-doc-rage .... I sortof figured it out, though | 20:08 |
| rosmaita | which docs? | 20:09 |
| imacdonn | I was trying to use the web-download import method, and failing to find anything that adequately describes how to do it | 20:09 |
| imacdonn | part of the problem was an old version of glanceclient that made my guesses fail ... but I really shouldn't have to guess | 20:09 |
| imacdonn | e.g. (I know this is "experimental"): https://docs.openstack.org/python-glanceclient/latest/cli/details.html#glance-image-create-via-import | 20:10 |
| imacdonn | there's nothing there that says you can add: --import-method web-download --uri <foo> | 20:10 |
| imacdonn | and I haven't found anything that describes the multi-step process (from a user perspective) at all, for web-download | 20:11 |
| rosmaita | yeah, the CLI docs don't get much love | 20:11 |
| imacdonn | very frustrating | 20:11 |
| rosmaita | well the api-ref has a pretty good description of what's supposed to happen | 20:11 |
| imacdonn | I'm being a user right now :P | 20:11 |
| rosmaita | i always think of "user" as "API user" | 20:12 |
| imacdonn | that may be a problem ... many users are just CLI users | 20:12 |
| *** Dinesh_Bhor has quit IRC | 20:12 | |
| rosmaita | anyway, i suppose you could adapt https://developer.openstack.org/api-ref/image/v2/index.html#interoperable-image-import to the CLI | 20:12 |
| imacdonn | (or horizon, even... but I'm thinking of typical sysadmins in a private cloud context) | 20:13 |
| rosmaita | yeah, you are probably right | 20:13 |
| imacdonn | the part I was getting frustrated about is "where do I put the image URI?" (when using the CLI) | 20:14 |
| rosmaita | i just noticed that we don't even have a link to the glanceclient CLI docs from the glance.oo.org landing page | 20:14 |
| imacdonn | I guess I know better than to look for a logical path to docs .. I just start with Google, cos I'll end up there anyway :) | 20:15 |
| imacdonn | so the reason I went on this little adventure .... I wanted to see how the "backend" appears in metadata, when using a store other than http | 20:17 |
| imacdonn | I'm not yet sure that using the default backend when adding a location is the right thing to do (although it will work OK for my use-case) | 20:17 |
| rosmaita | well, you are right that locations are weird | 20:18 |
| rosmaita | but in general, we don't want the user to have to worry about backend for the "normal" image operations | 20:18 |
| rosmaita | so in that sense, having the default is consistent with general expectations | 20:19 |
| imacdonn | I'm thinking of a possibly typical case of having, say, swift as the default store .. for anyone randomly creating an image and needing it stored somewhere .. but also having the http store available for published images | 20:19 |
| imacdonn | in that case, defaulting to swift when adding locations via patch isn't really ideal | 20:19 |
| rosmaita | i see what you mean | 20:19 |
| rosmaita | well, messing with locations should be restricted to operators anyway | 20:20 |
| rosmaita | so maybe the thing to do is make them be explicit about which store they want to use | 20:20 |
| rosmaita | after all, the admin will know all the store ids etc | 20:21 |
| imacdonn | yeah, I kinda sorta lean that way .... except it'll introduce incompatibility | 20:21 |
| rosmaita | yeah, but i think you have a good point | 20:22 |
| imacdonn | that's why I'm worrying about this now ... if it does go that way, I'll have to ask the developers of our provisioning system to add that metadata to their image creation proceess before we get to Stein | 20:22 |
| rosmaita | i mean, the whole idea of adding locations is to do something non-standard with glance's treatment of the backend | 20:22 |
| rosmaita | so in that way, it makes sense to require the backend be specified | 20:23 |
| imacdonn | except there's no (other) standard way to employ the http backend .... (?) | 20:23 |
| rosmaita | true, but the http backend is a pretty weird one | 20:25 |
| rosmaita | but i think the key thing here is, what makes for a good user experience? | 20:27 |
| rosmaita | thinking back to your swift example, i think if you try to set a http:// location for the swift backend, the call will fail because of an incorrect scheme for the store | 20:28 |
| imacdonn | right ... so is there any way to determine (in code) which schemes each backend supports ? | 20:30 |
| rosmaita | each store has to implement a get_schemes() method | 20:30 |
| rosmaita | but there are some other restrictions, e.g., can't set a file:// or swift+config:// (for security reasons) | 20:31 |
| imacdonn | pondering if there could be a way to iterate over the enabled backends and pick the (first?) one that supports the scheme presented? That'd probably be a "can or worms" | 20:32 |
| imacdonn | of* | 20:32 |
| imacdonn | or have a config option like ... default_backend_by_scheme | 20:33 |
| imacdonn | ("thinking out loud") | 20:33 |
| * rosmaita is liking forcing the user to specify which backend more and more | 20:34 | |
| rosmaita | i'm going to put a note on that bug so that erno, abhishek, and sean think about this issue | 20:34 |
| imacdonn | it does kinda seem like the right thing to do ... "if you're smart enough to be diddling with locations, you should be smart enough to know which backend you want" | 20:35 |
| rosmaita | exactly ... i think asking glance to be smart about this is simply asking for trouble! | 20:36 |
| imacdonn | I'm sorta shooting myself in the foot here, because it'll be incompatible with what we've implemented for Rocky (using patch to add a location - "we" = my "day job") ... but I think there's time to get that addressed before stein | 20:37 |
| imacdonn | the other thing I was wondering about ... haven't yet tested .... does the "backend" in metadata get used for anything after the image becomes active? If so, what happens to images that were created prior to enabling multi, which wouldn't have it? | 20:38 |
| rosmaita | that is a good question. I'm pretty sure if there's no backend specified, glance uses the default | 20:39 |
| rosmaita | abhishek did have a proposal to look through all backends with that scheme | 20:40 |
| rosmaita | but i can't remember for sure which one was implemented, i think *not* the search | 20:40 |
| imacdonn | if it uses the default, that could be a problem | 20:41 |
| rosmaita | you mean if the operator changes which backend is the current default? | 20:41 |
| imacdonn | the operator is going to be forced to enable multi in stein, right ? | 20:42 |
| rosmaita | i believe so | 20:42 |
| imacdonn | so say my configuration has always had swift as the default store, and http enabled | 20:43 |
| imacdonn | with multi, I'm forced to specify a "default_backend" | 20:43 |
| imacdonn | now all of the http images will default to using the swift backend ?!? | 20:44 |
| imacdonn | or vice-versa | 20:44 |
| rosmaita | yeah, that doesn't make sense. i'm pretty sure i'm wrong. glance currently uses the scheme of the location to determine wihch backend | 20:45 |
| rosmaita | that's why you couldn't have >1 of the same kind of backend before multistore in rocky | 20:45 |
| imacdonn | but isn't that "current" logic going away ? | 20:46 |
| rosmaita | i don't think so, because i don't think there's a migration for image locations | 20:47 |
| rosmaita | i need to go back and read the multistore spec, all this stuff was discussed in there | 20:47 |
| imacdonn | I guess I should do that too | 20:48 |
| rosmaita | https://specs.openstack.org/openstack/glance-specs/specs/rocky/implemented/glance/multi-store.html | 20:48 |
| imacdonn | "A very little hit on the performance of downloading the image from the old stores. For example, if existing user is using single rbd (say ceph) backend and now he has upgraded the environment and introduced two additional rbd stores as ceph1 and ceph2 with default store as ‘ceph1’ then if image which needs to be download from old store (ceph) will take some time as it needs to be looked in all the enabled backends." | 20:54 |
| rosmaita | ok, that makes sense | 20:55 |
| imacdonn | I wonder how it searches? like ... how would it determine if a given image "exists" in the http store ? | 21:04 |
| rosmaita | not sure. you may want to look at the tests | 21:07 |
| rosmaita | they probably don't test http store, but should give you an idea of whether there's a potential problem or not | 21:08 |
| rosmaita | but you know, since you are the biggest known http store user, you might want to add some tests just to protect yourself from regressions | 21:09 |
| imacdonn | that probably implies figuring out more about how the rest of glance works than I'd planned on | 21:18 |
| imacdonn | https://github.com/openstack/glance_store/blob/master/glance_store/multi_backend.py#L473-L489 | 21:28 |
| _alastor_ | rosmaita: Is there a way to either restrict the type of images that can be store on a backend or auto-convert them to the correct type on upload? | 21:39 |
| _alastor_ | For example, our backend can only quick-clone images into bootable volumes if they're stored in RAW format | 21:39 |
| _alastor_ | Currently I'm having users perform the conversions themselves, but ideally I'd like to have either Glance or my plugin do that conversion automagically | 21:40 |
| rosmaita | _alastor_: well, there is an image conversion plugin for the new image import process; or you can use the old glance tasks, which have a conversion task | 22:01 |
| _alastor_ | rosmaita: Are glance tasks due to be deprecated? | 22:21 |
| rosmaita | not sure what the plan is | 22:22 |
| rosmaita | if you think you might be interested, would be good to ask at the weekly glance meeting | 22:22 |
| _alastor_ | rosmaita: Ah, sorry. I still thought you were PTL | 22:23 |
| rosmaita | np, it's erno (jokke_) for rocky & stein | 22:24 |
| *** jmlowe has joined #openstack-glance | 23:46 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!