| *** gyee has quit IRC | 00:39 | |
| *** Liang__ has joined #openstack-glance | 00:40 | |
| *** jokke_ has quit IRC | 00:47 | |
| *** Liang__ is now known as LiangFang | 01:21 | |
| *** imacdonn has joined #openstack-glance | 01:38 | |
| *** imacdonn_ has quit IRC | 01:38 | |
| *** zhanglong has joined #openstack-glance | 01:58 | |
| *** udesale has joined #openstack-glance | 03:43 | |
| *** pdeore has joined #openstack-glance | 03:50 | |
| *** poojaj has joined #openstack-glance | 04:45 | |
| *** pdeore has quit IRC | 04:55 | |
| *** threestrands has joined #openstack-glance | 05:02 | |
| *** zul has quit IRC | 06:04 | |
| *** udesale has quit IRC | 06:16 | |
| *** udesale has joined #openstack-glance | 06:16 | |
| *** zhanglong has quit IRC | 07:18 | |
| *** mosulica has joined #openstack-glance | 07:37 | |
| *** mosulica has quit IRC | 07:58 | |
| *** lbragstad has joined #openstack-glance | 08:03 | |
| *** threestrands has quit IRC | 08:04 | |
| *** lbragstad has quit IRC | 08:06 | |
| *** mosulica has joined #openstack-glance | 08:07 | |
| *** lbragstad has joined #openstack-glance | 08:12 | |
| *** mosulica has joined #openstack-glance | 08:21 | |
| *** priteau has joined #openstack-glance | 08:35 | |
| *** itlinux has joined #openstack-glance | 08:49 | |
| *** itlinux has quit IRC | 08:50 | |
| *** lbragstad has quit IRC | 09:12 | |
| *** lbragstad has joined #openstack-glance | 09:24 | |
| *** mvkr has quit IRC | 09:29 | |
| *** brinzhang has joined #openstack-glance | 09:32 | |
| *** lbragstad has quit IRC | 09:39 | |
| *** priteau has quit IRC | 09:52 | |
| *** lbragstad has joined #openstack-glance | 09:54 | |
| *** mvkr has joined #openstack-glance | 09:55 | |
| *** mosulica has quit IRC | 09:56 | |
| *** pcaruana has joined #openstack-glance | 09:56 | |
| *** mosulica has joined #openstack-glance | 09:57 | |
| *** lbragstad has quit IRC | 10:04 | |
| *** mvkr has quit IRC | 10:38 | |
| *** mvkr has joined #openstack-glance | 10:38 | |
| *** LiangFang has quit IRC | 10:42 | |
| *** lbragstad has joined #openstack-glance | 10:44 | |
| *** poojaj is now known as poojajadhav | 11:09 | |
| *** poojajadhav is now known as poojaj | 11:10 | |
| *** poojaj has quit IRC | 11:26 | |
| *** lbragstad has quit IRC | 11:32 | |
| *** lbragstad has joined #openstack-glance | 11:40 | |
| *** lbragstad has quit IRC | 11:52 | |
| *** lbragstad has joined #openstack-glance | 11:53 | |
| *** lbragstad has quit IRC | 12:09 | |
| *** lbragstad has joined #openstack-glance | 12:12 | |
| *** lbragstad has quit IRC | 12:12 | |
| *** brinzhang has quit IRC | 13:43 | |
| *** Liang__ has joined #openstack-glance | 13:50 | |
| *** Liang__ is now known as LiangFang | 13:51 | |
| *** zul has joined #openstack-glance | 14:03 | |
| *** MattMan_ has quit IRC | 14:04 | |
| *** MattMan_ has joined #openstack-glance | 14:04 | |
| *** mosulica has quit IRC | 14:07 | |
| *** udesale has quit IRC | 14:16 | |
| *** Dinesh_Bhor has joined #openstack-glance | 14:22 | |
| *** udesale has joined #openstack-glance | 14:33 | |
| *** pvradu has joined #openstack-glance | 14:48 | |
| *** Dinesh_Bhor has quit IRC | 14:48 | |
| *** jistr is now known as jistr|mtg | 14:57 | |
| *** sapd1 has joined #openstack-glance | 15:19 | |
| openstackgerrit | zhouxinyong proposed openstack/glance master: Update the HTTP links to HTTPS in metadefs-index.rst. https://review.openstack.org/617688 | 15:23 |
|---|---|---|
| openstackgerrit | zhouxinyong proposed openstack/glance-specs master: Applying the HTTPS protocal in add-protected-filter.rst.To keep the website in this file be more robust,we'd better update the links to HTTPS type https://review.openstack.org/617690 | 15:26 |
| *** LiangFang has quit IRC | 15:29 | |
| *** Dinesh_Bhor has joined #openstack-glance | 15:29 | |
| openstackgerrit | zhouxinyong proposed openstack/glance_store master: Replacing the HTTP protocal with HTTPS in remove-s3-driver-f432afa1f53ecdf8.yaml. https://review.openstack.org/617692 | 15:31 |
| *** trident has quit IRC | 15:55 | |
| *** trident has joined #openstack-glance | 15:56 | |
| *** jistr|mtg is now known as jistr | 15:59 | |
| *** jmlowe has joined #openstack-glance | 15:59 | |
| *** gyee has joined #openstack-glance | 16:00 | |
| *** Luzi has joined #openstack-glance | 16:04 | |
| *** itlinux has joined #openstack-glance | 16:17 | |
| *** jmlowe has quit IRC | 16:22 | |
| *** Dinesh_Bhor has quit IRC | 16:23 | |
| *** itlinux has quit IRC | 16:32 | |
| *** jmlowe has joined #openstack-glance | 16:34 | |
| *** mvkr has quit IRC | 16:39 | |
| *** irclogbot_1 has joined #openstack-glance | 16:39 | |
| *** irclogbot_1 has quit IRC | 16:43 | |
| *** itlinux has joined #openstack-glance | 16:43 | |
| *** Luzi has quit IRC | 16:47 | |
| *** _alastor_ has joined #openstack-glance | 16:51 | |
| *** itlinux has quit IRC | 17:21 | |
| *** jmlowe has quit IRC | 17:25 | |
| *** udesale has quit IRC | 17:25 | |
| *** jmlowe has joined #openstack-glance | 17:28 | |
| *** jmlowe has quit IRC | 17:36 | |
| *** pvradu has quit IRC | 17:56 | |
| *** pvradu has joined #openstack-glance | 17:56 | |
| *** pvradu has quit IRC | 18:00 | |
| *** vishwanathj has joined #openstack-glance | 18:08 | |
| *** sapd1 has quit IRC | 18:11 | |
| imacdonn | rosmaita: are you around this week? not sure if you got to go to Berlin | 18:20 |
| rosmaita | i am sitting at home sulking | 18:20 |
| imacdonn | awww | 18:20 |
| rosmaita | what's up? | 18:20 |
| imacdonn | https://bugs.launchpad.net/glance/+bug/1802587 | 18:20 |
| openstack | Launchpad bug 1802587 in Glance "With multiple backends enabled, adding a location does not default to default store" [Undecided,New] | 18:20 |
| imacdonn | still trying to understand how this has anything to do with the validation_data stuff .. other than that they both involve setting locations via patch | 18:21 |
| rosmaita | well, when i have multiple backends set up (assuming i did it correctly), and i use your glanceclient patch without setting a specific backend, the call fails | 18:22 |
| imacdonn | so my question (https://review.openstack.org/602794) ... does it work (adding a location without specifying a backend) WITHOUT my patch ? | 18:23 |
| rosmaita | depends on whether you have multiple backends enabled or not | 18:23 |
| vishwanathj | hi, looking for links that explains how glance copies large size images (>4 GB) over to compute nodes .... have ansible kolla openstack installation and have bare metal compute nodes; noticing that the first time a VM is created using the image, it takes a long time to create the VM or fails to create VM due to timeout | 18:23 |
| imacdonn | right | 18:23 |
| imacdonn | my point is that it has nothing to do with my patch .. the problem already existed, and is completely separate ... I think ? | 18:24 |
| vishwanathj | or any links that explains how to handle my scenario is appreciated as well | 18:24 |
| rosmaita | imacdonn: yes, you have not broken anything, just revealed a problem | 18:25 |
| rosmaita | vishwanathj: glance doesn't do the copy ... nova requests the image data from glance via http | 18:26 |
| imacdonn | rosmaita: yeah, so I'm thinking that this shouldn't block the addition of validation_data to glanceclient ... unless you think wee need to do something radically different | 18:26 |
| vishwanathj | rosmaita oh I see | 18:26 |
| rosmaita | imacdonn: well, i'm not sure, just don't want to introduce new functionality that's going to fail a lot ... we need abhishek and erno's opinions, they've been driving the multiple backends | 18:27 |
| rosmaita | imacdonn: abhishek should be back tomorrow, he knows about the bug | 18:27 |
| rosmaita | vishwanathj: what backend are you using for glance? | 18:28 |
| vishwanathj | we are using IBM Swift as the glance backend...we have openstack installed in IBM Cloud | 18:29 |
| rosmaita | imacdonn: so if your question is, "what changes do i need to make to my patch to get it approved?", the answer from me is, "i don't know" | 18:29 |
| imacdonn | rosmaita: FWIW, I tried to enable multi-backend in Rocky, prompted by a deprecation warning, and was not able to get it to work, then I read something that said it's experimental and not recommended yet ... IMO, deprecation is been abused a bit these days, but that's s separate issue too | 18:30 |
| rosmaita | imacdonn: i didn't try it in stable/rocky, but did get it working in S-1 | 18:31 |
| imacdonn | rosmaita: I need to try it again, as I don't remember what failed | 18:32 |
| rosmaita | imacdonn: i have a paste of my glance-api.conf for multiple backends, gimme a minute to find it | 18:33 |
| vishwanathj | rosmaita using IBM swift as the glance backend, Openstack is installed in IBM cloud | 18:33 |
| rosmaita | imacdonn: http://paste.openstack.org/show/734694/ | 18:33 |
| imacdonn | rosmaita: I think I was attempting to configure it for multi, but with only a single backend ... i.e. I don't actually need multiple (I only every use http), but using the new config options | 18:33 |
| rosmaita | imacdonn: that's also why i want a second opinion about the bug, not sure i have everything configured properly | 18:34 |
| rosmaita | vishwanathj: i don't have any suggestions | 18:35 |
| vishwanathj | rosmaita ok, thanks | 18:35 |
| rosmaita | vishwanathj: when nova requests the image from glance, glance will get it from swift and cache it as it streams it to nova; the second request for the same image (after the caching is complete) should be much faster | 18:36 |
| rosmaita | vishwanathj: that doesn't speed up the first download, though | 18:36 |
| imacdonn | there may be some timeout that can be tweaked on the nova side? I don't recall offhand | 18:37 |
| vishwanathj | rosmaita oh ok, so it is expected that the first download would be slow and subsequent ones would be faster because it would be cached in the compute node...looks like every new compute node added will have this issue, right/ | 18:37 |
| imacdonn | every compute node would have to cache the image the first time it's needed there | 18:38 |
| *** gyee has quit IRC | 18:38 | |
| rosmaita | vishwanathj: yes, and like imacdonn points out, the compute also has a cache, but that's usually limited to only base images | 18:38 |
| imacdonn | a possible alternative approach is to create a volume from the image, then snapshot that volume .. then when you create a new VM, clone the snapshot to make a boot volume | 18:39 |
| imacdonn | all of that should happen on the Ceph side, and "be fast" | 18:40 |
| imacdonn | oh wait, you said Swift, not Ceph | 18:40 |
| imacdonn | nevermind ;) | 18:40 |
| vishwanathj | imacdonn so the workaround exists for ceph backend but not Swift? | 18:41 |
| rosmaita | vishwanathj: the image cache is middleware, to see if it's enabled check the paste_deploy.flavor config option in glance-api.conf | 18:41 |
| rosmaita | https://github.com/openstack/glance/blob/master/etc/glance-api.conf#L4413 | 18:41 |
| imacdonn | ... but if you do have a cinder service with a backend that supports cloning, it is a way to make booting from large images much more expedient | 18:42 |
| vishwanathj | I see | 18:43 |
| imacdonn | rosmaita: I enabled multi again, and it seems to start up OK .. the more I think about it, it seems likely I ran into the same issue you did | 18:58 |
| rosmaita | could be ... i get an "invalid location" error, which isn't very helpful | 18:59 |
| imacdonn | rosmaita: this reminds me of another issue ... the show_multiple_locations config option is supposedly deprecated since Newton, but.... https://github.com/openstack/glance/blob/master/glance/api/v2/images.py#L452-L455 | 18:59 |
| rosmaita | yeah, it turns out to be remarkably complicated to get rid of that option | 19:00 |
| rosmaita | imacdonn: https://docs.openstack.org/releasenotes/glance/ocata.html#relnotes-14-0-0-origin-stable-ocata-other-notes | 19:02 |
| imacdonn | yeah | 19:04 |
| imacdonn | the deprecation message still kindof lies, though | 19:04 |
| imacdonn | it says "the same functionality can be achieved with greater granularity by using policies.", which is not quite true | 19:05 |
| rosmaita | well, it's more an aspiration than a statement of fact, it turns out | 19:06 |
| imacdonn | heh | 19:07 |
| imacdonn | question .... is there any backend other than http that actually supports setting locations ? | 19:08 |
| imacdonn | trying to think of any other case where it'd actually make sense | 19:08 |
| rosmaita | well, locations are a glance thing, really, not a backend thing | 19:08 |
| *** irclogbot_1 has joined #openstack-glance | 19:09 | |
| rosmaita | i think when they were introduced, the idea was you could have an image stored in multiple places | 19:09 |
| rosmaita | you would use a location_strategy module to pick the location that was used for a particular request | 19:09 |
| imacdonn | I should have said "store", not "backend", perhaps | 19:10 |
| rosmaita | but i think you can configure nova to use locations to do the fast-snapshotting thing for images with the cepth backend | 19:11 |
| rosmaita | there's also the possibility to pre-deploy image data directly to the backend and then set the location | 19:12 |
| rosmaita | basically, what you're doing | 19:13 |
| rosmaita | except with the other stores | 19:13 |
| imacdonn | ok, hmmm | 19:13 |
| vishwanathj | rosmaita imacdonn what do you think about following the instructions at https://docs.openstack.org/glance/latest/admin/cache.html ... would it help alleviate a bit? I know it talks about caching on the server where glance-api is running and not on compute node | 19:13 |
| imacdonn | vishwanathj: it might ... it probably depends on where your bottleneck is | 19:14 |
| rosmaita | vishwanathj: if you use the prefetcher, it should speed things up | 19:15 |
| rosmaita | but imacdonn is right | 19:15 |
| rosmaita | you need to make sure a cached image is actually faster | 19:15 |
| vishwanathj | ok, I am using kolla ansible for deploying my openstack controller, all the services are running inside of a docker container, are you aware the cache will be supported on docker containers | 19:17 |
| rosmaita | vishwanathj: i really don't know. i think the software will work, but there may be resource constraints on how big your cache can be | 19:19 |
| vishwanathj | ok | 19:20 |
| _alastor_ | rosmaita: Hey, I'm trying to write some compatibility code in a driver for the Store.add function which changed signature in Dec 2017 | 19:58 |
| _alastor_ | rosmaita: The issue I'm hitting is with the capabilities.check decorator, it's not correctly passing the store as the first argument | 19:58 |
| _alastor_ | Example: http://paste.openstack.org/show/734785/ | 19:58 |
| rosmaita | _alastor_: looking | 20:00 |
| _alastor_ | rosmaita: This is the exception. http://paste.openstack.org/show/734786/ | 20:01 |
| _alastor_ | rosmaita: It occurs because the first argument passed to 'add' is the image_id instead of the store object | 20:01 |
| _alastor_ | rosmaita: I'm running on Pike, but it's pretty much the same on master | 20:02 |
| *** irclogbot_1 has quit IRC | 20:09 | |
| *** irclogbot_1 has joined #openstack-glance | 20:11 | |
| *** vishwanathj has quit IRC | 20:19 | |
| *** jmlowe has joined #openstack-glance | 20:20 | |
| rosmaita | _alastor_: i think i need more context for http://paste.openstack.org/show/734785/ | 20:30 |
| _alastor_ | rosmaita: I'm doing a simple import check to see if driver.back_compat_add exists. If it doesn't, then I need to overwrite the Store.add function with the older version | 20:32 |
| _alastor_ | rosmaita: Essentially I'm creating an adapter that will change the old function signature to the new one so I only have to maintain one version | 20:32 |
| rosmaita | _alastor_: not sure i understand ... why don't you just write the new function sig and use driver.back_compat_add to handle an "old style" call? | 20:35 |
| _alastor_ | rosmaita: Because that import won't function in a stable/pike environment because back_compat_add doesn't exist | 20:36 |
| rosmaita | oh, ok | 20:36 |
| rosmaita | do you have the capabilities.check decorator also on your definition of the store.add() function? | 20:46 |
| _alastor_ | rosmaita: Yes, that functions just fine. I add a passthrough back_compat_add function to the driver module when I detect it isn't present. The issue I'm running into is trying to correctly overwrite the 'add' function in that scenario | 20:50 |
| _alastor_ | When I use types.MethodType to overwrite I get an error like this: http://paste.openstack.org/show/734787/ | 20:50 |
| _alastor_ | Example using types.MethodType: http://paste.openstack.org/show/734788/ | 20:52 |
| rosmaita | thanks | 20:52 |
| *** mvkr has joined #openstack-glance | 21:02 | |
| *** pvradu has joined #openstack-glance | 21:04 | |
| rosmaita | _alastor_: i am not sure what's going on, but if you are already doing the capabilites check on add() and _add() is calling add(), then i wouldn't think you need to decorate _add()? what happens if you just leave the decorator off? | 21:09 |
| _alastor_ | rosmaita: When I take the decorator off I get this: http://paste.openstack.org/show/734789/ | 21:25 |
| rosmaita | looking | 21:25 |
| _alastor_ | rosmaita: I think it means the wrong number of arguments are getting passed to the function still | 21:26 |
| rosmaita | that looks just like the error when you used types.MethodType | 21:27 |
| rosmaita | not sure what that means, though | 21:28 |
| _alastor_ | rosmaita: This is where everything goes wrong. I've poked around with PDB and I can't figure out why the call fails: http://paste.openstack.org/show/734790/ | 21:33 |
| _alastor_ | rosmaita: You can see the argspec of the function should work with the arguments provided | 21:34 |
| rosmaita | if you are still in the debugger, what does it give you for the argspec for store._add ? | 21:36 |
| _alastor_ | rosmaita: can't call inspect on it because it's a nested method. I'll try promoting it and see if that helps | 21:39 |
| rosmaita | _alastor_: i have to go afk for a few hours, good luck | 21:43 |
| *** jmlowe has quit IRC | 21:43 | |
| _alastor_ | rosmaita: alrighty, thanks | 21:44 |
| *** pcaruana has quit IRC | 21:56 | |
| *** pvradu has quit IRC | 21:59 | |
| *** Sravan has joined #openstack-glance | 22:06 | |
| *** vishwanathj has joined #openstack-glance | 22:07 | |
| *** jiaopengju has quit IRC | 22:08 | |
| *** jiaopengju has joined #openstack-glance | 22:08 | |
| *** hoonetorg has quit IRC | 22:09 | |
| *** kukacz has quit IRC | 22:09 | |
| *** hoonetorg has joined #openstack-glance | 22:10 | |
| *** kukacz has joined #openstack-glance | 22:10 | |
| *** Sravan has quit IRC | 22:52 | |
| *** gyee has joined #openstack-glance | 23:01 | |
| *** pvradu has joined #openstack-glance | 23:14 | |
| *** pvradu has quit IRC | 23:19 | |
| *** Sravan has joined #openstack-glance | 23:27 | |
| *** Sravan has quit IRC | 23:32 | |
| *** Sravan has joined #openstack-glance | 23:38 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!