| opendevreview | OpenStack Proposal Bot proposed openstack/cinder master: Imported Translations from Zanata https://review.opendev.org/c/openstack/cinder/+/846842 | 02:16 |
|---|---|---|
| *** hemna9 is now known as hemna | 06:36 | |
| opendevreview | Tushar Trambak Gite proposed openstack/cinder master: Delete volume and cloned-volume concurrently https://review.opendev.org/c/openstack/cinder/+/843309 | 09:17 |
| opendevreview | Tushar Trambak Gite proposed openstack/cinder master: Configure a storage node in cinder https://review.opendev.org/c/openstack/cinder/+/820994 | 09:32 |
| opendevreview | Tushar Trambak Gite proposed openstack/cinder master: Update volume delete rejection message https://review.opendev.org/c/openstack/cinder/+/826608 | 09:40 |
| opendevreview | Tushar Trambak Gite proposed openstack/cinder master: Storage node configuration https://review.opendev.org/c/openstack/cinder/+/822822 | 09:55 |
| opendevreview | Luigi Toscano proposed openstack/cinder stable/yoga: SF: Remove compat clone image code https://review.opendev.org/c/openstack/cinder/+/848248 | 10:07 |
| opendevreview | Luigi Toscano proposed openstack/cinder stable/xena: SF: Remove compat clone image code https://review.opendev.org/c/openstack/cinder/+/848249 | 10:07 |
| opendevreview | Luigi Toscano proposed openstack/cinder stable/wallaby: SF: Remove compat clone image code https://review.opendev.org/c/openstack/cinder/+/848250 | 10:08 |
| tosky | whoami-rajat: I took the liberty of backporting some patches ^^ as the change seemed to apply cleanly | 10:08 |
| whoami-rajat | tosky, thanks, since the feature was removed in train, removing the compat code in wallaby (and above) seems safe (3 release gap) | 10:12 |
| *** bhagyashris_ is now known as bhagyashris | 10:25 | |
| opendevreview | Alexander Malashenko proposed openstack/cinder master: WIP: Add Cinder NFS driver for Dell EMC PowerStore https://review.opendev.org/c/openstack/cinder/+/797608 | 10:29 |
| opendevreview | Luigi Toscano proposed openstack/cinder-tempest-plugin master: DNM test ceph on CentOS Stream 9 https://review.opendev.org/c/openstack/cinder-tempest-plugin/+/729766 | 10:34 |
| opendevreview | Tushar Trambak Gite proposed openstack/cinder master: Include volume type constraints in internal API https://review.opendev.org/c/openstack/cinder/+/846146 | 10:35 |
| opendevreview | Rafael Weingartner proposed openstack/cinder master: Filter reserved image properties https://review.opendev.org/c/openstack/cinder/+/812685 | 10:59 |
| opendevreview | Tushar Trambak Gite proposed openstack/cinder master: Update volume_type_project bogus column https://review.opendev.org/c/openstack/cinder/+/822821 | 11:00 |
| *** dviroel_ is now known as dviroel | 11:20 | |
| amalashenko | Hello there! We can't have our zuul gate passed because of errors which seem not to be related to our code. https://review.opendev.org/c/openstack/cinder/+/807215. What should we do? | 11:23 |
| opendevreview | Merged openstack/devstack-plugin-ceph master: Deploy with cephadm https://review.opendev.org/c/openstack/devstack-plugin-ceph/+/826484 | 12:14 |
| felipe_rodrigues | .kl;.m,kmnb m | 12:25 |
| felipe_rodrigues | ,.lkiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii,mmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmkccfguiop[] | 12:25 |
| felipe_rodrigues | .lokihzzs by 678[].jm,lki,jklo;p[' | 12:26 |
| felipe_rodrigues | o9p;' | 12:27 |
| felipe_rodrigues | kiloçp~´´´´´´´´p0ol.kiklop | 12:27 |
| felipe_rodrigues | [=´~ | 12:27 |
| felipe_rodrigues | .=[=´oiool ç^ç´p0o9m,0-~ | 12:27 |
| tosky | say hello to the cat | 12:36 |
| felipe_rodrigues | sorry, kids | 12:36 |
| enriquetaso | hahaha | 13:00 |
| dc_arun | Hello core reviewers, https://review.opendev.org/c/openstack/cinder/+/836996 | 13:59 |
| hemna | geguileo, you around? | 14:17 |
| geguileo | hemna: hi, yeah, but I'm in a meeting | 14:21 |
| hemna | ok just ping me when you are done | 14:22 |
| hemna | thanks | 14:22 |
| geguileo | ok | 14:22 |
| amalashenko | Hello there! We can't have our zuul gate passed because of errors which seem not to be related to our code. https://review.opendev.org/c/openstack/cinder/+/807215. What should we do? | 14:38 |
| opendevreview | Rafael Weingartner proposed openstack/cinder master: Filter reserved image properties https://review.opendev.org/c/openstack/cinder/+/812685 | 14:40 |
| opendevreview | Alexander Malashenko proposed openstack/cinder master: WIP: Add Cinder NFS driver for Dell EMC PowerStore https://review.opendev.org/c/openstack/cinder/+/797608 | 14:46 |
| opendevreview | Fábio Oliveira proposed openstack/cinder master: NetApp ONTAP: Implemented REST Client for ONTAP https://review.opendev.org/c/openstack/cinder/+/836027 | 15:17 |
| geguileo | hemna: I'm back | 15:30 |
| hemna | yough | 15:31 |
| hemna | so i'm poking around and noticed that a few drivers report backend_state = 'up'/'down' | 15:31 |
| hemna | in their capabilities. I can't find where that is applied and used | 15:31 |
| hemna | does that somehow make it in here: https://github.com/openstack/cinder/blob/89eaf24c6a5f91e94902ba60d99a05c741e5af6b/cinder/scheduler/host_manager.py#L694 ? | 15:32 |
| hemna | from what I can tell that answer is now, because https://github.com/openstack/cinder/blob/master/cinder/objects/service.py#L206-L209 | 15:32 |
| hemna | err s/now/no | 15:32 |
| hemna | I want to be able to disable a pool from being scheduled, but not removed from the scheduler itself and reporting via pool stats | 15:33 |
| hemna | kinda surprised that doesn't exist already. like report 'pool_state': 'up'/'down' and if it's 'down', then the filters will filter that pool out until it's 'up' | 15:34 |
| geguileo | hemna: afaik it doesn't exist yet | 15:35 |
| geguileo | iirc they are only disabled if there is no heartbeat | 15:35 |
| hemna | the only way I see to do it now is to set an extra spec in every volume type of pool_state: up | 15:36 |
| hemna | several drivers set backend_state: up | 15:36 |
| hemna | lvm for one | 15:36 |
| hemna | 3par as well, rbd, few more. | 15:36 |
| geguileo | but is that actually been properly updated? | 15:36 |
| hemna | 3par does fwiw | 15:36 |
| hemna | it can go from down to up | 15:36 |
| geguileo | that's cool! | 15:37 |
| hemna | I kinda think we need both backend_state and pool_state | 15:37 |
| geguileo | LVM is always up | 15:37 |
| geguileo | I think it's tricky, right? Because we also need to report it right in cinder service-list | 15:37 |
| hemna | we have pools that go into maintenance mode for a bit and will fail to create a volume while in that state. so I wanted to mark that pool as 'down' until its out of maintenance mode | 15:37 |
| geguileo | So it's not only about making the scheduler do the right thing, but also Cinder report things right to end users | 15:38 |
| hemna | for pools I think it's easier | 15:38 |
| hemna | pool_state is down, then filter out always. | 15:38 |
| hemna | until it's back up | 15:38 |
| hemna | but if a driver reports backend_state = down, then the service is kinda down | 15:39 |
| hemna | well the backend is down, but the service is up, if that makes sense | 15:39 |
| hemna | the scheduler should filter out that entire backend at that point, re: all pools | 15:40 |
| geguileo | oooooh, cinder service-list already reports this information! | 15:40 |
| hemna | if the driver reports backend_state = 'down' ? | 15:40 |
| geguileo | yes | 15:40 |
| hemna | ok cool, what does the scheduler do with it? | 15:40 |
| geguileo | hemna: https://opendev.org/openstack/cinder/src/commit/89eaf24c6a5f91e94902ba60d99a05c741e5af6b/cinder/api/contrib/services.py#L75-L84 | 15:41 |
| geguileo | I'll look now | 15:41 |
| hemna | ok, I figured the BackendState object would have an attribute of backend_state or state | 15:42 |
| hemna | but it doesn't | 15:42 |
| geguileo | looks like the scheduler ignores this info :-( | 15:43 |
| hemna | https://github.com/openstack/cinder/blob/master/cinder/scheduler/host_manager.py#L96 | 15:43 |
| hemna | yah that's bad | 15:43 |
| hemna | most drivers don't report it | 15:44 |
| hemna | so I presume the default should be backend_state = 'up' | 15:44 |
| hemna | if I'm not mistaken, for a short term fix I could just specify pool_state = 'up' in the volume types, and then have the driver report the pool_state, and that should filter out pools that are down? | 15:46 |
| hemna | via the capabilities filter. But I think cinder in general should handle that w/o the volume type extra specs hack | 15:46 |
| geguileo | yeah, we need to fix that whole thing | 15:46 |
| geguileo | and if all pools are down the backend should report as down | 15:46 |
| hemna | true | 15:47 |
| geguileo | and we should have a field to report a reason why it's down | 15:47 |
| geguileo | the tricky part is that in a cluster we could have a host that returns that it's down while the others say they are ok | 15:47 |
| geguileo | so we will also have to consider that active-active case | 15:48 |
| hemna | yah that's a confusing state | 15:48 |
| hemna | I'd say we'd have to side on caution | 15:48 |
| geguileo | well, it could be that a specific service has lost network connection to the backend | 15:48 |
| hemna | but maybe one c-vol has a fuct network and can't talk to the backend....so that c-vol shouldn't accept requests, but the others are ok | 15:49 |
| geguileo | it we were to allow drivers to asynchronously report down state without waiting for stats | 15:49 |
| geguileo | that would help | 15:49 |
| hemna | heh | 15:49 |
| geguileo | this is definitely a nice feature to work on | 15:49 |
| hemna | so what about pool_state | 15:49 |
| geguileo | the pool state would be the same in active-active | 15:50 |
| hemna | the same thing could happen I suppose, due to having 'old' reporting from a few c-vols, but the most recent has a diffferent value | 15:50 |
| geguileo | I confirmed that creating a volume type with backend_state=up as extra specs works fine | 15:53 |
| hemna | ok perfect. I'll do that short term | 15:53 |
| geguileo | I haven't tried with the pool_state | 15:54 |
| geguileo | but it should be the same thing, since backend_state is reported at the pool level | 15:55 |
| hemna | yah I think the pools capabilities are compared when pools are reported | 15:57 |
| geguileo | hemna: another thing that should be improved is when c-sch stops receiving stats from c-vol | 15:58 |
| geguileo | right now that is ignored as long as we have some values | 15:58 |
| geguileo | and there is a heartbeat on the DB | 15:58 |
| hemna | doh yah | 15:58 |
| NiklasNorin[m] | Hi guys!! I'm new to openstack and I'm setting up my first all-in-one installation (on a decent home server). I have 2x RAID5/6 arrays, one with only SSDs and one with only HDDs.... (full message at https://matrix.org/_matrix/media/r0/download/matrix.org/mQnVdChjLHLHQNIhKfryKWhL) | 20:40 |
| *** dviroel is now known as dviroel|afk | 20:42 | |
| opendevreview | Merged openstack/cinder master: PowerMax Driver - allow for None values in metadata https://review.opendev.org/c/openstack/cinder/+/803785 | 21:11 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!