| *** akekane_ is now known as abhishekk | 07:15 | |
| *** akekane_ is now known as abhishekk | 08:57 | |
| *** dviroel|out is now known as dviroel | 11:02 | |
| *** mabrams1 is now known as mabrams | 11:09 | |
| opendevreview | Brian Rosmaita proposed openstack/python-cinderclient master: Add QuotaSet class to cinderclient.v3 https://review.opendev.org/c/openstack/python-cinderclient/+/800945 | 13:18 |
|---|---|---|
| whoami-rajat | geguileo, hey, can you take a look at this since you had a previous +2 on it? https://review.opendev.org/c/openstack/python-cinderclient/+/783628 | 13:45 |
| opendevreview | Brian Rosmaita proposed openstack/python-cinderclient master: Add missing classes to cinderclient.v3 https://review.opendev.org/c/openstack/python-cinderclient/+/800945 | 14:17 |
| opendevreview | Eric Harney proposed openstack/cinder master: mypy: service.py https://review.opendev.org/c/openstack/cinder/+/788367 | 14:19 |
| opendevreview | Eric Harney proposed openstack/cinder master: Add installation of mypy stubs packages https://review.opendev.org/c/openstack/cinder/+/800306 | 14:33 |
| opendevreview | Eric Harney proposed openstack/cinder master: Validate fields in encryption spec create/update https://review.opendev.org/c/openstack/cinder/+/789603 | 15:01 |
| Mr_Freezeex | geguileo: Hi! If you want to continue the discussion we had yesterday about snapshot mirroring in the backup driver and about your concerns that the backup driver will have to worry about managing snapshots. Well it should be mostly fine because ceph itself manage a max number of mirror snapshots and will remove snapshots as they are replayed on the other cluster. So this is | 15:02 |
| Mr_Freezeex | only about calling `mirror_image_create_snapshot` on the image after data transfer to improve the load on ceph and the RPO of the replication. There would not be any more logic on the cinder side than checking if the snapshot mirroring option is enabled basically. | 15:02 |
| geguileo | Mr_Freezeex: what I was referring to, is that if we have the scheduler anabled on the cluster, why would we need to enable the feature in cinder? won't the scheduler create the first snapshot and then not create more than 3? | 15:06 |
| Mr_Freezeex | geguileo: the max snapshot is a setting handled through librbd. Basically this will be pretty similar than what the scheduler does but every x minutes. This will allow to not enable the scheduler on the backup pool and give a "hint" to ceph on when it matters to replcate the data. | 15:10 |
| geguileo | Mr_Freezeex: that's the thing, if it can be done on the backend, we prefer that | 15:11 |
| Mr_Freezeex | geguileo: even if it's a < 10 line patch that would improve both perfomance and RPO :(? | 15:12 |
| Mr_Freezeex | (to be fair I will pass if you really say no, my goal is to make this possible first but this little imprvement would ease the load on our cluster) | 15:15 |
| geguileo | Mr_Freezeex: how would it easy it vs using the scheduler? | 15:16 |
| * geguileo is not understanding something... | 15:16 | |
| geguileo | s/easy/ease | 15:16 |
| Mr_Freezeex | geguileo: the scheduler will create many unecessary snapshots even if data didn't change. So then the rbd-mirror daemon will have to pickup all those snapshots try to diff with the previous one etc... In our case one daemon was able to handle 100 empty snapshots per minute. | 15:20 |
| geguileo | Mr_Freezeex: I'm sold | 15:20 |
| geguileo | the feature makes sense | 15:21 |
| geguileo | Mr_Freezeex: thanks for the explanation | 15:21 |
| geguileo | I would make it very clear that is the reason for adding the feature | 15:21 |
| geguileo | in the commit message, docs, etc | 15:22 |
| Mr_Freezeex | geguileo: thanks for the approval ahah :D | 15:22 |
| geguileo | oh, thank you for taking the time to explain the underlying issues that led to this driver improvement | 15:23 |
| Mr_Freezeex | geguileo: Actually i don't think there is much documentation for replication in the ceph cinder-backup | 15:23 |
| Mr_Freezeex | geguileo: sure no worries | 15:23 |
| geguileo | because up till now it was only configured in the ceph cluster | 15:23 |
| geguileo | and that's how we wanted to keep it | 15:23 |
| Mr_Freezeex | geguileo: Yes but well actually there is configuration difference between cinder and cinder-backup even on the pool | 15:24 |
| geguileo | but considering there is a considerable performance impact on the ceph cluster if done in the cluster, it makes sense to add it | 15:24 |
| Mr_Freezeex | geguileo: So that's a bit misleading imo | 15:24 |
| Mr_Freezeex | geguileo: I would happily add some docs for journaling and snapshot though :) | 15:25 |
| geguileo | what do you mean there is config diff between cinder and backup? | 15:25 |
| geguileo | Mr_Freezeex: docs are ALWAYS welcomed!!! | 15:25 |
| Mr_Freezeex | geguileo: In cinder you should activate the per image mode on the pool, on the backup side you have to enable it on the whole pool | 15:25 |
| geguileo | Mr_Freezeex: because one is the Cinder replication feature, and in the backup Cinder doesn't care | 15:26 |
| geguileo | so the difference is the level of integration | 15:26 |
| geguileo | in most cases people don't backup their backups | 15:26 |
| geguileo | so we leave it to the storage admins to decide how to implement that part | 15:27 |
| geguileo | though having docs would be great | 15:27 |
| Mr_Freezeex | geguileo: yep I will try add that soon :D | 15:28 |
| geguileo | thanks!! | 15:28 |
| opendevreview | Merged openstack/os-brick master: Drop lower-constraints jobs https://review.opendev.org/c/openstack/os-brick/+/789079 | 17:10 |
| amar7ibm | Hi team, if anyone has time could they please look at this fix : https://review.opendev.org/c/openstack/cinder/+/782405 and share your reviews/comments. | 17:12 |
| opendevreview | Merged openstack/cinder master: Add installation of mypy stubs packages https://review.opendev.org/c/openstack/cinder/+/800306 | 17:22 |
| *** amar7ibm is now known as amar__ | 17:32 | |
| opendevreview | Eric Harney proposed openstack/cinder master: mypy: create_volume flows https://review.opendev.org/c/openstack/cinder/+/780933 | 17:33 |
| opendevreview | Eric Harney proposed openstack/cinder master: mypy: image cache https://review.opendev.org/c/openstack/cinder/+/788363 | 17:35 |
| opendevreview | Eric Harney proposed openstack/cinder master: mypy: continued manager, scheduler, rpcapi https://review.opendev.org/c/openstack/cinder/+/784453 | 17:36 |
| opendevreview | Brian Rosmaita proposed openstack/cinder master: Run cinder-mypy job in the check https://review.opendev.org/c/openstack/cinder/+/792642 | 18:04 |
| opendevreview | Brian Rosmaita proposed openstack/os-brick master: Update mypy infrastructure https://review.opendev.org/c/openstack/os-brick/+/800568 | 18:09 |
| opendevreview | Andrei proposed openstack/cinder master: Open-E JovianDSS: Addes multi-attach support https://review.opendev.org/c/openstack/cinder/+/794962 | 18:38 |
| opendevreview | Andrei proposed openstack/cinder master: Open-E JovianDSS: Addes multi-attach support https://review.opendev.org/c/openstack/cinder/+/794962 | 18:43 |
| opendevreview | Eric Harney proposed openstack/cinder master: mypy: coordination.py https://review.opendev.org/c/openstack/cinder/+/788364 | 19:29 |
| *** dviroel is now known as dviroel|brb | 19:53 | |
| opendevreview | Eric Harney proposed openstack/os-brick master: mypy: initiator https://review.opendev.org/c/openstack/os-brick/+/786732 | 19:56 |
| *** dviroel|brb is now known as dviroel | 20:27 | |
| *** dviroel is now known as dviroel|out | 21:24 | |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!