| *** felipemonteiro__ has quit IRC | 00:53 | |
| *** iyamahat has quit IRC | 01:02 | |
| *** yamahata has quit IRC | 01:03 | |
| *** edmondsw has joined #openstack-meeting-cp | 01:21 | |
| *** edmondsw has quit IRC | 01:26 | |
| *** kmARC_ has joined #openstack-meeting-cp | 02:04 | |
| *** margaret_ has joined #openstack-meeting-cp | 02:04 | |
| *** hrybacki_ has joined #openstack-meeting-cp | 02:04 | |
| *** harlowja has quit IRC | 02:07 | |
| *** margaret has quit IRC | 02:07 | |
| *** hrybacki has quit IRC | 02:07 | |
| *** dhellmann has quit IRC | 02:07 | |
| *** dansmith has quit IRC | 02:07 | |
| *** kmARC has quit IRC | 02:07 | |
| *** margaret_ is now known as margaret | 02:07 | |
| *** hrybacki_ is now known as hrybacki | 02:07 | |
| *** kmARC_ is now known as kmARC | 02:07 | |
| *** dansmith has joined #openstack-meeting-cp | 02:08 | |
| *** dansmith is now known as Guest99680 | 02:08 | |
| *** dhellmann has joined #openstack-meeting-cp | 02:08 | |
| *** aselius has quit IRC | 02:36 | |
| *** _pewp_ has quit IRC | 02:46 | |
| *** _pewp_ has joined #openstack-meeting-cp | 02:46 | |
| *** yamahata has joined #openstack-meeting-cp | 03:03 | |
| *** edmondsw has joined #openstack-meeting-cp | 03:09 | |
| *** edmondsw has quit IRC | 03:14 | |
| *** iyamahat has joined #openstack-meeting-cp | 03:20 | |
| *** coolsvap has joined #openstack-meeting-cp | 03:54 | |
| *** edmondsw has joined #openstack-meeting-cp | 04:57 | |
| *** edmondsw has quit IRC | 05:02 | |
| *** margaret has quit IRC | 06:26 | |
| *** robcresswell has quit IRC | 06:26 | |
| *** ameade has quit IRC | 06:26 | |
| *** TheJulia has quit IRC | 06:26 | |
| *** kencjohnston has quit IRC | 06:26 | |
| *** persia has quit IRC | 06:26 | |
| *** knikolla has quit IRC | 06:26 | |
| *** margaret has joined #openstack-meeting-cp | 06:31 | |
| *** robcresswell has joined #openstack-meeting-cp | 06:31 | |
| *** ameade has joined #openstack-meeting-cp | 06:31 | |
| *** TheJulia has joined #openstack-meeting-cp | 06:31 | |
| *** kencjohnston has joined #openstack-meeting-cp | 06:31 | |
| *** persia has joined #openstack-meeting-cp | 06:31 | |
| *** knikolla has joined #openstack-meeting-cp | 06:31 | |
| *** edmondsw has joined #openstack-meeting-cp | 06:45 | |
| *** margaret has quit IRC | 06:48 | |
| *** robcresswell has quit IRC | 06:48 | |
| *** ameade has quit IRC | 06:48 | |
| *** TheJulia has quit IRC | 06:48 | |
| *** kencjohnston has quit IRC | 06:48 | |
| *** persia has quit IRC | 06:48 | |
| *** knikolla has quit IRC | 06:48 | |
| *** edmondsw has quit IRC | 06:50 | |
| *** TheJulia has joined #openstack-meeting-cp | 07:24 | |
| *** margaret has joined #openstack-meeting-cp | 07:24 | |
| *** robcresswell has joined #openstack-meeting-cp | 07:24 | |
| *** kencjohnston has joined #openstack-meeting-cp | 07:25 | |
| *** persia has joined #openstack-meeting-cp | 07:25 | |
| *** knikolla has joined #openstack-meeting-cp | 07:25 | |
| *** ameade has joined #openstack-meeting-cp | 07:25 | |
| *** edmondsw has joined #openstack-meeting-cp | 08:33 | |
| *** edmondsw has quit IRC | 08:38 | |
| *** yamahata has quit IRC | 09:00 | |
| *** edmondsw has joined #openstack-meeting-cp | 10:22 | |
| *** edmondsw has quit IRC | 10:26 | |
| *** edmondsw has joined #openstack-meeting-cp | 12:10 | |
| *** edmondsw has quit IRC | 12:14 | |
| *** coolsvap has quit IRC | 13:44 | |
| *** edmondsw has joined #openstack-meeting-cp | 13:58 | |
| *** edmondsw has quit IRC | 14:02 | |
| *** mriedem has joined #openstack-meeting-cp | 15:58 | |
| ildikov | #startmeeting cinder-nova-api-changes | 16:00 |
|---|---|---|
| openstack | Meeting started Thu Dec 21 16:00:22 2017 UTC and is due to finish in 60 minutes. The chair is ildikov. Information about MeetBot at http://wiki.debian.org/MeetBot. | 16:00 |
| openstack | Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. | 16:00 |
| *** openstack changes topic to " (Meeting topic: cinder-nova-api-changes)" | 16:00 | |
| openstack | The meeting name has been set to 'cinder_nova_api_changes' | 16:00 |
| ildikov | johnthetubaguy jaypipes e0ne jgriffith hemna mriedem patrickeast smcginnis diablo_rojo xyang1 raj_singh lyarwood jungleboyj stvnoyes | 16:00 |
| mriedem | o/ | 16:00 |
| jungleboyj | @! | 16:00 |
| _pewp_ | jungleboyj (。･д･)ﾉﾞ | 16:00 |
| ildikov | mriedem: jungleboyj: morning :) | 16:01 |
| stvnoyes | o/ | 16:01 |
| jungleboyj | Snowy morning. :-) | 16:01 |
| ildikov | I don't expect a crowd today and only have a few topics | 16:02 |
| ildikov | stvnoyes: hi :) | 16:02 |
| stvnoyes | hi :-) | 16:02 |
| ildikov | ok, let's start | 16:03 |
| ildikov | so we need to stamp the Cinder spec on multi-attach: https://review.openstack.org/#/c/523608/ | 16:03 |
| ildikov | I haven't seen an update from jgriffith as of yet since the last round of comments | 16:04 |
| jungleboyj | ildikov: Me either. I looked at that earlier. jgriffith seems to be AWOL. ;-) | 16:04 |
| ildikov | I think we are kind of clear on most the items | 16:05 |
| ildikov | I hope we can clean up the shelved offloaded stuff at some point with a Nova microversion bump | 16:05 |
| ildikov | running this late in the cycle it won't happen now though | 16:06 |
| mriedem | that's probably an easy cleanup for rocky | 16:06 |
| ildikov | I think the volume type aspect is clear now | 16:06 |
| ildikov | mriedem: +1 | 16:06 |
| jungleboyj | mriedem: ++ | 16:07 |
| ildikov | and we need some client clean up too on the Cinder side as pointed out by hemna | 16:07 |
| jungleboyj | Yeah, the --multi-attach option? | 16:07 |
| ildikov | yep, I meant that one | 16:08 |
| jungleboyj | ildikov: Ok, will have to go through deprecation and stuff on that, but not a big deal. | 16:08 |
| ildikov | jungleboyj: sure, we can look into that | 16:09 |
| ildikov | jungleboyj: it's a bit of a corner case though, but we can figure out how to handle it once we get there | 16:09 |
| ildikov | jungleboyj: do you want to give it a try to update the spec? | 16:10 |
| jungleboyj | ildikov: Looking at it, I was really hoping that jgriffith would jump in there but I can try to address the comments and see if they satisfy people. | 16:10 |
| ildikov | jungleboyj: let's give it a try and hopefully we can get it in soon now | 16:11 |
| ildikov | we can always update it later if needed | 16:11 |
| jungleboyj | Ok. I will give that a shot after the meeting here. | 16:12 |
| ildikov | jungleboyj: thank you | 16:12 |
| jungleboyj | ildikov: Welcome. | 16:12 |
| ildikov | I will hunt down jgriffith after the spec lands to see where he got with the code parts in the meantime | 16:13 |
| ildikov | on the Nova side I cleaned up the libvirt patch: https://review.openstack.org/#/c/267587/ | 16:14 |
| ildikov | I know that we have pseudo code in the Nova spec for the lock, but if someone would have a ton of time I could use some 'how to do locks' lessons | 16:15 |
| ildikov | as I'm not sure I got the full picture on how it should look like by the end... | 16:15 |
| mriedem | well the first thing there is we need the shared_targets flag off the volume | 16:15 |
| ildikov | yeah, that'll decide on whether or not we will have to lock | 16:16 |
| mriedem | so we have to get the volume at the new microversion, i assume when we know we're detaching from a volume that has multiple attachments? | 16:16 |
| mriedem | would have to look at the spec again | 16:16 |
| ildikov | didn't we kind of want to have that lock for shared_targets overall? | 16:17 |
| ildikov | https://specs.openstack.org/openstack/nova-specs/specs/queens/approved/cinder-volume-multi-attach.html | 16:17 |
| mriedem | so that can be attachments from different volumes? | 16:18 |
| mriedem | i.e. it's an existing problem? | 16:18 |
| jungleboyj | ildikov: The locking was for shared_targets, yes. | 16:18 |
| ildikov | I think we said that there's an existing problem with shared_targets it's just gets worse with multi-attach | 16:18 |
| mriedem | ok i missed that part | 16:18 |
| ildikov | *it | 16:19 |
| jungleboyj | Right. | 16:19 |
| mriedem | so if we're going to do that regardless of the volume saying it's multiattach or not, i guess that patch comes in before the multiattach one right? | 16:19 |
| ildikov | we had a chat about this at the PTG with the Cinder guys saying that brick is smarter now, etc, but there's in general a shared_targets issue that locks would help with | 16:19 |
| ildikov | yeah, in that case it should | 16:20 |
| ildikov | jungleboyj: is there any bug report about this issue or it's just an urban legend that we believe is true? | 16:20 |
| jungleboyj | ildikov: Just a legend. | 16:21 |
| ildikov | jungleboyj: never mind then :) | 16:21 |
| mriedem | ok, so i can maybe help a big here on the plumbing | 16:22 |
| mriedem | *bit | 16:22 |
| ildikov | I have very basic questions on how locks work in general | 16:23 |
| ildikov | like I saw instance lock in the code but that doesn't fully answer my how exactly can I lock the host | 16:23 |
| mriedem | depends on if the lock is external or not | 16:24 |
| ildikov | or well, the attach/detach volume operations to get them serialized | 16:24 |
| mriedem | we use https://github.com/openstack/nova/blob/master/nova/utils.py#L71 | 16:24 |
| jgriffith | Yes, that mv bump to expose shared_targets and lock name should be before multiattach | 16:24 |
| jgriffith | (sorry, joining the party late) | 16:24 |
| mriedem | https://github.com/openstack/oslo.concurrency/blob/master/oslo_concurrency/lockutils.py#L291 | 16:24 |
| mriedem | you can pass external=True here https://github.com/openstack/oslo.concurrency/blob/master/oslo_concurrency/lockutils.py#L231 | 16:24 |
| ildikov | external from what perspective? | 16:24 |
| mriedem | by default, the lock is just per-process | 16:25 |
| mriedem | an external lock is on the file system | 16:25 |
| * ildikov hides after that question now... | 16:25 | |
| mriedem | so across processes | 16:25 |
| ildikov | oh, ok, makes sense | 16:25 |
| mriedem | that's only per-host though, | 16:25 |
| mriedem | not across all hosts, that would be a DLM | 16:25 |
| mriedem | like etd | 16:25 |
| mriedem | *etcd | 16:25 |
| ildikov | we want it per host, so that's fine | 16:25 |
| mriedem | now, | 16:25 |
| mriedem | you can only run one nova-compute worker, | 16:25 |
| mriedem | so i'm not sure we need an external lock, | 16:26 |
| mriedem | unless someone is running multiple nova-compute processes on the same host | 16:26 |
| mriedem | which is possible, but i'm not sure how many people actually do that | 16:26 |
| ildikov | is it a huge extra burden to add 'external'? | 16:26 |
| mriedem | no, but it could unnecessarily impose a semaphore where one isn't needed | 16:27 |
| mriedem | which would impact performance | 16:27 |
| mriedem | i think starting with a non-external lock is fine, | 16:27 |
| ildikov | jgriffith: we were wondering if there's any explanation on the shared targets issue that exists independently from multi-attach that we discussed at the PTG | 16:27 |
| jgriffith | my 2 cents is that I didn't think we needed the external in this case | 16:27 |
| mriedem | if someone comes forward saying "omfg this breaks my crazy multi-nova-compute on single host deployment" then we can change it | 16:28 |
| ildikov | ok, fair enough | 16:28 |
| ildikov | just wanted to see the options here | 16:28 |
| mriedem | ok so i could probably wip something up | 16:28 |
| jgriffith | ildikov: the shared-targets issues outside of multi-attach? Just the "unsafe" removal of a target that's still in use | 16:29 |
| jgriffith | brick deals with it today but there are races there that can be problematic | 16:29 |
| ildikov | mriedem: that would be pretty great :) | 16:29 |
| jgriffith | although currently my understanding from folks is that it's pretty solid now | 16:29 |
| ildikov | jgriffith: I think I remembered the races discussion | 16:29 |
| jgriffith | ildikov: the conclusion we came to was that it was a nice improvement/enhancement for the non-multiattach world, and a requirements for multiattach | 16:30 |
| ildikov | jgriffith: thanks for confirming | 16:30 |
| jungleboyj | jgriffith: Welcome! | 16:30 |
| ildikov | jgriffith: I guess there isn't anything that would describe the need for this nice improvement/enhancement? :) | 16:31 |
| mriedem | https://specs.openstack.org/openstack/cinder-specs/specs/queens/add-shared-targets-to-volume-ref.html | 16:32 |
| mriedem | probably there ^ | 16:32 |
| ildikov | mriedem: ah, right, I forgot that it was written without any multi-attach poisoning in it... | 16:33 |
| ildikov | then we can document that change as well | 16:33 |
| mriedem | we should get this updated https://developer.openstack.org/api-ref/block-storage/v3/#show-a-volume-s-details | 16:34 |
| mriedem | for the new fields | 16:34 |
| ildikov | I can look into that | 16:34 |
| ildikov | mriedem: if you upload a WIP on the lock stuff, I will look into the API microversion bump and the needed Cinder changes | 16:35 |
| mriedem | https://bugs.launchpad.net/cinder/+bug/1739640 | 16:35 |
| openstack | Launchpad bug 1739640 in Cinder "3.48 volume response parameters are not in the v3 API reference" [Undecided,New] | 16:35 |
| mriedem | here you go | 16:35 |
| mriedem | i'll crank something out quick on the lock thing | 16:35 |
| ildikov | mriedem: sweet, tnx | 16:35 |
| ildikov | mriedem: is there anything in the libvirt patch that I should add/change? | 16:36 |
| mriedem | i haven't looked at it lately | 16:37 |
| ildikov | jgriffith: jungleboyj will update your spec on multi-attach; will you be around to answer questions and/or check the updated version? | 16:37 |
| ildikov | mriedem: no problem, I was just wondering | 16:38 |
| jungleboyj | jgriffith: Or can you make the updates so I can then approve it? | 16:38 |
| jgriffith | jungleboyj: will do | 16:38 |
| jungleboyj | jgriffith: You will make updates? If so, can we do that today? | 16:38 |
| jgriffith | jungleboyj: yes and yes | 16:39 |
| ildikov | jgriffith: even better, thank you so much | 16:39 |
| jungleboyj | jgriffith: You the man! Thank you. | 16:39 |
| ildikov | I haven't gotten to the Tempest test yet | 16:40 |
| *** iyamahat has quit IRC | 16:40 | |
| ildikov | I hope our expert, stvnoyes will be able to help out with that part in January :) | 16:40 |
| mriedem | i assume stvnoyes is frantically working on backporting all of this to kilo for oracle | 16:41 |
| ildikov | I hope he found slaves for that :) | 16:41 |
| mriedem | motion to adjourn | 16:41 |
| jgriffith | I second | 16:42 |
| jungleboyj | I third | 16:42 |
| ildikov | I give up :) | 16:42 |
| jgriffith | jungleboyj: no no no, procedure after a second is "all in favor" | 16:42 |
| jgriffith | :) | 16:42 |
| jungleboyj | all in favor? | 16:42 |
| mriedem | aye | 16:42 |
| ildikov | for today only though :) | 16:42 |
| jgriffith | aye | 16:42 |
| jungleboyj | They aye's have it. | 16:42 |
| ildikov | Happy Holidays All! | 16:43 |
| ildikov | that was my last topic | 16:43 |
| ildikov | no meeting next week | 16:43 |
| jungleboyj | Hope everyone has a Merry Christmas and a safe New Year. | 16:43 |
| jgriffith | Indeed Merry Christmas everyone | 16:43 |
| ildikov | I will check who's around on the 4th | 16:43 |
| jungleboyj | jgriffith: give me a ping when the spec is ready in case I don't see that come through. | 16:43 |
| jungleboyj | ildikov: I am in Workshops in Raleigh so I will probably only be peripherally there. | 16:44 |
| ildikov | jungleboyj: no worries, I'm an expert in talking to myself | 16:44 |
| * jungleboyj laughs | 16:44 | |
| ildikov | with all that said | 16:45 |
| ildikov | Thank You All for everything! | 16:45 |
| jungleboyj | ildikov: Thank you! | 16:45 |
| ildikov | it's a hell of a ride, but hopefully we will get there now | 16:45 |
| * jungleboyj sings "All I want for Christmas is multi-attach" | 16:46 | |
| ildikov | have fun and some rest and whatever makes you happy with family, friends, cats, etc next week(s) | 16:46 |
| ildikov | jungleboyj: that makes me nauseous... | 16:47 |
| ildikov | That's all Folks! | 16:47 |
| ildikov | :) | 16:47 |
| ildikov | #endmeeting | 16:47 |
| *** openstack changes topic to "OpenStack Meetings || https://wiki.openstack.org/wiki/Meetings" | 16:47 | |
| openstack | Meeting ended Thu Dec 21 16:47:41 2017 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4) | 16:47 |
| openstack | Minutes: http://eavesdrop.openstack.org/meetings/cinder_nova_api_changes/2017/cinder_nova_api_changes.2017-12-21-16.00.html | 16:47 |
| openstack | Minutes (text): http://eavesdrop.openstack.org/meetings/cinder_nova_api_changes/2017/cinder_nova_api_changes.2017-12-21-16.00.txt | 16:47 |
| openstack | Log: http://eavesdrop.openstack.org/meetings/cinder_nova_api_changes/2017/cinder_nova_api_changes.2017-12-21-16.00.log.html | 16:47 |
| *** nhelgeson has joined #openstack-meeting-cp | 17:01 | |
| *** mriedem has left #openstack-meeting-cp | 17:03 | |
| *** kmalloc has joined #openstack-meeting-cp | 17:13 | |
| *** aselius has joined #openstack-meeting-cp | 17:21 | |
| *** iyamahat has joined #openstack-meeting-cp | 17:29 | |
| *** iyamahat has quit IRC | 17:29 | |
| *** yamahata has joined #openstack-meeting-cp | 17:31 | |
| *** edmondsw has joined #openstack-meeting-cp | 17:39 | |
| *** edmondsw has quit IRC | 17:44 | |
| *** iyamahat has joined #openstack-meeting-cp | 17:45 | |
| *** harlowja has joined #openstack-meeting-cp | 18:16 | |
| *** iyamahat has quit IRC | 18:29 | |
| *** yamahata has quit IRC | 18:32 | |
| *** iyamahat has joined #openstack-meeting-cp | 18:49 | |
| *** yamahata has joined #openstack-meeting-cp | 19:07 | |
| *** edmondsw has joined #openstack-meeting-cp | 19:28 | |
| *** edmondsw has quit IRC | 19:33 | |
| *** diablo_rojo has quit IRC | 19:51 | |
| *** diablo_rojo has joined #openstack-meeting-cp | 20:51 | |
| *** andreaf has quit IRC | 21:55 | |
| *** andreaf has joined #openstack-meeting-cp | 21:56 | |
| *** diablo_rojo has quit IRC | 22:07 | |
| *** diablo_rojo has joined #openstack-meeting-cp | 22:13 | |
| *** diablo_rojo has quit IRC | 22:49 | |
| *** edmondsw has joined #openstack-meeting-cp | 23:04 | |
| *** stvnoyes has quit IRC | 23:06 | |
| *** edmondsw has quit IRC | 23:08 | |
| *** kbyrne has quit IRC | 23:33 | |
| *** kbyrne has joined #openstack-meeting-cp | 23:39 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!