16:00:08 <ildikov> #startmeeting cinder-nova-api-changes
16:00:09 <openstack> Meeting started Thu Aug  3 16:00:08 2017 UTC and is due to finish in 60 minutes.  The chair is ildikov. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:10 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:00:12 <openstack> The meeting name has been set to 'cinder_nova_api_changes'
16:00:19 <ildikov> johnthetubaguy jaypipes e0ne jgriffith hemna mriedem patrickeast smcginnis diablo_rojo xyang1 raj_singh lyarwood jungleboyj stvnoyes
16:00:36 <stvnoyes> o/
16:00:39 * jungleboyj is lurking
16:00:53 <jgriffith> o/
16:01:06 <xyang1> Hi
16:01:21 <ildikov> hi all :)
16:01:37 <_pewp_> hemna ヾ(-_-;)
16:02:53 <mriedem> o.
16:03:02 <ildikov> ok, let's start
16:03:14 <smcginnis> o/
16:03:33 <ildikov> so the big news since last week is that the uWSGI modifications broke the microversion check in the cinderclient
16:04:05 <ildikov> which means that all the gate jobs that try to use the new attach flow blow up when we check whether 3.27 is available or not
16:04:14 <hemna> neat
16:04:25 <ildikov> mriedem: if you have any uWSGI suggestions for us, that would be great :)
16:04:42 <mriedem> because the version discovery breaks?
16:05:02 <mriedem> andreykurilin fixed something like that in novaclient, sec
16:05:16 <ildikov> I reported a bug for this yesterday: https://bugs.launchpad.net/devstack/+bug/1708188
16:05:16 <openstack> Launchpad bug 1708188 in python-cinderclient "get_highest_client_server_version is failing with the new uWSGI changes in Devstack" [Undecided,New]
16:05:34 <mriedem> https://github.com/openstack/python-novaclient/commit/bd0a2adefe5e89c4269f6589837f871764c6f1f6
16:05:34 <ildikov> devstack patch: https://review.openstack.org/#/c/489715/
16:05:48 <ildikov> cinderclient patch: https://review.openstack.org/#/c/490040/
16:06:08 <smcginnis> mriedem: That looks like the same issue we're having.
16:06:23 * smcginnis was really hoping it wasn't something that required a client update...
16:06:40 <ildikov> mriedem: coolio, thanks
16:06:51 <ildikov> mriedem: I was looking for something like that :)
16:07:21 <ildikov> I will point e0ne to that novaclient fix
16:07:46 <smcginnis> ildikov: Looks like he's not online or I was going to do that right now.
16:07:48 <mriedem> i already put it in the cinderclient wip
16:07:59 <smcginnis> mriedem: Thanks
16:08:01 <ildikov> mriedem: awesome, thanks
16:08:02 <mriedem> so community goal for pike is cinder api in uwsgi,
16:08:17 <mriedem> but doing that breaks cinderclient version discovery, which is needed for version negotiation with the v3 api yes?
16:08:35 <smcginnis> Yep, it appears so
16:08:38 <mriedem> so wouldn't fixing that in cinderclient require a fix and release for pike?
16:08:45 <mriedem> post-freeze i mean
16:09:03 <mriedem> doesn't mean anything to nova really since the patches that rely on it wont be in pike, but just thinking about timelines and process
16:09:05 <smcginnis> Yes. We're well into freeze, so I don't think we can fix until unfreeze.
16:09:14 <mriedem> i guess you could backport the fix later
16:09:19 <mriedem> and release from stable/pke
16:09:21 <mriedem> *pike
16:09:25 <smcginnis> But we can do a stable release right away when it opens up.
16:09:30 <ildikov> mriedem: I guess for backward compatibility it would be great if it does not blow up
16:09:41 <ildikov> I guess we can handle random errors from the cinderclient
16:09:46 <smcginnis> We just can't raise the g-r minimum, so no guarantee that the fixed client is in use in Pike.
16:10:09 <smcginnis> So worse case - this fails and negotiates back to legacy attach, right?
16:10:32 <ildikov> smcginnis: yes
16:11:07 <smcginnis> So the world isn't blowing up, but the kitchen is on fire.
16:11:13 <smcginnis> SO normal state.
16:11:39 <jgriffith> smcginnis :)
16:11:47 <ildikov> I can catch the ValueError that it throws and fall back
16:12:30 * jungleboyj hands smcginnis a fire extinguisher
16:12:32 <ildikov> but honestly I would not like to have too many of those there as then we will silently fall back to the old flow for various reasons in the future that we might not catch
16:12:48 <ildikov> and we need to fix it quickly so we can finally move forward with the attach patch
16:12:56 <ildikov> for which I have clean test runs locally
16:13:02 <smcginnis> ++
16:13:11 <ildikov> with the blockdev error still in the periodic task
16:14:05 <ildikov> I didn't see anything else failing in the tempest runs, but I have an overly patched env right now...
16:14:16 <ildikov> jgriffith: do you have any additions to this?
16:14:44 <jgriffith> too much to even go into
16:15:02 <jgriffith> runs are extremely inconsistent right now which is worrisome
16:15:07 <jgriffith> and not just around attach/detach
16:15:19 <jgriffith> random failures.. as smcginnis mentioned "SO normal state"
16:15:39 <smcginnis> :)
16:15:50 <jgriffith> It does seem as though it's a race somewhere; running tests serially seems "ok" most of the time
16:17:02 <ildikov> does not sound good
16:17:29 <jgriffith> ildikov meh... nothing earth shatteringly new
16:17:36 <ildikov> true :)
16:17:44 <jgriffith> I'm optimistic that I'll figure it out :)
16:18:08 * ildikov puts on her optimistic hat too then :)
16:18:25 <ildikov> ok, I don't think we have enough info to talk about this more here
16:18:40 <ildikov> jgriffith: or do you have anything else to add?
16:18:51 <jgriffith> ildikov nope
16:18:55 <ildikov> ok
16:19:35 <ildikov> stvnoyes: the two patches posted in the cinderclient bug report are fixing things in a Devstack env if you would need to run tests locally
16:19:54 <stvnoyes> ok thx
16:20:08 <ildikov> stvnoyes: anything in your tempest test or for live migrate to talk about here?
16:20:47 <stvnoyes> no, it would be great to get a review but I understand that's pretty far down the priority list for now
16:21:15 <stvnoyes> before I forget what I did
16:21:27 <ildikov> sounds reasonable
16:22:04 <ildikov> as we hope to get this fixed and merged before the PTG I hope we can fix the version check in the cinderclient soon so you get clean test runs on the gate for your part
16:22:17 <stvnoyes> that would be great
16:22:35 <ildikov> smcginnis: keep in touch about the client release when the issue is fixed there
16:22:46 <ildikov> anything else from anyone for today?
16:22:57 <stvnoyes> not for me
16:23:54 <ildikov> alright, then that's it for today
16:23:58 <smcginnis> Thanks
16:24:02 <ildikov> mriedem: thanks again for the pointer
16:24:21 <ildikov> have a good day everyone!
16:24:30 <ildikov> #endmeeting