14:01:06 <edmondsw> #startmeeting PowerVM Driver Meeting 14:01:07 <openstack> Meeting started Tue Oct 9 14:01:06 2018 UTC and is due to finish in 60 minutes. The chair is edmondsw. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:01:09 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:01:11 <openstack> The meeting name has been set to 'powervm_driver_meeting' 14:01:32 <efried> ō/ 14:01:33 <edmondsw> #link agenda https://etherpad.openstack.org/p/powervm_driver_meeting_agenda 14:01:53 <edmondsw> mujahidali: mdrabe: ping 14:02:02 <edmondsw> #topic In-Tree Driver 14:02:41 <edmondsw> I don't believe there's been any work going on here for a while 14:03:09 <edmondsw> I'm hoping that mujahidali will be able to start working on more porting from OOT to IT eventually 14:03:24 <edmondsw> but the next stuff up requires multi-node CI, so he's gotta get that working first 14:03:32 <edmondsw> comments from anyone else? 14:03:36 <efried> nay 14:04:48 <edmondsw> #topic Out-of-Tree Driver 14:05:24 <edmondsw> We have opted-in for the community to help us upgrade to newer versions of neutron-lib as the release 14:05:36 <edmondsw> via https://review.openstack.org/#/c/607625/ 14:05:39 <efried> \o/ 14:05:58 <edmondsw> we also fixed a bug https://review.openstack.org/#/c/608023/ 14:06:17 <edmondsw> I think that's it since the last meeting 14:06:31 <efried> oh, does -^ need to be ported IT? 14:06:34 <efried> (I didn't look) 14:07:04 <edmondsw> efried ah, that's a good point... I thought about backporting (don't think that's needed at this point) but not IT 14:07:39 <edmondsw> I'll check after the meeting if nobody beats me to it 14:08:06 <efried> ✔ 14:08:19 <edmondsw> mdrabe: how is your schedule looking for working on MSP? 14:09:05 <edmondsw> I've been tied up in other things, so I might need you to take over getting the multi-node devstack env working 14:09:14 <edmondsw> but I can help advise you there at least 14:11:02 <edmondsw> alright, moving on 14:11:22 <edmondsw> #topic Device Passthrough 14:11:25 <edmondsw> efried ^ 14:11:43 <efried> Jay has agreed (again) to review kosamara's spec. 14:12:02 <efried> But we have a suggestion from Sylvain to break it apart into one spec for the file format and one for the passthrough itself. 14:12:16 <efried> because the file format has the potential to gain traction for use in a lot of areas other than just device passthrough. 14:12:25 <efried> Basically as a generic provider configurator 14:12:32 <edmondsw> interesting 14:12:48 <efried> which is getting interest from several high-priority blueprints right now 14:13:26 <efried> So if we can land something around that file spec in nova proper, we could implement passthrough fairly easily on top of it - whether IT or OOT. 14:14:03 <efried> Placement extraction is going well. Upgrade stuff is getting close to working. 14:14:16 <efried> Reshaping & nrp work is moving along 14:14:28 <efried> we keep hitting new and interesting issues (e.g. see gibi's ML post this morning) 14:15:12 <efried> I've been keeping abreast and weighing in, but not doing a lot of actual writing of spec/code content due to time split. 14:16:06 <efried> I think that's about it, unless someone wants more details on anything. 14:16:24 <edmondsw> thanks 14:16:38 <edmondsw> #topic PowerVM CI 14:16:46 <edmondsw> mujahidali how is the CI? 14:16:58 <mujahidali> After redeployment followed by Power outage, CI was broken and we added "AttachInterfacesUnderV243Test.test_add_remove_fixed_ip " to blacklist. 14:17:17 <mujahidali> CI looks good now. 14:17:36 <edmondsw> yep, I noticed we were getting some good votes. Tintri looks pretty clean 14:17:46 <edmondsw> thanks for fixing that up 14:18:02 <mujahidali> I have installed diskimage-builder and updated nodepool to 3.2.0 on stage env. 14:18:16 <mujahidali> I am currently trying to figure out the nodepool configuration to use the diskimage-builder. I am using this https://github.com/openstack-infra/project-config-example/blob/master/nodepool/nodepool.yaml as refernce. 14:19:40 <mujahidali> I need some more tweaking with the nodepool.yaml file and then we may get a wroking nodepool config. 14:20:39 <edmondsw> that'll be awesome 14:20:46 <mujahidali> :) 14:21:28 <mujahidali> Once nodepool is working with disimage-builder we need to figure out the SSP part for multinode. 14:22:04 <edmondsw> I can help you create an SSP that spans 2 nodes. I've done that before 14:22:20 <mujahidali> that will be a great help. 14:22:23 <mujahidali> thanks 14:22:39 <edmondsw> I'll leave it to you to figure out how to configure the CI to know which 2 nodes are paired 14:22:55 <mujahidali> okay 14:24:06 <mujahidali> that's it from me. 14:24:26 <edmondsw> mujahidali I'd suggest that we move to the new nodepool and diskimage-builder and all that first, and then worry about making the CI multinode in a later commit(s) 14:24:33 <edmondsw> i.e. try to stage the work out 14:24:43 <edmondsw> sound good? 14:24:54 <mujahidali> sure. 14:25:23 <mujahidali> so, let's break it out in two parts 14:25:29 <edmondsw> you didn't mention zookeeper. Have you been working with that as well? 14:25:36 <mujahidali> upgrade CI 14:25:45 <mujahidali> and then enable multinode 14:25:48 <edmondsw> yes 14:26:52 <edmondsw> alright, tx 14:26:57 <edmondsw> #topic Open Discussion 14:27:56 <edmondsw> efried the community has a "reason" arg for Unplug exceptions but not for plug: https://github.com/openstack/nova/blob/master/nova/exception.py#L132-L137 14:28:25 <efried> nice 14:28:26 <edmondsw> so while OOT we customize the Plug exception to give reason information (at the expense of saying it's a plug... which maybe we should change) 14:28:42 <edmondsw> IT we just stick with the generic message: https://github.com/openstack/nova/blob/master/nova/virt/powervm/vif.py#L116-L118 14:29:33 <edmondsw> I guess that's ok for IT, until/unless they add reason to the exception 14:29:37 <efried> tbh if we're using LOG.exception I'm not sure what benefit we would get by shoving a %reason in there 14:30:08 <edmondsw> but I'll probably propose a follow-on OOT to add a "Virtual interface plugin failed: " prefix to the message we create there 14:30:53 <edmondsw> efried yeah, I think the benefit would only be if the message gets used in the REST response 14:31:03 <efried> which it doesn't. 14:31:09 <efried> Because the REST operation is asynchronous. 14:31:23 <efried> You notice the instance go to ERROR state. 14:31:27 <efried> and then you go look at logs. 14:31:29 <edmondsw> hmm... then why did Arun care about this, I wonder 14:31:46 <efried> oh, because it would have been raising a spurious exception in the logs, obscuring the actual reason for the error. 14:32:14 <efried> Oh, I was talking about spawn - there may be a synchronous plug operation that does surface the error. 14:32:18 <efried> idk for sure 14:32:32 <edmondsw> I'll try to catch Arun and ask him for more detail 14:32:33 <efried> either way it's a correct fix 14:32:36 <edmondsw> yeah 14:32:59 <edmondsw> better than it was, for sure 14:33:09 <edmondsw> if that's it, I'll let y'all go 14:33:26 <edmondsw> #endmeeting