| *** dwayne has joined #openstack-powervm | 00:14 | |
| *** kylek3h has joined #openstack-powervm | 00:18 | |
| *** thorst has joined #openstack-powervm | 00:25 | |
| *** k0da has quit IRC | 00:28 | |
| *** thorst has quit IRC | 00:29 | |
| *** thorst has joined #openstack-powervm | 01:02 | |
| *** thorst has quit IRC | 01:02 | |
| *** esberglu has joined #openstack-powervm | 01:10 | |
| *** esberglu has quit IRC | 01:11 | |
| *** seroyer has joined #openstack-powervm | 01:33 | |
| *** thorst has joined #openstack-powervm | 02:03 | |
| *** seroyer has quit IRC | 02:05 | |
| *** thorst has quit IRC | 02:08 | |
| *** kjw3 has joined #openstack-powervm | 02:23 | |
| *** thorst has joined #openstack-powervm | 02:54 | |
| *** thorst has quit IRC | 02:54 | |
| *** kjw3 has quit IRC | 03:12 | |
| *** thorst has joined #openstack-powervm | 03:16 | |
| *** thorst has quit IRC | 03:17 | |
| *** apearson has joined #openstack-powervm | 04:07 | |
| *** thorst has joined #openstack-powervm | 04:17 | |
| *** thorst has quit IRC | 04:22 | |
| *** k0da has joined #openstack-powervm | 05:05 | |
| *** thorst has joined #openstack-powervm | 05:18 | |
| *** thorst has quit IRC | 05:23 | |
| *** edmondsw has joined #openstack-powervm | 06:13 | |
| *** apearson has quit IRC | 06:15 | |
| *** edmondsw has quit IRC | 06:17 | |
| *** thorst has joined #openstack-powervm | 06:19 | |
| *** thorst has quit IRC | 06:24 | |
| *** k0da has quit IRC | 07:19 | |
| *** thorst has joined #openstack-powervm | 07:20 | |
| *** thorst has quit IRC | 07:24 | |
| *** thorst has joined #openstack-powervm | 08:21 | |
| *** thorst has quit IRC | 08:25 | |
| *** k0da has joined #openstack-powervm | 08:39 | |
| *** thorst has joined #openstack-powervm | 09:22 | |
| *** thorst has quit IRC | 09:31 | |
| *** edmondsw has joined #openstack-powervm | 09:50 | |
| *** edmondsw has quit IRC | 09:54 | |
| *** edmondsw has joined #openstack-powervm | 10:22 | |
| *** edmondsw has quit IRC | 10:26 | |
| *** thorst has joined #openstack-powervm | 11:28 | |
| *** thorst has quit IRC | 11:33 | |
| *** seroyer has joined #openstack-powervm | 12:25 | |
| *** smatzek has joined #openstack-powervm | 12:31 | |
| *** smatzek_ has joined #openstack-powervm | 12:32 | |
| *** smatzek has quit IRC | 12:36 | |
| *** thorst has joined #openstack-powervm | 12:37 | |
| *** dwayne has quit IRC | 12:37 | |
| *** tblakes has joined #openstack-powervm | 12:59 | |
| *** esberglu has joined #openstack-powervm | 13:06 | |
| *** edmondsw has joined #openstack-powervm | 13:12 | |
| esberglu | #startmeeting powervm_ci_meeting | 13:31 |
|---|---|---|
| openstack | Meeting started Thu Mar 9 13:31:25 2017 UTC and is due to finish in 60 minutes. The chair is esberglu. Information about MeetBot at http://wiki.debian.org/MeetBot. | 13:31 |
| openstack | Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. | 13:31 |
| thorst | o/ | 13:31 |
| openstack | The meeting name has been set to 'powervm_ci_meeting' | 13:31 |
| efried | o/ | 13:31 |
| esberglu | #topic In Tree CI | 13:32 |
| esberglu | In tree CI is still looking good. One test that is hitting us every now and then | 13:33 |
| esberglu | http://184.172.12.213/85/442685/4/check/nova-in-tree-pvm/54cc6d5/powervm_os_ci.html | 13:34 |
| esberglu | Where it can't find an image | 13:34 |
| esberglu | Which is something I wanted to debug today | 13:34 |
| esberglu | #action esberglu: Debug failing IT CI test | 13:34 |
| thorst | hmm...that doesn't seem to be related to our driver | 13:34 |
| thorst | :-/ | 13:35 |
| esberglu | Other than that, I had to redeploy the staging CI. But the plan there is | 13:35 |
| esberglu | to have a set of changes to the whitelist | 13:35 |
| esberglu | So that as we start getting IT patches approved, we can have the update whitelist ready | 13:36 |
| thorst | neat | 13:36 |
| esberglu | #topic OOT CI | 13:37 |
| esberglu | OOT also has a couple failures I wanted to debug today | 13:37 |
| esberglu | http://184.172.12.213/66/443266/1/check/nova-out-of-tree-pvm/c04198b/powervm_os_ci.html | 13:37 |
| esberglu | These two tests have been causing problems on and off for a while | 13:38 |
| esberglu | It seemed like they were a side effect of another test we disabled, as they stopped showing up | 13:39 |
| esberglu | after we disabled another test from that same class | 13:39 |
| thorst | but are still an issue now? Just not as often? | 13:39 |
| esberglu | Yeah still an issue now. They seem to have started hitting us more this week | 13:40 |
| thorst | hmm....could debug in staging if we're failing >20% or so | 13:41 |
| esberglu | #action esberglu: Debug 2 failing rebuild tests for OOT CI | 13:41 |
| esberglu | Then we have the newton/ocata failures | 13:42 |
| esberglu | The fix has been merged for g-r master, just has to be cherry-picked to the other branches | 13:42 |
| esberglu | And the other part of the fix (in tempest) is still under review | 13:43 |
| esberglu | But we have a fix incoming | 13:43 |
| esberglu | #topic OSA CI | 13:44 |
| esberglu | I finally got the OSA CI dev. env. deployed after some struggles. | 13:44 |
| esberglu | I need to increase the size of the SSP though, I forgot that OSA requires a larger flavor | 13:45 |
| esberglu | And didn't make it big enough | 13:45 |
| thorst | staging CI? | 13:45 |
| esberglu | No 3rd environment | 13:45 |
| thorst | I just got a bunch of capacity for the v7k...so we can make all the SSPs larger at some point if needed. | 13:46 |
| esberglu | Cool. I don't think we need to atm. But might have to later | 13:46 |
| efried | You can do that without rebuilding the whole SSP, as I'm sure you know. | 13:47 |
| thorst | yep, just add new disks | 13:47 |
| esberglu | Okay. Might have questions later, I don't think I've done that before | 13:48 |
| esberglu | #topic Other Items | 13:49 |
| esberglu | At some point I want to move the undercloud from newton to ocata | 13:49 |
| esberglu | Just kind of waiting for a lull in the action to get that done | 13:49 |
| thorst | heh | 13:50 |
| thorst | lull | 13:50 |
| thorst | but yeah, agree that'd be a good update to do | 13:50 |
| esberglu | Also I've been working with bjayasan as he is trying to get tempest running for an SDE setup | 13:50 |
| esberglu | Still just trying to help him understand the flow | 13:51 |
| esberglu | What is the plan for that? Is that going to be a whole new CI setup? Or are we trying to integrate that into our CI eventually? | 13:51 |
| thorst | esberglu: no, he's just trying to run some tests with systems configured in a different way | 13:52 |
| thorst | so right now you run SSPs | 13:52 |
| thorst | so we have a coverage gap with anything not SSP | 13:52 |
| thorst | he's trying to validate some of the other style configurations - or get to that point really | 13:52 |
| esberglu | Oh okay. Sounded kind of like he was trying to get a full fledged CI running when we were working | 13:54 |
| esberglu | I'll touch base with him again today | 13:54 |
| thorst | esberglu: hmmm...don't think so | 13:54 |
| thorst | I am 90% sure he only has one server. | 13:54 |
| thorst | :-) | 13:54 |
| esberglu | Yeah that's why I brought it up. He was asking about the staging environment and how he could test stuff | 13:55 |
| thorst | hmm...I may touch base with him | 13:55 |
| esberglu | But staging definitely can't support that | 13:55 |
| thorst | the intention is simply to test different flavors of configurations | 13:55 |
| thorst | Jay to test the in-tree stuff (initially) with tempest. That'll eventually move to a SDE style env | 13:55 |
| thorst | nbante to test the oot stuff with temptest, on a SDE style env. | 13:56 |
| thorst | and we don't know how to bring SDE style env's into CI because they can't do the same tricks we're doing here... | 13:56 |
| esberglu | That's all I had today | 13:57 |
| esberglu | Thanks for joining | 13:58 |
| esberglu | #endmeeting | 13:58 |
| openstack | Meeting ended Thu Mar 9 13:58:51 2017 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4) | 13:58 |
| openstack | Minutes: http://eavesdrop.openstack.org/meetings/powervm_ci_meeting/2017/powervm_ci_meeting.2017-03-09-13.31.html | 13:58 |
| openstack | Minutes (text): http://eavesdrop.openstack.org/meetings/powervm_ci_meeting/2017/powervm_ci_meeting.2017-03-09-13.31.txt | 13:58 |
| openstack | Log: http://eavesdrop.openstack.org/meetings/powervm_ci_meeting/2017/powervm_ci_meeting.2017-03-09-13.31.log.html | 13:58 |
| *** apearson has joined #openstack-powervm | 14:01 | |
| efried | thorst esberglu We can test SSP and localdisk (and probably even LIO stuff?) on the same system. Just have to twiddle the conf and restart the compute service each time, neh? | 14:04 |
| thorst | potentially, yes | 14:09 |
| thorst | would need LIO and VIOSes co-existing | 14:09 |
| thorst | I don't know that apearson would say that's supported if we find issues :-) | 14:10 |
| adreznec | efried: thorst in general that should work (and we've had systems in similar configs in dev environments), though likely not for the way we do CI | 14:10 |
| efried | Since we build and stack a separate node for each run, we would just build that node with the appropriate config. | 14:11 |
| efried | ...except they share a nvl, riiight. | 14:11 |
| thorst | so it goes back to how we do the API | 14:12 |
| apearson | thorst - I'd say that's supported...no reason for it not to be. And it gives the user the ability to run with their traditional enterprise attach storage on the same system as they try out the cloud attach storage. So with standard NovaLink + OpenStack, I think we could support that model. | 14:12 |
| thorst | which is the discussion from yesterday | 14:12 |
| *** tjakobs has joined #openstack-powervm | 14:27 | |
| *** seroyer has quit IRC | 14:34 | |
| *** smatzek_ has quit IRC | 14:36 | |
| openstackgerrit | Eric Fried proposed openstack/networking-powervm master: WIP: Event-driven heal_and_optimize for SR-IOV https://review.openstack.org/436078 | 14:44 |
| *** waler has joined #openstack-powervm | 14:45 | |
| efried | thorst D'oh bug ^^ | 14:47 |
| thorst | I like the change...UT of course | 14:49 |
| thorst | that event is in 1.0.0.5? | 14:49 |
| esberglu | efried: thorst: I put up a change for the test that was failing IT CI | 14:49 |
| esberglu | It was failing because snapshot isn't implemented yet | 14:50 |
| *** alainfisher has joined #openstack-powervm | 14:50 | |
| esberglu | What I don't understand (and we have seen this on other tests for IT) | 14:50 |
| esberglu | Is why this test is not failing every time | 14:50 |
| thorst | what's the change set number? | 14:50 |
| efried | thorst 4968 | 14:51 |
| esberglu | Yeah | 14:51 |
| thorst | LGTM, but yeah...no idea how that would ever pass | 14:52 |
| *** seroyer has joined #openstack-powervm | 14:55 | |
| esberglu | Failing and passing both have the same NotImplementedError in the logs. But it only gets picked up as a failure sometimes. Idk | 14:57 |
| *** smatzek_ has joined #openstack-powervm | 14:57 | |
| *** jpasqualetto has joined #openstack-powervm | 14:57 | |
| *** smatzek_ has quit IRC | 14:57 | |
| thorst | lol | 14:58 |
| efried | I had an interesting conversation about this the other day in #openstack-nova - Just because we don't implement something doesn't mean it doesn't "succeed". | 14:58 |
| *** smatzek_ has joined #openstack-powervm | 14:58 | |
| adreznec | Well it's not like you want your tests to have consistent results, right? | 14:58 |
| efried | At least from horizon, the ops are asynchronous. So horizon telling you it succeeded just means it successfully submitted the request to the (nova) API. | 14:58 |
| efried | So I believe the tempest tests are similarly submitting the request, and then instead of checking the return from that command, they're polling for certain VM states or other conditions. | 14:59 |
| thorst | yep, you're right | 15:00 |
| efried | In other words, us raising NotImplementedError doesn't necessarily cause a failure. | 15:00 |
| efried | Now, if it's possible for a test to succeed sometimes, even when one of the critical ops in its path fails, I submit to you that it is a bad test. | 15:01 |
| efried | Not that I'm necessarily in favor of submitting bugs every time we find one of these... | 15:02 |
| efried | ...but that might be The Right Thing To Do. | 15:02 |
| waler | Morning, sorry to disturb, I'm seeking some clarification regarding integrating a POWER8/PowerVM compute node, into a broader OpenStack environment. I was advised this was the best place to ask? | 15:03 |
| efried | waler Probably so. What's up? | 15:03 |
| waler | am seeking some enlightenment to how (deployable) images are handled/treated? I know from a traditional PowerVM background how LPARs are handled storage wise. And I know from using PowerVC how you can "dd" images in and out. But it wasn't obvious from wiki.openstack.org/wiki/PowerVM how the images were handled when you don't have PowerVC in the middle | 15:05 |
| waler | I was presuming images would be "discovered", as opposed to pushed down from the controller node? | 15:06 |
| *** thorst is now known as thorst_afk | 15:06 | |
| efried | Well, for starters, you'll have to have images that are specifically built for the ppc64le architecture. So, like, you won't be able to use the same images you use for e.g. Intel. | 15:07 |
| waler | yep, appreciate that | 15:07 |
| efried | When we do it, we stuff the images in glance just like any other. | 15:07 |
| waler | so it's not the discovered approach, like it used to be with ICM <> PowerVC? | 15:07 |
| efried | Sorry, I don't have the background to answer that. Where would those guys "discover" from? | 15:08 |
| efried | I believe PowerVC uses cinder for everything. The native PowerVM compute driver can use ephemeral boot disks. | 15:09 |
| efried | You need to configure what kind of disk driver you want to use - either 'localdisk' or 'ssp'. | 15:09 |
| efried | 'localdisk' creates LVs on a specified VG on your VIOS(es). 'ssp' uses a Shared Storage Pool that you've got set up. | 15:10 |
| efried | Do you know which one of those you want to use? | 15:10 |
| waler | yes, I think we've got that far. In that our simple setup (one VIOS, one novalink, one SSP). | 15:11 |
| efried | Okay. And you've told the conf that you want to use the ssp disk driver? | 15:12 |
| efried | Should just need that one line in the conf, and the driver will discover your SSP. | 15:12 |
| efried | [powervm] | 15:13 |
| efried | disk_driver = ssp | 15:13 |
| waler | yep, and our upstream controllers, report correctly (RAM/CPU/Storage space) | 15:13 |
| efried | Cool. So have you tried a deploy yet? | 15:13 |
| waler | it's then what todo with an image? I'd presumed from the VC ways, that you'd "dd" the image into the SSP and OpenStack would somehow "discover" what was in the SSP for use? | 15:14 |
| efried | Oh, no, you don't have to do that. | 15:14 |
| efried | If the image is in glance, you just select that image when you deploy. | 15:14 |
| efried | Our driver will upload it for you. | 15:15 |
| efried | into a shiny new LU that it creates for your VM in the SSP. | 15:15 |
| efried | It's actually slightly cooler than that: If it's the first time you've used that image, the whole thing gets uploaded into an "image LU", but each VM you subsequently create against that image - even if it's on a separate host on the same SSP - will get a "thin clone" disk LU backed by that image LU. So the first one is slow, but subsequent ones are really fast. | 15:16 |
| waler | so where on the compute, would is store the glaced image, in the process? Does that transition through the novalink LPAR or the VIOS? | 15:17 |
| waler | ah, an image LU, that's the special type of LU within the SSP, if I'm following you correctly | 15:17 |
| efried | Right. | 15:17 |
| efried | I didn't quite catch your question above. If you've already put the image into glance (e.g. via `openstack image create`), then it'll have a name/UUID. | 15:18 |
| waler | so would that first time, transfer it through the novalink LPAR? or direct from controller to vios? | 15:19 |
| efried | When you do your spawn, you pick that image (either in horizon, or via --image on the CLI) | 15:19 |
| efried | Okay, I think I understand the question. | 15:19 |
| efried | It goes through the NovaLink partition to the VIOS. | 15:19 |
| efried | But glance is talking to the NovaLink partition, not to the controller node. | 15:20 |
| efried | The NovaLink partition is where the compute service runs. | 15:20 |
| efried | The compute process is what talks to the glance image API. | 15:20 |
| waler | yep, understand that. And from what you explained above, we need to treat an image like other images (ie. it's not discovered like with VC - good to know). I'm now trying to see how the flow works, as some AIX/IBMi images could be of the larger size. If they get staged down to the SSP, via the novalink LPAR, I presume I need to ensure the novalink LPAR has enough free disk space for temp storage? Or does the image just route through nov | 15:24 |
| waler | alink, it's not staged/copied there? | 15:24 |
| efried | waler - shouldn't need temp space on the NovaLink. | 15:27 |
| efried | It's streamed through. | 15:27 |
| waler | ah great, streamed, thank you! Thanks for the enlightenment. | 15:28 |
| efried | You bet. Good luck, and let us know if you have any issues. | 15:29 |
| waler | Just to check, my expectation that manually creating an LPAR elsewhere (on another POWER8, with another SSP) and then using the documented process to dd that out of the SSP, is a recommended route to creating images? I know it's the documented approach for backing up images when using VC, so hoped the same trick would be "good enough" here? | 15:34 |
| thorst_afk | waler: that should work. Just make sure you get the cloud-init on there | 15:39 |
| thorst_afk | and in a standard openstack install, swap to this interfaces.template on the powervm compute nodes | 15:39 |
| thorst_afk | https://ibm.ent.box.com/s/pjfy571tg6v5imu8kq59v32rdgmu7x37 | 15:39 |
| thorst_afk | that allows a mix of IPv4 and IPv6. The IPv6 on a NovaLink POWER8 is used for RMC | 15:40 |
| *** alainfisher has quit IRC | 15:40 | |
| waler | thorst_afk, efried thanks. We've actually got some S822L's running Ubuntu+KVM happily hooked into our broader OpenStack environment already. We're just trying todo the same with S822's now for AIX and IBMi deployments | 15:43 |
| waler | appreciate the clarification and enlightenment... | 15:44 |
| thorst_afk | waler: let us know how it goes | 15:45 |
| thorst_afk | and be sure to fill out the user survey too :-) | 15:45 |
| thorst_afk | we like having PowerVM show up in the openstack survey :-D | 15:46 |
| *** tjakobs has quit IRC | 16:03 | |
| *** dwayne has joined #openstack-powervm | 16:16 | |
| esberglu | thorst: efried: Getting rest errors on those 2 failing OOT CI tests trying to closevterm | 16:23 |
| esberglu | http://paste.openstack.org/show/602090/ | 16:23 |
| *** k0da has quit IRC | 16:35 | |
| thorst_afk | esberglu: yikes. efried any idea? That URL looks right...and that code has been in there forever. | 16:38 |
| *** thorst_afk is now known as thorst | 16:38 | |
| thorst | wonder if that's what you'd get if the console was already open? | 16:39 |
| *** tjakobs has joined #openstack-powervm | 16:39 | |
| esberglu | I'm still looking through the logs to see if anything else happened before that | 16:41 |
| *** Guest27328 has joined #openstack-powervm | 16:50 | |
| *** jpasqualetto has quit IRC | 17:05 | |
| *** alainfisher has joined #openstack-powervm | 17:12 | |
| *** Guest27328 has quit IRC | 17:13 | |
| *** alainfisher has quit IRC | 17:13 | |
| *** jpasqualetto has joined #openstack-powervm | 17:21 | |
| *** tjakobs has quit IRC | 17:50 | |
| *** alainfisher has joined #openstack-powervm | 18:28 | |
| *** tjakobs has joined #openstack-powervm | 18:29 | |
| efried | thorst esberglu That looks to me like the LPAR is already gone. | 18:39 |
| efried | I thought we were ignoring that one. Is it causing a failure? | 18:39 |
| *** alainfisher has left #openstack-powervm | 18:39 | |
| esberglu | We were talking about ignoring them. We never came to a conclusion and they stopped showing up for a while | 18:41 |
| efried | thorst eshaseth tested PS4 of https://review.openstack.org/#/c/436078/ and it works. Should I keep those debug messages in there or rip 'em out? | 18:42 |
| thorst | efried: I think btang had thoughts/feels about what we do when the lpar doesn't exist. | 19:00 |
| efried | Do tell. | 19:01 |
| thorst | efried: honestly I don't remember. It was some use case they were driving where they wanted better messages | 19:08 |
| thorst | we should probably ping him...he doesn't hang out in this channel | 19:08 |
| thorst | as for the messages...I do think those debugs are a bit aggressive. If doing another rev, I'd say rip em | 19:09 |
| thorst | or at least prune them... | 19:09 |
| efried | Had to do another rev for UT and one bugfix. | 19:11 |
| esberglu | efried: Thoughts on disabling those 2 tests? Or continue debugging | 19:15 |
| efried | esberglu Debug. | 19:15 |
| efried | The logs should hopefully make it clear whether the LPAR was deleted (or never existed) before we got to that point. | 19:15 |
| efried | In either case, I think a 404 on CloseVTerm should be ignored. | 19:16 |
| thorst | efried: I agree. | 19:23 |
| *** k0da has joined #openstack-powervm | 19:25 | |
| *** jpasqualetto_ has joined #openstack-powervm | 19:31 | |
| esberglu | efried: It looks like the LPAR never actually existed | 19:34 |
| *** jpasqualetto has quit IRC | 19:34 | |
| efried | esberglu Fine. | 19:34 |
| efried | Feel like proposing that fix? | 19:34 |
| *** jpasqualetto_ has quit IRC | 19:36 | |
| esberglu | Sure | 19:36 |
| *** alainfisher has joined #openstack-powervm | 19:38 | |
| *** jpasqualetto has joined #openstack-powervm | 19:45 | |
| openstackgerrit | Eric Fried proposed openstack/networking-powervm master: Event-driven heal_and_optimize for SR-IOV https://review.openstack.org/436078 | 19:49 |
| efried | thorst ^^ before you leave. | 19:50 |
| thorst | anything for you efried | 19:51 |
| efried | :) | 19:51 |
| thorst | so we still poll on the heal and optimize, this just escalates it a bit | 19:51 |
| thorst | for sr-iov | 19:51 |
| efried | though I guess svenkat has the power to approve. | 19:51 |
| thorst | well, lets get two +2's | 19:52 |
| efried | Yes, the original h-a-o behavior hasn't changed. This just triggers it extra on certain events. | 19:52 |
| thorst | svenkat will be owning this more than me - he's the one catching defects for it | 19:52 |
| efried | eshaseth tested it earlier (at PS4) and it works. | 19:52 |
| efried | So at least make sure I didn't break the logic with PS5 ;-) | 19:52 |
| efried | Note bug in PS3 that made it not work at all. Subtle. | 19:52 |
| thorst | hah | 19:53 |
| thorst | yeah, this looks solid to me | 19:54 |
| thorst | wish we could get a heal and optimize for the sea...but meh. | 19:54 |
| thorst | in terms of priorities...that's so low | 19:54 |
| efried | We gonna kick SEA to the kerb at some point anyway? | 19:55 |
| efried | For that matter, we'll need SR-IOV direct VIF. | 19:55 |
| efried | ...with some kind of answer for mobility... | 19:55 |
| efried | Okay, now back to trying to figure out how to break SSP into smaller change sets. | 19:55 |
| thorst | efried: SEA to the kerb...in like 50 years | 19:57 |
| thorst | I don't think we'll be enhancing it much, but its still super broadly used | 19:57 |
| efried | Sure. IBM time. | 19:57 |
| thorst | like Linux Bridge :-) | 19:57 |
| *** svenkat has joined #openstack-powervm | 19:57 | |
| efried | thorst Man, I can get rid of like 50LOC by ripping out the config option, but that hardly seems worth it. | 19:59 |
| efried | The only other gains I can make would be by ripping out test or doc. | 19:59 |
| thorst | svenkat: I'm good with https://review.openstack.org/#/c/436078/5 | 20:00 |
| efried | Or splitting out functionality - like create/connect in one change set, delete/disconnect in another - which I hate the idea of. | 20:00 |
| thorst | but since your team is the one supporting that, I want you to be the second +2 on it | 20:00 |
| svenkat | thorst: looking… | 20:00 |
| thorst | efried: yeah...I hate the idea too, but it is a bit overwhelming to look through right now | 20:00 |
| thorst | even to me...and I know the code. | 20:00 |
| svenkat | efried: looking good, is_hao_event filters out physicalport change (anything on phyisical port i guess, not just label) and triggers healandoptimize... | 20:03 |
| svenkat | looking ut | 20:03 |
| efried | svenkat Yeah, unfortunately it's as granular as we're able to get right now. | 20:04 |
| efried | But that should be okay. | 20:04 |
| svenkat | yess… adding my +2. | 20:06 |
| svenkat | efried: done. | 20:07 |
| efried | svenkat Okay, merging. Does this need an ocata cherry-pick, or are you going to carry it internally? | 20:07 |
| svenkat | is it possible to cherry pick into ocata? | 20:08 |
| efried | I would think so. thorst ? | 20:08 |
| efried | We have a bug for it. Have had for a while. | 20:08 |
| efried | https://review.openstack.org/#/c/443956/ Will let thorst dictate whether this goes in or not. | 20:10 |
| svenkat | after pep8 and py27 gates i will add my +2 | 20:11 |
| openstackgerrit | Merged openstack/networking-powervm master: Event-driven heal_and_optimize for SR-IOV https://review.openstack.org/436078 | 20:14 |
| svenkat | efried: there is an issue with ocata.. i think it is related to neutron_lib | 20:16 |
| efried | svenkat That shouldn't have happened, gr. Means networking-powervm is probably entirely busted in ocata. | 20:21 |
| efried | Will need a new fix for that. | 20:21 |
| svenkat | checking | 20:23 |
| svenkat | efried : from neutron import context as ctx - this is supposed to work in ocata isitnt? in master it is moved to neutron_lilb. is it correct | 20:28 |
| efried | svenkat I believe that's right. But this is a different error. | 20:29 |
| svenkat | first error is see is : from neutron import context as ctx ImportError: cannot import name context. | 20:31 |
| efried | svenkat Are you going to open a bug and backport https://review.openstack.org/#/c/442540/ ? | 20:44 |
| svenkat | yes… few minutes | 20:44 |
| *** k0da has quit IRC | 20:46 | |
| *** k0da has joined #openstack-powervm | 20:47 | |
| *** k0da has quit IRC | 20:54 | |
| *** alainfisher has left #openstack-powervm | 20:58 | |
| *** alainfisher has joined #openstack-powervm | 20:58 | |
| *** alainfisher has quit IRC | 20:58 | |
| svenkat | efried : opened bug : https://bugs.launchpad.net/networking-powervm/+bug/1671628 | 21:00 |
| openstack | Launchpad bug 1671628 in networking-powervm "Need to back port 442540 to coat" [High,New] - Assigned to Sridhar Venkat (svenkat) | 21:00 |
| *** smatzek_ has quit IRC | 21:01 | |
| efried | svenkat Roger that. Cherry-pick away. Then I'll rebase the HAO change set on top. | 21:01 |
| svenkat | sure… one moment | 21:01 |
| svenkat | done : https://review.openstack.org/#/c/443960/ | 21:02 |
| *** k0da has joined #openstack-powervm | 21:07 | |
| efried | svenkat I'll wait for jenkins. | 21:08 |
| svenkat | efreid sure. | 21:08 |
| thorst | efried svenkat: https://bugs.launchpad.net/bugs/1671628 | 21:11 |
| openstack | Launchpad bug 1671628 in networking-powervm "Need to back port 442540 to coat" [High,New] - Assigned to Sridhar Venkat (svenkat) | 21:11 |
| thorst | is that the right way to name the bug? | 21:11 |
| thorst | shouldn't the bug be a description of the issue? | 21:12 |
| efried | Yeah, that's not a good title. | 21:12 |
| efried | I didn't read it, mahbad. | 21:12 |
| thorst | efried: I mean, I know you've classified me as a project manager, so I'm just doing my part | 21:12 |
| efried | I'm actually betting there's already a bug for all the other stuff that got broke by that change. | 21:13 |
| svenkat | thorst: i added description in the details. i was looking for a short title - i see now - ocata became coat, my browser replaced it | 21:13 |
| efried | We should just add networking-powervm to that bug. | 21:13 |
| svenkat | I am rephrasing title | 21:15 |
| svenkat | called it Import context from neutron_lib | 21:17 |
| svenkat | cherrypicked into ocata also failed with similar error | 21:19 |
| efried | Hm, passes locally. Wonder what that's about. | 21:24 |
| efried | I wonder if this is a global reqs problem. | 21:25 |
| efried | thorst adreznec Should https://review.openstack.org/#/c/439805/2/requirements.txt be requiring neutron? | 21:26 |
| efried | ...or python-neutronclient? | 21:26 |
| thorst | efried: so...for stable branches yeah | 21:26 |
| thorst | not python-neutronclient | 21:26 |
| thorst | but the issue is we need it to pull down the LATEST neutron | 21:27 |
| thorst | so we had that dependency in the tox.ini and setup.cfg I think? | 21:27 |
| efried | uh, looking... | 21:27 |
| efried | Yeah. | 21:28 |
| thorst | it was in line with what some other neutron ml2 plugin was doing at the time | 21:28 |
| efried | One sec. | 21:28 |
| efried | thorst svenkat Like this: https://review.openstack.org/#/c/385486/ | 21:30 |
| thorst | correct | 21:30 |
| efried | We need to do that for ocata. svenkat may as well do it in this change set. | 21:30 |
| thorst | fair enough | 21:30 |
| thorst | I wonder if that will solve some of esberglu's issues with CI there | 21:30 |
| svenkat | efried lok… | 21:31 |
| efried | thorst Yeah, we need to do that for the other projects. | 21:32 |
| *** apearson has quit IRC | 21:32 | |
| svenkat | i will generate another review for 443960 . | 21:33 |
| *** apearson has joined #openstack-powervm | 21:35 | |
| esberglu | Just a heads up | 21:36 |
| esberglu | until | 21:36 |
| esberglu | https://review.openstack.org/#/c/443558/1 | 21:36 |
| esberglu | gets in ocata CI is failing | 21:36 |
| *** edmondsw has quit IRC | 22:03 | |
| svenkat | second patchset ready for https://review.openstack.org/#/c/443960/ | 22:15 |
| *** jpasqualetto has quit IRC | 22:19 | |
| svenkat | jenkins job went through except out-of-tree-pvm. Error: IOError: [Errno 2] No such file or directory: '/opt/stack/tempest/.tox/tempest/lib/python2.7/site-packages/appdirs-1.4.2.dist-info/METADATA' | 22:31 |
| svenkat | but py27 jobs are going through now. | 22:31 |
| esberglu | svenkat: Yep. https://review.openstack.org/#/c/443558/1 | 22:33 |
| *** edmondsw has joined #openstack-powervm | 22:33 | |
| svenkat | esberglu : thanks. | 22:33 |
| esberglu | Needs to merge for ocata CI to pass | 22:34 |
| openstackgerrit | Eric Berglund proposed openstack/nova-powervm master: Except HttpError 404 close_vterm https://review.openstack.org/444031 | 22:36 |
| *** edmondsw has quit IRC | 22:38 | |
| *** thorst has quit IRC | 22:39 | |
| *** apearson has quit IRC | 22:40 | |
| efried | esberglu Where's that pastebin with the 404 error? | 22:45 |
| esberglu | http://paste.openstack.org/show/602090/ | 22:46 |
| efried | Aha. | 22:46 |
| efried | Okay, so I don't agree with your solution. | 22:46 |
| efried | We're already trapping the 404 at the driver level, which is as it should be. | 22:46 |
| efried | The problem is, we're trapping it based on a specific REST path, which is horrible. | 22:47 |
| efried | See nova_powervm/virt/powervm/driver.py:685 | 22:47 |
| efried | Arguably, based on the logic in the other methods in that module (e.g. power_on and power_off), we should be trapping and ignoring the 404 where you're doing it, and NOT at the driver level at all. | 22:50 |
| efried | ...which would mean your change isn't necessarily wrong, just not the complete solution. | 22:51 |
| esberglu | Okay so you're proposing I remove the 404 from the driver level. And then add an exception for 404 on the lpar delete as well? | 22:59 |
| esberglu | At the vm level | 23:00 |
| *** apearson has joined #openstack-powervm | 23:01 | |
| *** dwayne has quit IRC | 23:01 | |
| efried | esberglu Yeah, something like that. | 23:02 |
| efried | Wouldn't mind getting thorst's nod, as that's a nontrivial change. | 23:02 |
| efried | But it should be okay. We only use dlt_lpar in two places: the destroy flow, and the spawn revert. | 23:02 |
| efried | In both cases, I'm pretty sure we don't want to bubble up if the LPAR isn't there. | 23:02 |
| efried | BUT | 23:02 |
| *** thorst has joined #openstack-powervm | 23:03 | |
| efried | We should first make sure that 404 on the CloseVTerm URI really means the LPAR ain't there. | 23:03 |
| efried | That should be easy enough to reproduce, I should think. | 23:04 |
| *** thorst has quit IRC | 23:07 | |
| *** tjakobs has quit IRC | 23:12 | |
| *** esberglu has quit IRC | 23:21 | |
| *** esberglu has joined #openstack-powervm | 23:33 | |
| *** esberglu has quit IRC | 23:37 | |
| *** tblakes has quit IRC | 23:46 | |
| *** seroyer has quit IRC | 23:48 | |
| *** apearson has quit IRC | 23:50 | |
| *** dwayne has joined #openstack-powervm | 23:59 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!