| *** Ashana has joined #openstack-powervm | 00:00 | |
| *** Ashana has quit IRC | 00:05 | |
| *** Ashana has joined #openstack-powervm | 00:06 | |
| *** Ashana has quit IRC | 00:11 | |
| *** Ashana has joined #openstack-powervm | 00:12 | |
| *** seroyer has joined #openstack-powervm | 00:14 | |
| *** Ashana has quit IRC | 00:16 | |
| *** Ashana has joined #openstack-powervm | 00:18 | |
| *** Ashana has quit IRC | 00:22 | |
| *** Ashana has joined #openstack-powervm | 00:24 | |
| *** thorst has joined #openstack-powervm | 00:25 | |
| *** Ashana has quit IRC | 00:29 | |
| *** Ashana has joined #openstack-powervm | 00:30 | |
| *** thorst has quit IRC | 00:32 | |
| *** Ashana has quit IRC | 00:34 | |
| *** Ashana has joined #openstack-powervm | 00:37 | |
| *** Ashana has quit IRC | 00:41 | |
| *** Ashana has joined #openstack-powervm | 00:43 | |
| *** Ashana has quit IRC | 00:47 | |
| *** Ashana has joined #openstack-powervm | 00:49 | |
| *** thorst has joined #openstack-powervm | 00:49 | |
| *** Ashana has quit IRC | 00:53 | |
| *** Ashana has joined #openstack-powervm | 00:57 | |
| *** Ashana has quit IRC | 01:02 | |
| *** Ashana has joined #openstack-powervm | 01:03 | |
| openstackgerrit | Drew Thorstensen proposed openstack/nova-powervm: Driver cleanup work https://review.openstack.org/338573 | 01:03 |
|---|---|---|
| *** tlian2 has quit IRC | 01:04 | |
| *** Ashana has quit IRC | 01:08 | |
| *** Ashana has joined #openstack-powervm | 01:09 | |
| openstackgerrit | Drew Thorstensen proposed openstack/nova-powervm: Driver cleanup work https://review.openstack.org/338573 | 01:09 |
| thorst | efried: 338573 and 338536 when you get a chance....hoping to simplify that code. Been bugging me for a while. | 01:11 |
| *** thorst has quit IRC | 01:12 | |
| *** thorst has joined #openstack-powervm | 01:13 | |
| *** Ashana has quit IRC | 01:13 | |
| *** tlian has joined #openstack-powervm | 01:14 | |
| *** Ashana has joined #openstack-powervm | 01:15 | |
| *** Ashana has quit IRC | 01:19 | |
| *** Ashana has joined #openstack-powervm | 01:21 | |
| *** thorst has quit IRC | 01:22 | |
| *** Ashana has quit IRC | 01:25 | |
| *** Ashana has joined #openstack-powervm | 01:26 | |
| *** Ashana has quit IRC | 01:31 | |
| *** arnoldje has joined #openstack-powervm | 01:55 | |
| *** thorst has joined #openstack-powervm | 02:22 | |
| *** arnoldje has quit IRC | 02:26 | |
| *** kriskend has joined #openstack-powervm | 02:26 | |
| *** thorst has quit IRC | 02:27 | |
| *** thorst has joined #openstack-powervm | 02:34 | |
| *** thorst has quit IRC | 02:34 | |
| *** kriskend has quit IRC | 02:45 | |
| *** seroyer has quit IRC | 03:05 | |
| *** ManojK has joined #openstack-powervm | 03:22 | |
| *** jwcroppe has quit IRC | 03:24 | |
| *** jwcroppe has joined #openstack-powervm | 03:24 | |
| *** jwcroppe has quit IRC | 03:29 | |
| *** tlian has quit IRC | 03:33 | |
| *** thorst has joined #openstack-powervm | 03:34 | |
| *** thorst has quit IRC | 03:43 | |
| *** ManojK has quit IRC | 04:01 | |
| *** tjakobs has joined #openstack-powervm | 04:05 | |
| *** tjakobs has quit IRC | 04:07 | |
| *** jwcroppe has joined #openstack-powervm | 04:26 | |
| *** jwcroppe has quit IRC | 04:28 | |
| *** thorst has joined #openstack-powervm | 04:41 | |
| *** thorst has quit IRC | 04:48 | |
| *** tlian has joined #openstack-powervm | 05:35 | |
| *** tlian has quit IRC | 05:43 | |
| *** thorst has joined #openstack-powervm | 05:47 | |
| *** Ashana has joined #openstack-powervm | 05:48 | |
| *** Ashana has quit IRC | 05:53 | |
| *** thorst has quit IRC | 05:54 | |
| *** Ashana has joined #openstack-powervm | 05:54 | |
| *** Ashana has quit IRC | 05:59 | |
| *** Ashana has joined #openstack-powervm | 06:00 | |
| *** Ashana has quit IRC | 06:04 | |
| *** Ashana has joined #openstack-powervm | 06:06 | |
| *** jwcroppe has joined #openstack-powervm | 06:09 | |
| *** Ashana has quit IRC | 06:10 | |
| *** Ashana has joined #openstack-powervm | 06:12 | |
| *** Ashana has quit IRC | 06:16 | |
| *** Ashana has joined #openstack-powervm | 06:17 | |
| *** Ashana has quit IRC | 06:22 | |
| *** Ashana has joined #openstack-powervm | 06:23 | |
| *** jwcroppe has quit IRC | 06:24 | |
| *** Ashana has quit IRC | 06:28 | |
| *** Ashana has joined #openstack-powervm | 06:29 | |
| *** Ashana has quit IRC | 06:33 | |
| *** Ashana has joined #openstack-powervm | 06:35 | |
| *** Ashana has quit IRC | 06:39 | |
| *** Ashana has joined #openstack-powervm | 06:41 | |
| *** Ashana has quit IRC | 06:45 | |
| *** Ashana has joined #openstack-powervm | 06:47 | |
| *** Ashana has quit IRC | 06:51 | |
| *** thorst has joined #openstack-powervm | 06:52 | |
| *** Ashana has joined #openstack-powervm | 06:53 | |
| *** Ashana has quit IRC | 06:57 | |
| *** Ashana has joined #openstack-powervm | 06:59 | |
| *** thorst has quit IRC | 06:59 | |
| *** Ashana has quit IRC | 07:03 | |
| *** Ashana has joined #openstack-powervm | 07:04 | |
| *** Ashana has quit IRC | 07:09 | |
| *** Ashana has joined #openstack-powervm | 07:10 | |
| *** Ashana has quit IRC | 07:15 | |
| *** Ashana has joined #openstack-powervm | 07:16 | |
| *** Ashana has quit IRC | 07:21 | |
| *** Ashana has joined #openstack-powervm | 07:22 | |
| *** Ashana has quit IRC | 07:27 | |
| *** Ashana has joined #openstack-powervm | 07:28 | |
| *** jwcroppe has joined #openstack-powervm | 07:32 | |
| *** Ashana has quit IRC | 07:33 | |
| *** k0da has joined #openstack-powervm | 07:33 | |
| *** Ashana has joined #openstack-powervm | 07:34 | |
| *** Ashana has quit IRC | 07:39 | |
| *** Ashana has joined #openstack-powervm | 07:41 | |
| *** Ashana has quit IRC | 07:45 | |
| *** Ashana has joined #openstack-powervm | 07:47 | |
| *** Ashana has quit IRC | 07:51 | |
| *** Ashana has joined #openstack-powervm | 07:52 | |
| *** thorst has joined #openstack-powervm | 07:55 | |
| *** Ashana has quit IRC | 07:57 | |
| *** Ashana has joined #openstack-powervm | 07:58 | |
| *** Ashana has quit IRC | 08:03 | |
| *** thorst has quit IRC | 08:03 | |
| *** Ashana has joined #openstack-powervm | 08:04 | |
| *** jwcroppe has quit IRC | 08:05 | |
| *** Ashana has quit IRC | 08:09 | |
| *** Ashana has joined #openstack-powervm | 08:10 | |
| *** Ashana has quit IRC | 08:15 | |
| *** Ashana has joined #openstack-powervm | 08:16 | |
| *** openstackgerrit has quit IRC | 08:18 | |
| *** openstackgerrit has joined #openstack-powervm | 08:18 | |
| *** Ashana has quit IRC | 08:20 | |
| *** Ashana has joined #openstack-powervm | 08:25 | |
| *** Ashana has quit IRC | 08:29 | |
| *** Ashana has joined #openstack-powervm | 08:31 | |
| *** Ashana has quit IRC | 08:35 | |
| *** jwcroppe has joined #openstack-powervm | 08:35 | |
| *** jwcroppe has quit IRC | 08:37 | |
| *** jwcroppe has joined #openstack-powervm | 08:41 | |
| *** thorst has joined #openstack-powervm | 09:01 | |
| *** jwcroppe has quit IRC | 09:04 | |
| *** thorst has quit IRC | 09:09 | |
| *** thorst has joined #openstack-powervm | 10:07 | |
| *** thorst has quit IRC | 10:14 | |
| *** jwcroppe has joined #openstack-powervm | 10:26 | |
| *** jwcroppe has quit IRC | 10:27 | |
| *** jwcroppe has joined #openstack-powervm | 10:58 | |
| *** thorst has joined #openstack-powervm | 11:06 | |
| *** jwcroppe has quit IRC | 11:30 | |
| *** thorst_ has joined #openstack-powervm | 11:45 | |
| *** thorst has quit IRC | 11:49 | |
| *** jwcroppe has joined #openstack-powervm | 11:49 | |
| *** jwcroppe has quit IRC | 11:54 | |
| *** thorst_ is now known as thorst | 12:02 | |
| *** seroyer has joined #openstack-powervm | 12:20 | |
| *** mdrabe has joined #openstack-powervm | 12:35 | |
| *** Ashana has joined #openstack-powervm | 12:35 | |
| *** jwcroppe has joined #openstack-powervm | 12:52 | |
| *** jwcroppe has quit IRC | 12:57 | |
| *** ManojK has joined #openstack-powervm | 13:06 | |
| thorst | efried adreznec: Any chance I could get you to look over 338573 and 338536 quick? | 13:10 |
| *** tblakeslee has joined #openstack-powervm | 13:16 | |
| efried | thorst, on it. | 13:25 |
| thorst | thx dude | 13:25 |
| efried | Any particular order? | 13:26 |
| thorst | 536 first | 13:26 |
| thorst | 573 depends on 536 | 13:26 |
| efried | thorst, 536 atcha | 13:37 |
| openstackgerrit | Shyama proposed openstack/nova-powervm: Save valid udid on pre_live_migration of vscsi volumes using multi VIOS https://review.openstack.org/336783 | 13:46 |
| *** esberglu has joined #openstack-powervm | 13:48 | |
| *** lmtaylor1 has joined #openstack-powervm | 13:52 | |
| efried | thorst, 573 atcha | 13:53 |
| openstackgerrit | Drew Thorstensen proposed openstack/nova-powervm: Refactor validate vopt media repo to pypowervm https://review.openstack.org/338536 | 13:58 |
| thorst | efried: on 573 - I plan to do more in phases. More of the host_uuid can be removed in a lot of places. I guess this one isn't too big yet so I can keep iterating on that one | 14:00 |
| efried | cool. | 14:00 |
| efried | Just saw you get rid of it in one place, thought it would be cool to get rid of it everywhere while yer at it. | 14:01 |
| *** apearson has joined #openstack-powervm | 14:04 | |
| thorst | don't think we can get rid of it everywhere unfortunately | 14:07 |
| thorst | some things in pypowervm still take it as input | 14:07 |
| thorst | and that's a trickier thing to change... | 14:07 |
| efried | thorst, the things in pypowervm that take it don't use it, do they? | 14:10 |
| efried | And if they do, we should get rid of that too. | 14:10 |
| *** ManojK has quit IRC | 14:11 | |
| thorst | efried: I'm not necessarily asserting that the things in pypowervm need them | 14:18 |
| thorst | just that its harder to change the API there. | 14:18 |
| thorst | though we could just make that a legacy variable that could be None. | 14:18 |
| efried | Easy to pass None, though | 14:18 |
| efried | Yuh | 14:18 |
| thorst | so we can/should identify those | 14:18 |
| thorst | but getting through your comments first | 14:18 |
| efried | Yeah, I'm interested to see what in pypowervm still takes a host_uuid. | 14:18 |
| efried | I hope it's minimal. | 14:18 |
| thorst | some of the jobs | 14:19 |
| efried | And if it's not minimal, I'm going to be making it my personal mission to squash those like cockroaches. | 14:19 |
| openstackgerrit | Drew Thorstensen proposed openstack/nova-powervm: Driver cleanup work https://review.openstack.org/338573 | 14:20 |
| thorst | efried: want to knock out 536 so that my dependency chain is nicer? | 14:20 |
| efried | stand by | 14:20 |
| efried | thorst, +2 | 14:21 |
| thorst | yay! | 14:21 |
| thorst | I did something today | 14:21 |
| thorst | lol | 14:21 |
| thorst | efried: the build_vscsi_mapping for instance takes in a host_uuid | 14:23 |
| thorst | and I'm sure I'm the one that put it in there too. | 14:23 |
| efried | thorst, well, I'll be ripping that apart when we have VSCSI mappings as first-class objects, and host_uuid won't even be *possible* to use since the ROOT will have to be the VIOS. | 14:24 |
| thorst | efried: makes sense...but we're in transition period for the time being | 14:25 |
| thorst | so we'll have to keep host_uuid around in some places | 14:25 |
| thorst | but...definitely can minimize | 14:25 |
| efried | nod | 14:25 |
| *** kriskend has joined #openstack-powervm | 14:26 | |
| efried | thorst, 573 +2; one comment you can address if you wish. Though the issue has become more and more embarrassing for me. | 14:27 |
| thorst | efried: if it is about xags... | 14:28 |
| efried | heh, no. | 14:28 |
| efried | Not THAT embarrassing. | 14:28 |
| thorst | efried: I'm going to change them back to setup and then update the doc string | 14:29 |
| thorst | unless you prefer set_up? | 14:29 |
| openstackgerrit | Merged openstack/nova-powervm: Refactor validate vopt media repo to pypowervm https://review.openstack.org/338536 | 14:29 |
| efried | thorst, your call. I can certainly tolerate the method name either way. | 14:29 |
| thorst | k | 14:29 |
| *** ManojK has joined #openstack-powervm | 14:30 | |
| thorst | efried: looks like we should be investigating python 3.5 as well. | 14:31 |
| efried | thorst, what brings that up? | 14:34 |
| *** tjakobs has joined #openstack-powervm | 14:34 | |
| efried | Has core openstack gotten anything working in py3? | 14:34 |
| efried | Last I heard, you couldn't even tox the thing. | 14:34 |
| adreznec | efried: Most projects actually support it now... but not Nova or Swift | 14:35 |
| adreznec | Or one other I can't remember | 14:35 |
| efried | Oh, well, who needs those? | 14:35 |
| adreznec | Well, most of the main ones | 14:35 |
| adreznec | Right? | 14:35 |
| openstackgerrit | Shyama proposed openstack/nova-powervm: Save valid udid on pre_live_migration of vscsi volumes using multi VIOS https://review.openstack.org/336783 | 14:35 |
| thorst | efried: They have a non-voting job going for it now | 14:36 |
| thorst | its passing for us | 14:36 |
| thorst | but would be good to officially add it | 14:36 |
| thorst | and I think ubuntu 16.04 is what causes it. They ship python 3.5 by default. | 14:37 |
| adreznec | thorst: Well that and py34 isn't going to get fixes beyond this year iirc | 14:37 |
| *** seroyer has quit IRC | 14:37 | |
| thorst | efried: a lot of the stuff in pypowervm's cna.py also has host_uuid... Not as sure we can rip easily from there. | 14:38 |
| thorst | maybe we can... | 14:38 |
| thorst | not sure | 14:38 |
| efried | CNA is a child of the LPAR. | 14:39 |
| efried | So we again should be *unable* to use host UUID. | 14:39 |
| thorst | not just CNAs though...vNETs | 14:39 |
| efried | ah, where vnet is actually a CHILD of the managed system. That makes more sense. | 14:39 |
| efried | For those, a host_uuid param would be appropriate insofar as it would allow us to save ManagedSystem GETs if the caller already has that cached. | 14:40 |
| efried | But are vnets also ROOTs these days in NovaLink? apearson | 14:40 |
| *** burgerk has joined #openstack-powervm | 14:43 | |
| thorst | efried: just for you, I got rid of several more host_uuids that were completely unnecessary | 14:51 |
| thorst | though, before we merge this, I'd like to run it through our CI | 14:51 |
| thorst | esberglu: Do you think we could get the CI running with just neo38/39 for now? | 14:51 |
| thorst | while the memory upgrades are running on the others? | 14:51 |
| openstackgerrit | Drew Thorstensen proposed openstack/nova-powervm: Driver cleanup work https://review.openstack.org/338573 | 14:52 |
| esberglu | thorst: Yeah. Any update on the upgrades? Still waiting on access | 14:52 |
| *** jwcroppe has joined #openstack-powervm | 14:54 | |
| *** seroyer has joined #openstack-powervm | 14:55 | |
| thorst | esberglu: nothing on the upgrades yet. I suspect its a slow week due to July 4th. A lot of people return on the 11th... | 14:56 |
| thorst | but neo38/39 won't be getting any upgrades...so we're fine there. They're already max'd | 14:56 |
| *** ManojK has quit IRC | 14:58 | |
| *** jwcroppe has quit IRC | 15:00 | |
| *** svenkat has joined #openstack-powervm | 15:00 | |
| *** ManojK has joined #openstack-powervm | 15:01 | |
| svenkat | thorst_, efried: this is about comments in nova-powervm sriov bp : https://review.openstack.org/#/c/322203/7/specs/newton/powervm-sriov-nova.rst | 15:01 |
| svenkat | comment 1 is about compute driver changes. i am aware of vif driver invocation changes only. | 15:02 |
| svenkat | and comment 2 is about discussion on redundancy | 15:02 |
| apearson | @efried - no, ethernet adapters are always child objects - not root obbjects. | 15:04 |
| thorst | apearson: question was about vnets | 15:04 |
| thorst | not eth adapters | 15:04 |
| thorst | (I think) | 15:04 |
| thorst | svenkat: for comment 1, if its a vif driver change only - which I find a little suspect - then we can remove the bits about other driver changes. I guess all that whitelist stuff is part of the vif driver then? | 15:05 |
| svenkat | yes. hold on, waiting for Eric to join | 15:06 |
| *** efried has quit IRC | 15:06 | |
| *** efried has joined #openstack-powervm | 15:08 | |
| efried | thorst svenkat: ping? | 15:08 |
| svenkat | @efried, yes | 15:09 |
| efried | okay, now that THAT senseless delay is out of the way. | 15:09 |
| svenkat | ok!! | 15:09 |
| efried | I haven't seen anything since "but neo38/39 won't be getting upgrades..." | 15:10 |
| svenkat | so, the changes in compute driver will be specific to vif invocation to attach nic to vm. | 15:10 |
| efried | Right; I assumed that would entail changes in methods like spawn. | 15:10 |
| efried | Perhaps not. | 15:10 |
| efried | We've got to parse the configs to detect that we need to use the SRIOV VIF driver. Does that happen in nova driver init? | 15:11 |
| svenkat | spawn ends up invoking plug.. so i thioght all needed changes in plug. | 15:12 |
| thorst | svenkat efried: where is this 'whitelist' stuff encapsulated? | 15:12 |
| svenkat | similar to community code, which branches off to invoke various vif drivers in plug in vif driver | 15:12 |
| efried | Well, once we know we're using the SR-IOV VIF driver, I agree all that can be within the VIF driver itself. | 15:13 |
| thorst | efried: this is what svenkat is referring to: https://github.com/openstack/nova-powervm/blob/master/nova_powervm/virt/powervm/vif.py#L46-L48 | 15:13 |
| thorst | used by the method at line 60 | 15:13 |
| svenkat | yes. the high level driver level vif | 15:13 |
| svenkat | @thorst: to answer your question about whitelist, it can be part of vif parameter | 15:14 |
| thorst | can be or is a? | 15:14 |
| thorst | 'can be' implies you need to massage the data somehow | 15:14 |
| thorst | 'is part of' indicates it is already there... | 15:15 |
| svenkat | can be . unless we want to prepare this data at a higher level and refer to it in plug. | 15:15 |
| thorst | so I believe what you're saying is that the whitelist can be completely encapsulated within the vif plug. | 15:15 |
| thorst | and the 'vif' as a parameter is enough to work off of. | 15:15 |
| efried | thorst, I think I agree with that. | 15:16 |
| svenkat | thorst: yes. | 15:16 |
| thorst | ok | 15:16 |
| svenkat | contain all necessary data and logic in plug itself. | 15:16 |
| thorst | then I think that there aren't really any driver changes | 15:16 |
| svenkat | ok… | 15:16 |
| efried | So the first bullet in the blueprint (line 96) | 15:16 |
| svenkat | yes. | 15:17 |
| efried | should state that there are no compute driver changes, and provide the rationale as discussed above. | 15:17 |
| svenkat | ok... | 15:17 |
| svenkat | will do. | 15:17 |
| efried | cool. | 15:17 |
| efried | Now the fun one. | 15:17 |
| svenkat | for line 127 in bp, about redundancy | 15:17 |
| efried | thorst and I had some discussions about this | 15:17 |
| svenkat | we have a paragraph for redundancy. | 15:18 |
| svenkat | ok... | 15:18 |
| efried | And I can't freakin remember which chat service we were using, so it would take me a while to look it up. | 15:18 |
| svenkat | oh ok… np. | 15:18 |
| efried | But I do remember where we landed was that we would, at least initially, not expose any tunables for redundancy levels. | 15:18 |
| svenkat | ok… | 15:18 |
| thorst | 1 or 2... | 15:19 |
| thorst | that's my understanding | 15:19 |
| thorst | if you have 2 available, go redundant. | 15:19 |
| efried | We would implement a default policy that would figure out an appropriate min_redundancy and max_redundancy based on the number of cards, number of VIOSes, number of ports. | 15:19 |
| thorst | if you only have 1...don't use redundancy | 15:19 |
| thorst | or...make it a conf option. | 15:19 |
| thorst | I'm fine with either. | 15:19 |
| svenkat | why stop at 2, can we call 1 or more insetad of 1 or 2 | 15:19 |
| thorst | but I don't want to go crazy on the config. | 15:19 |
| thorst | you stop at 1 so you don't flood the ports. | 15:20 |
| efried | Right, the goal was to keep the config simple initially | 15:20 |
| thorst | I mean at 2. | 15:20 |
| efried | and only provide ability to tune if demanded by customer. | 15:20 |
| svenkat | ok… | 15:20 |
| thorst | If I have 100 VMs and 4 SR-IOV ports...I'd run out of ports super fast if I had them flooding all the ports | 15:20 |
| efried | We may be talking about redundancy on a couple of different levels here, thorst. | 15:20 |
| efried | First of all - seroyer, correct me if I got this wrong - but redundant vNIC is only going to be running over one VF at a time. | 15:21 |
| thorst | this is not the # of VFs backing a vNIC? | 15:21 |
| thorst | well...server vNICs | 15:21 |
| thorst | I know my terminology is a bit off... | 15:21 |
| efried | Right. I'm pretty sure they don't all run at once. I'm pretty sure only one is on at any given time. | 15:21 |
| efried | So we're not worried about flooding traffic. The closest we come there is worrying about running out of VFs. | 15:21 |
| thorst | efried: yeah, that's what I meant about flooding | 15:22 |
| thorst | even though only one is active at a time, the others are reserved and can't be used | 15:22 |
| thorst | you can't over commit them | 15:22 |
| efried | So let's talk max redundancy. | 15:22 |
| efried | With two (multi-port) cards and two VIOSes, I say we should have 4 backing VFs. | 15:23 |
| efried | That eliminates SPoF. | 15:23 |
| efried | Actually eliminates two PoF. | 15:23 |
| efried | Is that too much redundancy? | 15:23 |
| thorst | efried: Are you assuming two fabrics? | 15:23 |
| thorst | or a single fabric | 15:24 |
| efried | Shrug, that's outside our purview. | 15:24 |
| thorst | not necessarily. The reason FC does 4 ports is because they want two ports per fabric | 15:24 |
| efried | If they've got 2x2 ports hooked up to the same external net, it's reasonable to assume they've got switch redundancy. | 15:24 |
| efried | Remember, this is all for sure on the same external net. | 15:24 |
| efried | Otherwise they'd be different whitelist entries | 15:24 |
| thorst | I think we can make a case for 4. I doubt that it should ever be the default for max redundancy | 15:24 |
| efried | or they configured it wrong. | 15:24 |
| thorst | I think the desired is almost always going to be 2. | 15:24 |
| thorst | and 4 is going to be for your extremely paranoid workloads | 15:25 |
| thorst | and we shouldn't let the 1% of workloads dictate a desired default. | 15:25 |
| efried | thorst, not saying you're wrong, but upon what are you basing your assumption? | 15:25 |
| efried | "desired almost always 2" | 15:25 |
| thorst | SEA today is typically only a failover between two cards | 15:26 |
| efried | Across two VIOSes. | 15:26 |
| *** ManojK has quit IRC | 15:26 | |
| thorst | ports rarely fail. The reason they do two ports on a given SEA (within a VIOS) is for perf. I guess a port could fail...but I just don't see it making sense to default to 4 | 15:26 |
| thorst | also because we can only get like 20 VFs per physical port. | 15:26 |
| efried | Okay, so let's flip it around: do you see a need for us to *allow* the consumer to specify more than 2? | 15:27 |
| thorst | so if you default to 4...and you have two 4 port cards...you're getting a max of 40 workloads on that box | 15:27 |
| thorst | which is nothin | 15:27 |
| *** ManojK has joined #openstack-powervm | 15:28 | |
| efried | thorst, ^^ - assuming we default to max of 2, do you think we need a config option to increase? | 15:28 |
| *** k0da has quit IRC | 15:29 | |
| svenkat | it will be a good idea to drive max via a configuration | 15:29 |
| svenkat | gives us room to eliminate future code changes | 15:29 |
| efried | what about min? | 15:29 |
| svenkat | why not drive both via configuration changes | 15:29 |
| efried | By enforcing a min greater than 1, the user is saying they want deploy to fail if we can't achieve the desired minimum redundancy level. | 15:30 |
| efried | Is that a likely use case? | 15:30 |
| thorst | efried: I'm OK with it being handled via a configuration change. As noted earlier...I think at the OpenStack level its just a conf option of 'I want X VFs backing this thing' | 15:30 |
| thorst | no min...no max....just X | 15:30 |
| efried | svenkat, sure, we *can* do that, but I'm trying to minimize conf clutter. | 15:30 |
| efried | thorst, so we would set min = max = X? | 15:30 |
| svenkat | so one entry for both min and max in configuration | 15:31 |
| thorst | efried: that's what I'd start with | 15:31 |
| efried | (internally, as params to the anti-affinity algo) | 15:31 |
| thorst | you can always make things more complex later | 15:31 |
| svenkat | ok.. | 15:31 |
| efried | So this is exactly what svenkat has described in the blueprint. | 15:32 |
| svenkat | i will update BP with min/max configuration option | 15:32 |
| efried | No, I think this is exactly what you've already described. | 15:32 |
| svenkat | oh ok. got it… DRL | 15:32 |
| efried | Though we should talk about the conf option names. | 15:32 |
| svenkat | desired_redundnacy_level as mentioned in bp already | 15:33 |
| svenkat | desired_redundancy_level | 15:33 |
| efried | thorst, maybe I misread, did you hint that openstack already has a conf option for this? | 15:33 |
| thorst | efried: I don't think it does | 15:33 |
| efried | okay. | 15:33 |
| efried | So | 15:33 |
| thorst | svenkat: change the parameter to required_redundancy_level | 15:33 |
| svenkat | sure. | 15:33 |
| efried | Our name needs to be somehow specific to PowerVM and to SR-IOV | 15:34 |
| thorst | well, it'll be in the powervm section | 15:34 |
| thorst | so its specific that way | 15:34 |
| svenkat | ok… | 15:34 |
| efried | Okay, good. | 15:34 |
| efried | That should be specified | 15:34 |
| efried | In the blueprint | 15:34 |
| thorst | vnic_required_vf_count? | 15:34 |
| svenkat | ok.. | 15:34 |
| thorst | with a default of 2? | 15:34 |
| efried | vnic_required_vfs would be okay by me | 15:35 |
| efried | agree default 2 | 15:35 |
| svenkat | this is only if redundancy is needed.. if no redundancy is needed, this is not relevant.. did i understand this right | 15:35 |
| efried | no | 15:35 |
| efried | redundancy is implied by this setting. | 15:35 |
| efried | If you don't want redundancy, you set it explicitly to 1. | 15:35 |
| thorst | if they set it to 1...then no redundancy... | 15:36 |
| efried | There's no "redundancy on/off" switch. | 15:36 |
| svenkat | ok… | 15:36 |
| svenkat | i will describe it in bp for clarity | 15:36 |
| svenkat | ok… | 15:36 |
| svenkat | i will update bp and send it out for review soon. | 15:36 |
| efried | So to be clear, pci_passthrough_whitelist is an actual OpenStack setting. It goes in whatever section it goes in today. | 15:36 |
| svenkat | ok.. | 15:36 |
| efried | This new vnic_required_vfs will go in the [powervm] section. | 15:36 |
| svenkat | ok. | 15:37 |
| efried | Thanks, guys. | 15:37 |
| svenkat | thanks | 15:38 |
| thorst | rockin | 15:38 |
| thorst | thx | 15:38 |
| seroyer | efried: Yes, vNIC failover is “failover”. Active and some number of backups. Only one physical link is used at a time. | 15:43 |
| efried | thanks | 15:43 |
| *** dwayne_ has quit IRC | 15:49 | |
| *** ManojK has quit IRC | 16:04 | |
| *** ManojK has joined #openstack-powervm | 16:24 | |
| thorst | kriskend: got a sec? | 16:36 |
| thorst | https://review.openstack.org/#/c/339094/1/nova_powervm/virt/powervm/driver.py <-- line 592. Did you hit an error there when not using config drive? Or did you proactively comment that out? | 16:37 |
| kriskend | @thorst sure | 16:41 |
| kriskend | we commented that out and it fixed our problem | 16:42 |
| kriskend | the problem we hit was updating the VG | 16:42 |
| kriskend | cuz REST support is not there for that yet | 16:42 |
| kriskend | We did not make the nova.conf change to not use config drive | 16:43 |
| kriskend | Would you like us to do that instead??? | 16:43 |
| thorst | kriskend: I understand... | 16:44 |
| thorst | OK - I can work with that. Thx | 16:45 |
| kriskend | we are currently updating our code and going to try with the changes you have put up | 16:46 |
| kriskend | and the one Taylor put up | 16:46 |
| *** dwayne_ has joined #openstack-powervm | 16:48 | |
| thorst | tjakobs: I posted a suggestion on how to get that through sonar. | 16:48 |
| *** dwayne_ has quit IRC | 16:52 | |
| *** jwcroppe has joined #openstack-powervm | 16:58 | |
| *** dwayne_ has joined #openstack-powervm | 17:00 | |
| *** jwcroppe has quit IRC | 17:02 | |
| *** Ashana has quit IRC | 17:12 | |
| *** Ashana has joined #openstack-powervm | 17:14 | |
| *** Ashana_ has joined #openstack-powervm | 17:16 | |
| *** Ashana has quit IRC | 17:18 | |
| openstackgerrit | Drew Thorstensen proposed openstack/nova-powervm: Remove delete vopt if nothing to delete https://review.openstack.org/339151 | 17:22 |
| thorst | kriskend tjakobs: ^^ See that. I think if you set config drive off and use that...you should be good | 17:22 |
| *** catintheroof has joined #openstack-powervm | 17:53 | |
| *** jwcroppe has joined #openstack-powervm | 17:58 | |
| *** jwcroppe has quit IRC | 18:04 | |
| *** ManojK has quit IRC | 18:07 | |
| kriskend | thorst: We tried changing force_config_drive=True to False | 18:29 |
| kriskend | in nova.conf | 18:29 |
| kriskend | and pulled down the change you put up | 18:29 |
| kriskend | but we are still failing in media.py | 18:30 |
| kriskend | looking for the vopt | 18:30 |
| thorst | stack trace? | 18:30 |
| thorst | can you put it on pastebin or something? | 18:30 |
| tjakobs | thorst: http://pastebin.com/pzcvTd1P | 18:32 |
| thorst | https://github.com/openstack/nova-powervm/blob/master/nova_powervm/virt/powervm/driver.py#L442 | 18:33 |
| kriskend | it is that same issue where it is looking for rootvg | 18:33 |
| thorst | kriskend tjakobs: can you put a breakpoint in there and see why configdrive.required_by is returning true? | 18:33 |
| kriskend | sure | 18:34 |
| thorst | wait | 18:35 |
| thorst | stop | 18:35 |
| thorst | its not that.. | 18:35 |
| thorst | is this a spawn? | 18:35 |
| thorst | or a destroy? Cause that is definitely being called from destroy. | 18:35 |
| tjakobs | we are spawning | 18:36 |
| tjakobs | also just put a breakpoint where you wanted and configdrive.required_by(instance) return False | 18:36 |
| thorst | yeah... | 18:39 |
| thorst | so here's what's happening | 18:39 |
| thorst | 1) Something else in spawn is breaking | 18:39 |
| thorst | 2) It rolls back | 18:40 |
| thorst | 3) Destroy is invoked | 18:40 |
| thorst | 4) Destroy itself is failing | 18:40 |
| thorst | and thus the real error is hidden | 18:40 |
| thorst | the ole double failure hiding the real failure | 18:40 |
| thorst | tjakobs kriskend: I'll have a new patch for you to try out soon... | 18:41 |
| kriskend | yeah there is a different error before this | 18:42 |
| kriskend | that we need to look at | 18:43 |
| kriskend | oh this other one is also on the delete I think... it is in unplug vif | 18:44 |
| kriskend | looks like we have an issue there too | 18:44 |
| kriskend | but still not the real problem | 18:44 |
| tjakobs | thorst: here was the first error message that happened: http://pastebin.com/SXV1sSaT | 18:45 |
| thorst | eek...that's a good one. | 18:47 |
| *** ManojK has joined #openstack-powervm | 18:50 | |
| thorst | kriskend tjakobs: that one is in pypowervm...gross | 18:51 |
| openstackgerrit | Drew Thorstensen proposed openstack/nova-powervm: Remove delete vopt if nothing to delete https://review.openstack.org/339151 | 18:52 |
| thorst | you'll need that patch....getting you something for above as well | 18:52 |
| *** lmtaylor1 has left #openstack-powervm | 18:52 | |
| kriskend | yeah we are still trying to update a vg cuz we are back to getting a REST error trying to do it | 18:54 |
| kriskend | I think when we do this... | 18:56 |
| kriskend | 2016-07-07 13:37:52.260 ^[[00;36mINFO nova_powervm.virt.powervm.disk.localdisk [^[[01;36mreq-c978fd39-fcc2-4549-aa31-74b084058311 ^[[00;36madmin admin^[[00;36m] ^[[01;35m^[[00;36mCreate disk.^ | 18:56 |
| kriskend | [[00m | 18:56 |
| kriskend | ick | 18:56 |
| kriskend | <RequestURI kb="ROR" kxe="false">/rest/api/uom/VirtualIOServer/57DFF425-1FCB-4E2D-A7AE-022DAA684983/VolumeGroup/1ea243c7-08ef-375e-a0c7-8dbde3344d88</RequestURI> | 18:58 |
| kriskend | <ReasonCode kxe="false" kb="ROR">Unknown internal error.</ReasonCode> | 18:58 |
| kriskend | <Message kxe="false" kb="ROO">Current state of ResourceMonitoringControl does not allow to perform operation on VIOS with ID 1 in System 9119-MME*106CCC7 | 18:58 |
| thorst | kriskend: well, nonetheless... | 18:58 |
| thorst | these are good errors | 18:58 |
| kriskend | yeah | 18:58 |
| thorst | that other one needs to be taken up with changh | 18:58 |
| kriskend | Should we be able to create a disk on a lparHosting VG? | 18:59 |
| kriskend | I thought that should work... | 18:59 |
| tjakobs | thorst: tried your nova-powervm patch, no longer hitting the issue in that first pastebin | 18:59 |
| kriskend | adreznec ^ | 18:59 |
| thorst | kriskend: yes...you should | 19:02 |
| thorst | (at least *I* think you should be able to) | 19:03 |
| kriskend | Looks to me like there is an RMC check in that path | 19:03 |
| thorst | yeah, but in REST... | 19:03 |
| kriskend | that is keeping it from working in the VIOSless enviroment | 19:03 |
| kriskend | yep | 19:03 |
| thorst | so changh | 19:03 |
| thorst | or apearson | 19:03 |
| *** jwcroppe has joined #openstack-powervm | 19:05 | |
| *** jwcroppe has quit IRC | 19:05 | |
| *** jwcroppe has joined #openstack-powervm | 19:05 | |
| apearson | kriskend - so this volume group is on a linux VIOS? We'd need to get the FFDC log to see exactly where it was failing...@changh can help... | 19:12 |
| kriskend | yeah apearson it is | 19:12 |
| kriskend | switched over to Slack to discuss... since it is REST | 19:13 |
| *** k0da has joined #openstack-powervm | 19:24 | |
| *** mdrabe has quit IRC | 19:41 | |
| *** mdrabe has joined #openstack-powervm | 19:45 | |
| *** k0da has quit IRC | 19:50 | |
| *** jwcroppe has quit IRC | 20:03 | |
| *** apearson has quit IRC | 20:11 | |
| *** apearson has joined #openstack-powervm | 20:15 | |
| *** k0da has joined #openstack-powervm | 20:25 | |
| *** apearson has quit IRC | 20:30 | |
| *** Ashana has joined #openstack-powervm | 20:46 | |
| thorst | efried kriskend tjakobs: 3543 for the cna fix | 20:49 |
| *** Ashana_ has quit IRC | 20:49 | |
| *** Ashana has quit IRC | 20:50 | |
| efried | thorst, I really don't like that method. | 20:54 |
| efried | It's too overloaded. | 20:54 |
| thorst | _1 | 20:54 |
| thorst | +1 | 20:54 |
| efried | Where's the sonar rule for "ugly"? | 20:54 |
| thorst | I don't like that method either | 20:54 |
| thorst | tried to make it better | 20:54 |
| thorst | I guess the question is...did that method go into a release. | 20:54 |
| efried | It's in 1.0.0.3 | 20:55 |
| efried | But with totally different semantics. | 20:56 |
| efried | So having changed it thus is just as bad as completely refactoring it. | 20:56 |
| efried | Sorry, the completely different semantics is the same as what this patch was before you changed it. | 20:56 |
| efried | So I don't think what you've done here is better than ripping it apart and making it (probably into multiple methods that) make sense. | 20:57 |
| efried | thorst ^^ | 20:57 |
| efried | def get_partitions(adapter, get_lpars=True, get_vioses=True, get_mgmt=False): | 21:00 |
| efried | """Get a list of partitions possibly including LPARs, VIOSes, and the management partition. | 21:00 |
| efried | :param get_lpars: If True, the result will include all LPARs. | 21:00 |
| efried | :param get_vioses: If True, the result will include all VIOSes. | 21:00 |
| efried | :param get_mgmt: If True, the result is guaranteed to include the management partition, even if | 21:00 |
| efried | it would not otherwise have been included based on get_lpars/get_vioses. | 21:00 |
| efried | """ | 21:00 |
| efried | thorst something like ^^ | 21:00 |
| *** svenkat has quit IRC | 21:00 | |
| *** kriskend has quit IRC | 21:06 | |
| thorst | efried: let me see whats in 1.0.0.3... | 21:06 |
| thorst | cause I just assumed this weird semantic was there too... | 21:06 |
| efried | thorst, it's the same as this was before you changed it. | 21:07 |
| efried | My point is, you changed it from one weird semantic to a completely different weird semantic. | 21:07 |
| thorst | seeing is believing | 21:07 |
| *** apearson has joined #openstack-powervm | 21:07 | |
| thorst | just to be clear...we're talking get_all_lpars? | 21:08 |
| efried | y | 21:08 |
| thorst | I don't actually see any code using that... | 21:08 |
| thorst | except in pypowervm itself | 21:08 |
| thorst | given the grossness of the method...I'm pretty inclined to just change it | 21:09 |
| thorst | at a minimum, make it private. | 21:09 |
| thorst | thoughts? I know that's not ideal... | 21:09 |
| efried | thorst, I'm on board. | 21:10 |
| thorst | ripping it | 21:10 |
| efried | What do you think of above suggestion? | 21:10 |
| efried | ...plus making it private | 21:10 |
| thorst | yeah, I dig | 21:10 |
| efried | k | 21:10 |
| *** k0da has quit IRC | 21:12 | |
| efried | thorst, in other news, the options on SRIOV physical ports are way messed up in the schema. Considering forcing them to change it. | 21:14 |
| thorst | now is the time | 21:15 |
| thorst | efried: I put up a new patch. It won't pass test...but would like to see if you think that's the right approach | 21:19 |
| efried | thorst, beautiful. I think this is much clearer in all ways. | 21:21 |
| thorst | yeah, me too | 21:21 |
| thorst | lol - the unit tests still work | 21:24 |
| thorst | which do actually test that method I think | 21:24 |
| thorst | that's kinda hilarious | 21:24 |
| thorst | I will be making a new test though... | 21:24 |
| *** burgerk has quit IRC | 21:30 | |
| *** apearson has quit IRC | 21:31 | |
| *** apearson has joined #openstack-powervm | 21:37 | |
| thorst | efried: posted a new version up | 21:46 |
| efried | done | 21:50 |
| *** apearson has quit IRC | 21:52 | |
| thorst | esberglu: http://184.172.12.213/73/338573/5/check/nova-powervm-pvm-dsvm-tempest-full/885f2c9/powervm_os_ci.html | 21:54 |
| thorst | I'm not sure that those test_trunk things are valid for our networking-powervm agent | 21:54 |
| thorst | either that or our openstack config...not sure | 21:54 |
| thorst | and I think that they're leaving artifacts around that screw up the other tests... | 21:55 |
| esberglu | Okay. There’s new qos tests too that I’m removing right now | 21:55 |
| thorst | esberglu: yeah...I don't think we wipe them all out...but just a select few | 21:55 |
| thorst | trunks search criteria seems odd | 21:56 |
| thorst | we should figure out what a 'network trunk' is to openstack... | 21:56 |
| thorst | any takers? | 21:56 |
| esberglu | thorst: Wipe all the qos tests though right? | 21:57 |
| thorst | esberglu: yep | 21:57 |
| thorst | https://wiki.openstack.org/wiki/Neutron/TrunkPort | 21:57 |
| thorst | alright...I've got to head out | 22:00 |
| thorst | see ya | 22:00 |
| *** ManojK has quit IRC | 22:02 | |
| *** thorst has quit IRC | 22:03 | |
| *** ManojK has joined #openstack-powervm | 22:04 | |
| *** mdrabe has quit IRC | 22:08 | |
| *** ManojK has quit IRC | 22:09 | |
| *** tjakobs has quit IRC | 22:12 | |
| *** apearson has joined #openstack-powervm | 22:12 | |
| *** esberglu has quit IRC | 22:14 | |
| *** apearson has quit IRC | 22:16 | |
| *** apearson has joined #openstack-powervm | 22:19 | |
| *** tblakeslee has quit IRC | 22:35 | |
| *** seroyer has quit IRC | 22:36 | |
| *** catintheroof has quit IRC | 22:40 | |
| *** thorst has joined #openstack-powervm | 22:50 | |
| *** thorst has quit IRC | 22:54 | |
| *** ManojK has joined #openstack-powervm | 23:22 | |
| *** ManojK has quit IRC | 23:30 | |
| *** thorst has joined #openstack-powervm | 23:31 | |
| *** ManojK has joined #openstack-powervm | 23:33 | |
| *** thorst has quit IRC | 23:36 | |
| *** apearson has quit IRC | 23:48 | |
| *** thorst has joined #openstack-powervm | 23:52 | |
| *** thorst has quit IRC | 23:57 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!