| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-01-18 | |||
| 13:13:16 | bauzas | ildikov: +Wd | |
| 13:13:22 | bauzas | looking at the 3rd in the series | |
| 13:13:36 | ildikov | bauzas: thank you!!! | |
| 13:14:27 | efried | jaypipes Is there a plan in the future to implement any kind of allocation healing/discovery, whereby we try to validate allocations in the placement db against the VMs nova actually has deployed? | |
| 13:14:58 | jaypipes | efried: no plan, no | |
| 13:15:03 | efried | thx | |
| 13:15:16 | bauzas | efried: WDYM ? | |
| 13:15:19 | bauzas | self-healing ? | |
| 13:15:36 | bauzas | that should be done by computes, nope ? | |
| 13:15:40 | efried | bauzas Yeah, basically anything to make sure the placement DB is in fact in sync with what's going on IRL | |
| 13:15:42 | efried | bauzas No | |
| 13:15:55 | efried | bauzas The computes get to report info to sync the *inventory* | |
| 13:16:00 | efried | bauzas Not the *allocations* | |
| 13:16:19 | bauzas | efried: but the compute nodes know if the instances are good or not | |
| 13:16:52 | efried | bauzas But other than e.g. removing allocations for destroyed instances, they don't sync up the placement DB. | |
| 13:17:05 | efried | bauzas If what you're saying is that the two "should" never get out of sync... | |
| 13:17:16 | efried | bauzas ...that may be what we're banking on :) | |
| 13:17:43 | bauzas | efried: so, if it wouldn't be the compute service which would look at all the instances, which service should be doing that ? | |
| 13:18:03 | efried | mdbooth At least enough to open the bug | |
| 13:18:08 | ameeda | jaypipes: thank you sir, I understand that :) | |
| 13:18:12 | efried | mdbooth A new bug, that is | |
| 13:18:25 | mdbooth | efried: Yep | |
| 13:18:26 | efried | bauzas No, I agree it should be the compute service. | |
| 13:18:40 | efried | bauzas Point is, we have no plans for the compute service to do that. | |
| 13:19:03 | bauzas | anyway, let's discuss that in the PTG | |
| 13:19:08 | efried | bauzas (That said, the compute service is only responsible for the resource providers it's responsible for.) | |
| 13:19:34 | bauzas | efried: yup, and I agree with what you just said | |
| 13:19:44 | efried | Others, e.g. the ones controlled by neutron or a shared provider manager... | |
| 13:19:56 | efried | And when there's supposed to be coordination between/among such sources... | |
| 13:19:58 | bauzas | efried: if the self-healing is done by the compute service, then that service would only heal the related instances it has | |
| 13:20:01 | efried | I can see where stuff could get out of sync. | |
| 13:20:12 | efried | Though here's an interesting case: | |
| 13:20:24 | bauzas | efried: I'm not against that, I mean, only Neutron would know which ports are fine | |
| 13:20:34 | bauzas | or which routers are there | |
| 13:21:39 | efried | If I have some allocations, and then I go and change my allocation ratio, and then I delete some allocations, I assume they get deleted as multiplied by the *new* allocation ratio, which would certainly lead to the db being out of sync. jaypipes am I missing something here? | |
| 13:22:24 | jaypipes | efried: yes, you are missing the fact that when you delete an allocation, there's nothing to do with the allocation ratio of the inventory. | |
| 13:22:50 | efried | I'll think it through again some time when there aren't more important things going on :) | |
| 13:23:00 | jaypipes | efried: allocation ratio is only used as a constraint at claim time. | |
| 13:23:23 | jaypipes | efried: so if you change it, it will affect claim decisions in the future after the change | |
| 13:23:32 | jaypipes | efried: but will have no effect on existing allocations | |
| 13:23:47 | efried | ack | |
| 13:24:02 | jaypipes | heh | |
| 13:24:41 | jaypipes | efried: if you could get rid of this flu for me, that would be appreciate too :( still going after 11 days :( | |
| 13:24:56 | efried | aw dude, that sucks. | |
| 13:25:14 | efried | You think you've got it rough, it's 26 degrees here. | |
| 13:25:22 | bauzas | jaypipes: seriously ? | |
| 13:25:29 | bauzas | man, change your doctor | |
| 13:25:39 | efried | Yeah, get you some Tamiflu | |
| 13:25:59 | bauzas | 26 degrees is excellent by a winter time, here | |
| 13:26:14 | bauzas | of course, it's just not the same degrees that you folks :p | |
| 13:26:19 | bauzas | than* | |
| 13:26:38 | jaypipes | efried: it's only 36 degrees here :( | |
| 13:26:47 | efried | Unacceptable. | |
| 13:26:56 | efried | Anything below 45 makes me grouchy. | |
| 13:27:01 | jaypipes | indeed. | |
| 13:27:07 | jaypipes | grrrroaaarrr. | |
| 13:46:24 | Spazmotic | Flu is spreading out the workplace like a wildfire.. I am going to Korea in 2 weeks so have been working from home to isolate myself. | |
| 13:47:54 | efried | FYI, added nova to https://wiki.openstack.org/wiki/PTG/Rocky/Etherpads | |
| 13:49:04 | jaypipes | efried: hold off responding to my review comments for at least half an hour. | |
| 13:49:15 | efried | jaypipes That bad? | |
| 13:49:26 | efried | or because you're queueing up other stuff? | |
| 13:49:27 | jaypipes | efried: I'm saving my comments on higher patches as I go along while doing a Reply comment on the bottom patch ... | |
| 13:49:34 | efried | gotcha. | |
| 13:50:15 | jaypipes | efried: doing it this way because otherwise I will end up with about 50 patchsets with draft review comments all just sitting around and not saved... | |
| 13:50:28 | efried | I dig it | |
| 13:50:35 | jaypipes | you speak jive? | |
| 13:51:00 | jaypipes | Jive ass dude don't got no brains anyhow. | |
| 13:51:09 | jaypipes | ok, back to my review... | |
| 13:51:18 | jaypipes | enough airplane! for one day | |
| 13:51:43 | Spazmotic | I need to find a new bug to fix while I wait on this one.. going stircrazy. I really need to bust out of stupid xenapi | |
| 13:52:52 | jaypipes | Spazmotic: efried can point you at a whole list of placement bugs we could use help on :) | |
| 13:53:23 | Spazmotic | Haha i'm not super good at.. things. I'm not sure how much of the insanity I see him commit that I could keep up with :D | |
| 13:53:35 | efried | Wow | |
| 13:54:07 | efried | Spazmotic If you're so inclined: https://bugs.launchpad.net/nova/+bugs?field.tag=placement | |
| 13:54:08 | jaypipes | hehe | |
| 13:54:13 | Spazmotic | Yeah sure i'll definately take a look | |
| 13:54:24 | jaypipes | I agree efried is indeed insane. | |
| 13:54:24 | Spazmotic | Might at least introduce me to new section of code | |
| 14:00:11 | stephenfin | jaypipes: Could you send this through? https://review.openstack.org/#/c/534712 | |
| 14:05:11 | jaypipes | stephenfin: done | |
| 14:06:21 | stephenfin | Cheers! | |
| 14:06:53 | Spazmotic | Never looked at this API before since we don't really use it.. I like it.. and I wanna look closer at this local delete issue. | |
| 14:07:15 | Spazmotic | An issue we face a lot in Xen land with xen falling on its face and landing wrong on the curb and never getting back up heh | |
| 14:08:04 | Spazmotic | Will start reviewing the code when I wake up.. time for sleep, thanks for the idea, gentlemen :D | |
| 14:09:21 | jaypipes | Spazmotic: oh, trust me... you do use it. you just may not realize it :) | |
| 14:20:10 | openstackgerrit | Aditya Vaja proposed openstack/nova master: remove IVS plug/unplug as they're moved to separate plugin https://review.openstack.org/534371 | |
| 14:21:41 | Spazmotic | Yeah I'm assuming that is where we are storing our allocation locking, just never thought about it before hehehehe | |
| 14:21:48 | Spazmotic | Excited to look at it. | |
| 14:24:59 | jaypipes | efried: keep bearing with me... | |
| 14:25:46 | efried | jaypipes Still working through my morning review mail. | |
| 14:25:54 | jaypipes | k | |
| 14:33:40 | Spazmotic | Stuck reading this provider code now rather than sleeping | |
| 14:33:41 | Spazmotic | Doomed | |
| 14:33:50 | Spazmotic | Pretty interesting and clean | |
| 14:34:31 | efried | Spazmotic Are you reading the database side? | |
| 14:34:38 | Spazmotic | API side currently | |
| 14:34:53 | Spazmotic | tracing down some calls to get to that point :D | |
| 14:35:23 | efried | Spazmotic If you're a sql wiz, we could sure use some help on that side of things. | |
| 14:35:40 | efried | If jaypipes gets hit by a bus... ahem, wins the lottery, we're effed. | |
| 14:35:59 | Spazmotic | I don't think i'm enough of a wiz to offer what you need..but some of these bugs I think I could at least learn and join the discussion. | |