| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-01-06 | |||
| 16:46:06 | mriedem | i don't know which change that is referring to in oslo.messaging | |
| 16:48:51 | melwitt | mriedem: ah, right. that TODO is invalid but there were specific changes in oslo.messaging to make it use [DEFAULT]/transport_url, so need to see if those are part of newton | |
| 16:49:18 | mriedem | not only part of newton, but probably part of the minimum required version of oslo.messaging in newton | |
| 16:49:32 | mriedem | to ensure that it works with the minimum package dependencies | |
| 16:49:41 | melwitt | right | |
| 16:52:58 | openstackgerrit | Julian proposed openstack/nova: Transform keypair.create notification https://review.openstack.org/408689 | |
| 17:03:07 | mriedem | BobBall: oh so close https://review.openstack.org/#/c/365505/ | |
| 17:06:24 | openstackgerrit | Stephen Finucane proposed openstack/nova: Ignore IOError when creating 'console.log' https://review.openstack.org/392643 | |
| 17:07:23 | sfinucan | sahid, vladikr: Any chance ye could have a look at ^ when you have a chance? | |
| 17:07:47 | vladikr | sfinucan, sure | |
| 17:08:15 | BobBall | mriedem: I think we're OK - aren't all API calls here synchronized on the VM object? | |
| 17:08:23 | BobBall | Let me find that... | |
| 17:08:26 | mriedem | BobBall: nope | |
| 17:08:35 | BobBall | I thought they were in the XenAPI driver | |
| 17:08:51 | mriedem | oh, well that might be news to me | |
| 17:09:06 | BobBall | Nope, it's not | |
| 17:09:16 | BobBall | I was sure they were at one point :D | |
| 17:09:55 | BobBall | Easy enough to add a utils.synchronized to the plug method though? | |
| 17:10:19 | mriedem | yeah | |
| 17:10:28 | mriedem | although | |
| 17:10:34 | mriedem | attach interface is an rpc call | |
| 17:10:48 | mriedem | but i guess if you lock on the instance then that should be ok | |
| 17:11:03 | BobBall | yeah - I meant on the instance. | |
| 17:11:07 | mriedem | hopefully 2 plugs for the same instance at the same time could complete in under 60 seconds | |
| 17:11:13 | leakypipes | dansmith: mind taking a look at https://review.openstack.org/#/c/413068/3/oslo_versionedobjects/base.py? I have a suspicion that doing that is not intended... | |
| 17:11:27 | BobBall | The synchronization I was thinking of is already there for VBD plug / unplug | |
| 17:11:35 | BobBall | oh yes, for sure. | |
| 17:12:24 | BobBall | Could probably just synchronize the creation (not hotplug) - once we have the record in the XenServer DB it will not announce that as an allowed device | |
| 17:12:43 | dansmith | leakypipes: yep, you're right | |
| 17:13:03 | leakypipes | phew. | |
| 17:15:54 | openstackgerrit | Matt Riedemann proposed openstack/nova: Fix TypeError in _update_from_compute_node race https://review.openstack.org/416822 | |
| 17:20:54 | dansmith | mriedem: what do I do for this? use |
|
| 17:22:42 | dansmith | or %(int)s I guess | |
| 17:24:21 | dansmith | hmm, nope | |
| 17:24:40 | leakypipes | dansmith: couldn't you just do: \d+ | |
| 17:24:59 | leakypipes | BTW, is superdan in hiding? ;) | |
| 17:24:59 | dansmith | leakypipes: yes, but just trying to figure out how to use the integrated one he suggested | |
| 17:25:10 | dansmith | oh jeez is it friday? | |
| 17:25:14 | leakypipes | sure is! | |
| 17:25:16 | leakypipes | :) | |
| 17:25:19 | leakypipes | there he is! | |
| 17:25:29 | leakypipes | to save the day. | |
| 17:25:38 | sean-k-mooney | leakypipes: hi jay would you have time to review sfinucan mtu patch to os-vif https://review.openstack.org/#/c/408779 | |
| 17:25:40 | bauzas | superdan: leakypipes: mriedem: folks, I have an important question for you | |
| 17:25:48 | bauzas | oh man, my manners | |
| 17:25:55 | leakypipes | sean-k-mooney: yup, will do. thx for the heads up. | |
| 17:26:37 | leakypipes | superdan: yeah, looks like %(int)s is correct there. | |
| 17:26:45 | bauwser | superdan: leakypipes: mriedem: so, since I wanted to write the scheduler-side of the placement call for getting the list of RPs, I thought about the fact we wrote the methods within the client.report module | |
| 17:26:53 | superdan | leakypipes: doesn't work | |
| 17:27:22 | bauwser | superdan: leakypipes: mriedem: that's why I began to cut that from the module and rather create a new module with a PlacementClient | |
| 17:27:38 | leakypipes | sdague: perhaps you might know the answer to superdan's dilemma? | |
| 17:27:50 | bauwser | superdan: leakypipes: mriedem: but given it will need some tests modifications, that change will be huge to review | |
| 17:28:05 | superdan | sdague: I'm trying (desperately) to run devstack locally with cellsv1 and I keep getting use_neutron=True, despite using the ENABLED_SERVICES list from the gate run | |
| 17:28:16 | superdan | sdague: any idea if there's something else that will trigger neutron? I need n-net for cellsv1 | |
| 17:28:26 | bauwser | superdan: leakypipes: mriedem: do you guys think we should maybe just use the existing and cut the placementclient afterwards?N | |
| 17:28:40 | leakypipes | bauwser: yes. | |
| 17:29:12 | bauwser | okay, I'll put my existing modifications in a WIP change and -2 it | |
| 17:29:52 | bauwser | tbh, I'm a bit afraid of the time flying | |
| 17:29:57 | superdan | sdague: hmm, maybe the interface/network configs | |
| 17:31:47 | openstackgerrit | Dan Smith proposed openstack/nova: Make the SingleCellSimple fixture a little more comprehensive https://review.openstack.org/410472 | |
| 17:31:47 | openstackgerrit | Dan Smith proposed openstack/nova: Fix non-parameterized service id in hypervisors sample tests https://review.openstack.org/410471 | |
| 17:32:42 | openstackgerrit | Gábor Antal proposed openstack/nova: Transform instance.reboot notifications https://review.openstack.org/382959 | |
| 17:37:40 | melwitt | mriedem: so newton requires oslo.messaging>=5.2.0 and all of the rpc backends already supported config via [DEFAULT]/transport_url at that time. so it should be okay | |
| 17:37:55 | superdan | melwitt: I have no idea why, but suddenly the host calls are failing in cellsv1 with my patch: http://logs.openstack.org/75/396775/21/check/gate-tempest-dsvm-cells-ubuntu-xenial/2d615a2/logs/screen-n-api.txt.gz?#_2017-01-06_16_01_16_390 | |
| 17:38:19 | superdan | melwitt: timing out calling through cells to the compute host.. super weird since that patch only touches instance listings | |
| 17:38:41 | openstackgerrit | Julian proposed openstack/nova: Transform instance.add_fixed_ip notification https://review.openstack.org/332876 | |
| 17:39:19 | leakypipes | sfinucan, sean-k-mooney: questions on commit message. | |
| 17:39:28 | melwitt | superdan: lettuce see | |
| 17:39:37 | superdan | melwitt: just talking out loud | |
| 17:40:11 | melwitt | oh, yeah. I was thinking maybe since listing does a /detail call, it's trying to pull some host info? that doesn't really make sense though | |
| 17:41:14 | sean-k-mooney | leakypipes: thanks jay ill take a look at it now. im just opening a bug for the windows issues that sfinucan change uncovered | |
| 17:42:05 | superdan | melwitt: it's just the test hitting the /hypervisors api afaict | |
| 17:42:34 | melwitt | yeah, I see that now. weird | |
| 17:46:48 | leakypipes | superdan: so did you figure out the substitution thing? | |
| 17:47:18 | superdan | leakypipes: no, I went with the + instead of * thing, and will change it if mriedem requires and/or tells me how | |
| 17:47:36 | superdan | leakypipes: but I'm hoping he'll just let it go as is | |
| 17:47:48 | leakypipes | superdan: but I see "%(int:service_id)s" | |
| 17:48:11 | leakypipes | superdan: should that just be "%(service_id)s" | |
| 17:48:13 | leakypipes | ? | |
| 17:49:06 | superdan | none of those work for me | |
| 17:54:05 | openstackgerrit | Gábor Antal proposed openstack/nova: Added instance.reboot.error to the legacy notifications https://review.openstack.org/411816 | |
| 17:54:06 | openstackgerrit | Gábor Antal proposed openstack/nova: Transform instance.reboot.error notification https://review.openstack.org/411791 | |
| 17:55:11 | melwitt | superdan: I guess the only thing I notice is if you look at the test_results all of the tests failed with a timeout of some kind, but for some reason only the hypervisors api tests are showing any traces in the logs of a timeout | |
| 17:55:54 | melwitt | so I wonder if it's a side effect | |
| 17:56:09 | superdan | but.. look at the change.. it's just adding stuff to get_all instances | |
| 17:56:26 | superdan | seems like.. totally unrelated | |
| 17:57:06 | melwitt | I know. I was thinking maybe other tests that do a listing bombarded the queue so hard that it made all the other tests time out too? I dunno | |
| 17:57:36 | superdan | I dunno, or if we locked up the compute node somehow | |
| 17:57:46 | melwitt | hm, yeah | |
| 17:57:48 | superdan | the hypervisors one is the only one I saw with any tracebacks anywhere, | |
| 17:57:54 | superdan | so I was trying to follow that thread | |
| 17:59:08 | melwitt | there's a guru meditation report in the n-cpu.log | |
| 17:59:14 | melwitt | is that usual? | |
| 17:59:53 | superdan | any time we fail the run we do that I think | |
| 18:00:02 | superdan | also, I see debug in n-api that shows us listing cells prperly and stuff | |
| 18:00:08 | superdan | Found 2 cells: cell0,cell1 _get_instances_by_filters_all_cells | |
| 18:00:13 | superdan | which is what this thing changes | |
| 18:00:39 | superdan | although, the two patches before this don't run the cells job | |
| 18:00:43 | superdan | because they're test only | |
| 18:00:58 | superdan | so I wonder if this is just failing recently by chance? seems too unlikely, but... | |