| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-03-07 | |||
| 16:37:36 | mriedem | there is | |
| 16:37:37 | mriedem | https://github.com/openstack/tempest/blob/master/tempest/api/compute/admin/test_auto_allocate_network.py#L150 | |
| 16:37:52 | dansmith | mriedem: that tests no allocate right? | |
| 16:37:57 | mriedem | correct | |
| 16:38:15 | dansmith | but the multi one tests that there _is_ allocation | |
| 16:38:17 | mriedem | but we still create a server | |
| 16:38:30 | mriedem | why would it matter for multi cell? | |
| 16:38:58 | dansmith | I don't know why it's failing, and was looking for a non-multi-create test to see why that works and multi doesn't | |
| 16:39:17 | mriedem | log link me? | |
| 16:39:19 | dansmith | changing it to one server doesn't make it pass, so it must not be related to multi-create, I was just curious | |
| 16:39:47 | dansmith | mriedem: http://logs.openstack.org/94/436094/12/check/gate-tempest-dsvm-neutron-full-ubuntu-xenial/051cffa/logs/testr_results.html.gz | |
| 16:42:30 | openstackgerrit | Ed Leafe proposed openstack/nova master: Add test for inventory with invalid ResourceClass https://review.openstack.org/442498 | |
| 16:43:42 | dansmith | I'm not even sure what to look at | |
| 16:46:25 | mriedem | just look into my eyes and don't be afraid | |
| 16:47:09 | mriedem | dansmith: so we're failing on https://github.com/openstack/tempest/blob/master/tempest/api/compute/admin/test_auto_allocate_network.py#L188 | |
| 16:47:19 | mriedem | are we going to the proper cell to get the info_cache? | |
| 16:47:37 | mriedem | that test creates 3 servers and waits for them all to be ACTIVE, | |
| 16:47:55 | mriedem | then iterates over them and lists their addresses to make sure there was a nic assigned | |
| 16:48:09 | dansmith | should be.. you think that's the only test that looks at stuff from info cache? | |
| 16:48:13 | mriedem | so either the nic isn't getting assigned, or we're not pulling the info_cache properly | |
| 16:48:35 | mriedem | there should be other tests that do that for attaching interfaces | |
| 16:48:51 | dansmith | yeah, and those appear to work | |
| 16:49:06 | mriedem | i'm not sure what they assert though | |
| 16:49:11 | mriedem | checking the n-cpu logs now | |
| 16:49:12 | dansmith | mriedem: and if I boot a server manually I see an address | |
| 16:49:21 | dansmith | let me try with auto networks | |
| 16:49:24 | mriedem | dansmith: with devstack? | |
| 16:49:27 | dansmith | yeah | |
| 16:49:30 | mriedem | dansmith: that's not the same, | |
| 16:49:37 | mriedem | because your tenant in devstack has access to the public network already | |
| 16:49:43 | mriedem | so we don't hit the auto-allocate paths | |
| 16:49:45 | mriedem | in n-cou | |
| 16:49:47 | dansmith | right, I'm saying we're pulling those nic details | |
| 16:49:48 | mriedem | *n-cpu | |
| 16:49:50 | dansmith | in the api | |
| 16:50:06 | mriedem | yeah, but this specific test has a tenant which has no access to any existing networks | |
| 16:50:13 | openstackgerrit | Dan Smith proposed openstack/nova master: Make os-availability-zones know about cells https://review.openstack.org/442163 | |
| 16:50:13 | openstackgerrit | Dan Smith proposed openstack/nova master: WIP: Get instance availability_zone without hitting the api db https://review.openstack.org/439754 | |
| 16:50:14 | openstackgerrit | Dan Smith proposed openstack/nova master: Make scheduler target cells to get compute node instance info https://review.openstack.org/439891 | |
| 16:50:14 | openstackgerrit | Dan Smith proposed openstack/nova master: Change MQ targeting to honor only what is in the context https://review.openstack.org/438023 | |
| 16:50:15 | openstackgerrit | Dan Smith proposed openstack/nova master: WIP: Make conductor ask scheduler to limit migrates to same cell https://review.openstack.org/438025 | |
| 16:50:15 | openstackgerrit | Dan Smith proposed openstack/nova master: Add cell field to Destination object https://review.openstack.org/438024 | |
| 16:50:16 | openstackgerrit | Dan Smith proposed openstack/nova master: Teach os-aggregates about cells https://review.openstack.org/442200 | |
| 16:50:51 | dansmith | right, I'm just saying I think api is pulling info cache from the right cell db because I see nic details normally | |
| 16:56:04 | mriedem | hmm http://logs.openstack.org/94/436094/12/check/gate-tempest-dsvm-neutron-full-ubuntu-xenial/051cffa/logs/screen-n-cpu.txt.gz#_2017-03-07_00_14_21_656 | |
| 16:56:14 | mriedem | No network configured allocate_for_instance | |
| 16:56:19 | mriedem | Instance network_info: |[]| | |
| 16:56:51 | dansmith | is that before it goes to allocate or after? | |
| 16:57:02 | mriedem | during | |
| 16:57:50 | dansmith | once we're on the compute node, there's no cell switching involved, so I'm not sure why multicell has anything to do with this | |
| 16:58:02 | dansmith | is there something we put into info cache from the api side on instance create or something? | |
| 16:58:19 | dansmith | on my test system, I have nothing in cell0 for info caches, so it's not like we're creating them there by accident or something | |
| 16:58:29 | mriedem | i don't see "Automatically allocating a network for project" | |
| 16:59:17 | dansmith | should that be logged by n-cpu? | |
| 16:59:21 | mriedem | yeah | |
| 16:59:45 | mriedem | https://github.com/openstack/nova/blob/master/nova/network/neutronv2/api.py#L1495 | |
| 17:00:51 | dansmith | yeah | |
| 17:00:59 | mriedem | i'm also not seeing this https://github.com/openstack/nova/blob/master/nova/network/neutronv2/api.py#L1542 | |
| 17:01:04 | mriedem | ^ is called via the api | |
| 17:01:20 | mriedem | oh well i should look for that in the n-api logs | |
| 17:02:44 | mriedem | i don't see that in n-api logs either | |
| 17:02:58 | mriedem | i wonder if somehow the auto_allocate flag isn't getting set on the NetworkRequest object? | |
| 17:03:46 | dansmith | not sure why that would be.. we don't even persist that right? | |
| 17:06:33 | mriedem | we don't persist it | |
| 17:06:41 | mriedem | are we somehow mangling what we pass between n-api and conductor? | |
| 17:07:17 | mriedem | i would doubt it because otherwise we wouldn't have any tests that create a server on a specific nic passing | |
| 17:07:27 | mriedem | which i believe is most of tempest | |
| 17:07:36 | mriedem | they all have tenant-isolated networks | |
| 17:09:12 | dansmith | mriedem: no mangling that I know of, and this all passes on the same stack of patches with normal devstack, AFAICT | |
| 17:11:09 | mriedem | ok. notifications meeting atm and then i'll keep digging. | |
| 17:11:12 | mriedem | we might need a debug logging patch | |
| 17:13:42 | dansmith | okay thanks for helping | |
| 17:14:45 | mriedem | https://www.youtube.com/watch?v=RWc5gsU-eeY | |
| 17:15:13 | mriedem | mtreinish: ^ | |
| 17:16:33 | openstackgerrit | Eric Fried proposed openstack/nova master: PowerVM Driver: spawn/destroy #3: TaskFlow https://review.openstack.org/438729 | |
| 17:16:51 | openstackgerrit | Eric Fried proposed openstack/nova master: PowerVM Driver: spawn/destroy #4: full flavor https://review.openstack.org/391288 | |
| 17:16:51 | dansmith | heh | |
| 17:17:45 | efried | thorst adreznec esberglu Need re-+1 on https://review.openstack.org/#/c/438729/ (TaskFlow) please. | |
| 17:24:48 | mdbooth | lyarwood: Before I looked, I thought there was a swap_volume flow where we simply replaced 1 volume with another, contents and all. However, looking now it appears we'll always attempt to rebase the contents of the old volume on to the new volume. Is that right, or am I just blind today? | |
| 17:25:46 | lyarwood | mdbooth: we always rebase, even with the cinder retype flow | |
| 17:26:00 | mdbooth | Well the cinder retype flow is the one I knew did it | |
| 17:26:15 | mdbooth | I thought the crackpot flow didn't do it though, for some reason | |
| 17:26:27 | mriedem | efried: you guys are using taskflow huh | |
| 17:26:32 | lyarwood | mdbooth: yeah volume-update does it as well | |
| 17:26:41 | efried | mriedem Yes. | |
| 17:26:45 | mdbooth | lyarwood: Thanks | |
| 17:26:49 | efried | That's... good, right mriedem ? | |
| 17:26:56 | mriedem | efried: we've never used it in nova | |
| 17:27:03 | mriedem | i know cinder uses it, sort of, for some tihngs | |
| 17:27:09 | mriedem | but never liked it enough to use it for all operations | |
| 17:27:10 | efried | Oh. I was under the impression that e.g. libvirt used it. | |
| 17:27:16 | mriedem | maybe it's fine on a small scale confied to a single virt driver | |
| 17:28:17 | efried | mriedem We find it super useful for rollbacks. | |
| 17:30:25 | mriedem | like spawn fails, detach interfaces and block devices? | |
| 17:30:31 | efried | yup | |
| 17:30:53 | mriedem | dude you can totally just have a 300LOC mega method with lots of try/excepts :) | |
| 17:31:44 | efried | Hey, y'all are the ones who wanted smaller change sets. | |
| 17:31:53 | efried | We'll bloat it later for ya. | |
| 17:33:53 | mdbooth | jaypipes: Did somebody bust you out of some meetings? | |
| 17:33:54 | jaypipes | dansmith: and yes, I'm listening to "Business Time" | |