| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2019-10-16 | |||
| 18:44:06 | mriedem | Instance.save(expected_state_state=) | |
| 18:44:08 | mriedem | that's a tuple or lits, | |
| 18:44:11 | mriedem | *list | |
| 18:44:17 | mriedem | the singular in the name *always* screws me up | |
| 18:44:31 | dansmith | except that it takes None too, IIRC | |
| 18:44:31 | mriedem | we should maybe guard against that in the save method | |
| 18:44:48 | mriedem | yeah, | |
| 18:44:55 | mriedem | so if not None, assert it's an iterable | |
| 18:45:16 | mriedem | https://review.opendev.org/#/c/627891/63/nova/conductor/tasks/cross_cell_migrate.py@423 | |
| 18:46:39 | efried | melwitt: in case you haven't been following closely, note that we no longer do fast approvals for unchanged spec reproposals. | |
| 18:47:13 | melwitt | efried: I had read that in one of the emails, then forgot about it, so thanks for reminding | |
| 18:50:24 | mriedem | yeah we have several places where we pass a string to expected_task_state | |
| 18:53:32 | artom | expected_attrs all over again | |
| 18:53:52 | openstack | Launchpad bug 1848373 in OpenStack Compute (nova) "Instance.save(expected_task_state=) is passed string in many locations" [Medium,Triaged] | |
| 18:53:52 | mriedem | https://bugs.launchpad.net/nova/+bug/1848373 | |
| 18:54:31 | melwitt | dynamic types ftw | |
| 18:57:50 | mriedem | back in 2012 the db api handled it.... https://review.opendev.org/#/c/12222/5/nova/db/sqlalchemy/api.py | |
| 18:57:53 | efried | gmann: was os-cells a cellsv1 thing? | |
| 18:57:57 | mriedem | yes | |
| 18:58:16 | efried | thx | |
| 18:58:28 | gmann | yeah | |
| 18:58:30 | melwitt | hah, I wonder why that was removed | |
| 18:59:12 | mriedem | https://review.opendev.org/#/c/202593/ | |
| 18:59:22 | mriedem | since liberty | |
| 18:59:34 | mriedem | so i don't even know if expected_task_state is even enforced in the DB API anymore | |
| 18:59:40 | mriedem | dansmith: ^ does that jive with you? | |
| 19:00:09 | mriedem | https://github.com/openstack/nova/blob/1a226aaa9e8c969ddfdfe198c36f7966b1f692f3/nova/db/sqlalchemy/api.py#L2783 is definitely no longer true | |
| 19:01:19 | mriedem | wtf, that expected= kwarg in instance_update_and_get_original isn't even used anywhere | |
| 19:01:54 | dansmith | mriedem: I don't think I know anything about that change | |
| 19:02:06 | mriedem | i remember mdbooth pushing for it way back when | |
| 19:02:35 | dansmith | I remember talking about the comnpare-and-swap thing | |
| 19:03:21 | mriedem | and i reverted the first attempt https://review.opendev.org/#/c/201771/ | |
| 19:05:38 | melwitt | time for another revert | |
| 19:05:39 | melwitt | JUST KIDDING | |
| 19:05:52 | mriedem | don't tempt me | |
| 19:07:02 | mriedem | oh i see where expected_task_state is used | |
| 19:07:04 | mriedem | tricky | |
| 19:07:24 | mriedem | https://github.com/openstack/nova/blob/1a226aaa9e8c969ddfdfe198c36f7966b1f692f3/nova/db/sqlalchemy/api.py#L2850 | |
| 19:09:06 | dansmith | # Coerce all single values to singleton lists | |
| 19:09:45 | mriedem | yeah i see that too | |
| 19:10:05 | mriedem | i'll abandon the bug since even though we're using the object interface incorrectly the db api is hiding it | |
| 19:11:50 | mriedem | the expected= kwarg is still never used from anywhere | |
| 19:12:03 | mriedem | so it just adds to confusion | |
| 19:25:41 | openstack | bug 1848343 in OpenStack Compute (nova) "Reverting migration-based allocations leaks allocations if the server is deleted" [Medium,Triaged] https://launchpad.net/bugs/1848343 | |
| 19:25:41 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add compute side revert allocation test for bug 1848343 https://review.opendev.org/689013 | |
| 19:27:52 | mriedem | heh, just got a random code review comment from someone on a merged commit in github, | |
| 19:28:01 | mriedem | i appreciate the stat padding but | |
| 19:38:42 | mriedem | stable/train is open so i've start +2ing things we were holding https://review.opendev.org/#/q/status:open+project:openstack/nova+branch:stable/train | |
| 19:38:46 | mriedem | *started | |
| 19:39:22 | mriedem | bauzas: since you brought up queens reviews last week maybe you can go through some of these tomorrow https://review.opendev.org/#/q/status:open+project:openstack/nova+branch:stable/queens+label:Code-Review=1 | |
| 19:42:53 | openstackgerrit | Andreas Jaeger proposed openstack/nova master: Switch to opensuse-15 nodeset https://review.opendev.org/689020 | |
| 19:51:14 | mriedem | efried: some rambling in https://review.opendev.org/#/c/687416/ if it means anything to you, probably not worth pursuing | |
| 20:07:45 | efried | mriedem: separate pursuit though, yah? | |
| 20:08:07 | efried | I was also considering making False the default throughout, since that's the more "modern" thing, but that would obv be a separate change as well. | |
| 20:08:14 | mriedem | efried: yeah | |
| 20:44:31 | KeithMnemonic | frickler: we found the issue, they needed -on-shared-storage. | |
| 21:37:11 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Delete source allocations in move_allocations if target no longer exists https://review.opendev.org/689049 | |
| 22:05:27 | openstackgerrit | Takashi NATSUME proposed openstack/nova master: Update keypairs in saving an instance object https://review.opendev.org/683043 | |
| 22:05:46 | openstackgerrit | Takashi NATSUME proposed openstack/nova master: Move Destination object tests to their own test class https://review.opendev.org/683017 | |
| 23:24:46 | openstackgerrit | Merged openstack/nova stable/train: Fix unit of hw_rng:rate_period https://review.opendev.org/688083 | |
| 23:52:48 | openstackgerrit | Brin Zhang proposed openstack/python-novaclient master: Add functional test for migration-list in v2.80 https://review.opendev.org/688635 | |
| #openstack-nova - 2019-10-17 | |||
| 00:02:09 | openstackgerrit | Brin Zhang proposed openstack/nova-specs master: Remove the todo in the migrations spec https://review.opendev.org/689056 | |
| 00:43:51 | openstackgerrit | Merged openstack/nova-specs master: Remove the todo in the migrations spec https://review.opendev.org/689056 | |
| 02:14:32 | SonPham | hi. how to install python-novaclient package from github. I installed from git but horizon error | |
| 03:07:26 | openstackgerrit | Liang Fang proposed openstack/nova-specs master: Support volume local cache https://review.opendev.org/689070 | |
| 04:14:06 | LiangFang | hi cores, I want to propose a spec about volume local cache. Can anybody be my Feature liaison？Thank you so much. spec: Support volume local cache https://review.opendev.org/689070 | |
| 04:14:15 | LiangFang | I'm new to Nova | |
| 06:58:08 | eandersson | VMs stuck in BUILD is such a pain :'( | |
| 07:02:52 | gibi_off | eandersson: can I do something to help with that? | |
| 07:03:26 | eandersson | gibi, been fighting these for a while in Rocky. We got a few issus fixed upstream. | |
| 07:03:49 | eandersson | We first though these were due to restarted computes | |
| 07:03:57 | eandersson | https://review.opendev.org/#/c/687535/ | |
| 07:04:08 | eandersson | but I think the issues are due to race conditions in this case | |
| 07:04:18 | eandersson | I see two instances scheduled at the same second on the same compute | |
| 07:04:30 | eandersson | I see an allocation, but both VMs are stuck in building / scheduling | |
| 07:04:50 | eandersson | with no compute visible in the instance info | |
| 07:04:59 | eandersson | (had to track the compute down using the db) | |
| 07:05:40 | gibi | eandersson: interesting. So you see both server having allocation in placement or only one of the server has allocation? | |
| 07:05:58 | eandersson | both have allocation in placement | |
| 07:06:09 | eandersson | both are stuck in BUILD | |
| 07:06:40 | gibi | and none of the servers has instance.host updated to point to the compute? | |
| 07:07:23 | eandersson | correct | |
| 07:08:24 | eandersson | I see on the compute 2x Final resource view, image xxx at /var/lib | |
| 07:08:27 | eandersson | and then nothing else | |
| 07:17:47 | eandersson | hmm maybe the compute is bad in this case | |
| 07:17:51 | eandersson | I just got 3 more stuck on that host | |
| 07:18:07 | eandersson | but don't understand why they never fail | |
| 07:18:16 | gibi | eandersson: I guess instance.task_state is None. Unfortunately there is no logs coming out from the compute between the build request reaching the compute and setting the instance.vm_state to BUILDING and between the instance_claim that will set instance.host | |
| 07:20:20 | eandersson | btw we are running the very latest rocky | |
| 07:21:01 | gibi | eandersson: did you happen to see instance.create.start versioned notification or compute.instance.create.start legacy notification for these servers? | |
| 07:21:14 | eandersson | I can check designate | |
| 07:21:43 | eandersson | or nvm they only catch end | |
| 07:23:06 | gibi | eandersson: does pci devices or sriov ports are requested for these servers? | |
| 07:24:35 | eandersson | nothing special | |
| 07:25:41 | gibi | eandersson: the two Final resource view logged considers the allocation of the servers? | |
| 07:26:23 | eandersson | Sorry, don't fully understand that | |
| 07:27:00 | eandersson | last part | |
| 07:27:03 | gibi | you mentioned that you saw two "Final resouce view " logs from the compute | |
| 07:27:18 | gibi | that contains how much vcpu and ram is used on the compute | |
| 07:27:42 | gibi | does that usage contains the usage of your two servers stuck in build? | |