| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2023-07-04 | |||
| 12:39:37 | sean-k-mooney | so that you can pass a dict of object | |
| 12:39:58 | sean-k-mooney | and then in the relvent tests that fial just mock out get_packed_virtqueue_constraint to return true | |
| 12:40:42 | sean-k-mooney | im just not sure how many tests you will have to fix if you take that approch | |
| 12:56:19 | opendevreview | Merged openstack/nova stable/xena: Fix segment-aware scheduling permissions error https://review.opendev.org/c/openstack/nova/+/884808 | |
| 12:56:28 | ykarel | gibi, sean-k-mooney bauzas can you please revisit https://review.opendev.org/c/openstack/nova/+/868419 | |
| 12:56:56 | sean-k-mooney | sure i have a meeting at the top of the hour but ill look quickly | |
| 12:57:08 | bauzas | ykarel: ack, will look | |
| 12:57:10 | ykarel | Thanks | |
| 12:57:26 | sean-k-mooney | ack so you move this to the libvirt section from workarounds | |
| 12:57:32 | ykarel | yes | |
| 12:58:15 | sean-k-mooney | that was the only outstding issue as far as i was concerned | |
| 12:58:19 | sean-k-mooney | so i think this is good as is now | |
| 13:04:58 | sahid | o/ quick question guys, we are trying to use nova-manage db archive_deleted_rows but for some reason we dont't see anything migrated in the shadow tables even if the output is reporting something | |
| 13:05:30 | sahid | more of that when we list instances with deleted_at not null we can still see them | |
| 13:06:05 | sahid | and the final strp of crazy is that, when re-executing the script it sees nothing else to archive... | |
| 13:06:50 | sahid | i was thinking about something related to the cells... | |
| 13:08:23 | sahid | list_cells only report cell0 so I guess we are good | |
| 13:10:02 | sahid | one point that i don't understand, we have nova and nova_cell0 databases | |
| 13:10:45 | sahid | in nova we have a certain number of instances deleted | |
| 13:11:02 | sahid | in nova_cell0 we have nothing | |
| 13:14:11 | opendevreview | Merged openstack/nova stable/xena: Fix get_segments_id with subnets without segment_id https://review.opendev.org/c/openstack/nova/+/883725 | |
| 13:22:05 | dvo-plv | looks like we did it :) | |
| 13:22:06 | dvo-plv | congratulations :) | |
| 13:22:06 | dvo-plv | py38: commands succeeded | |
| 13:33:29 | bauzas | sahid: tbc, there is no longer a 'nova' DB, there is either a nova API DB or a nova cell DB | |
| 13:33:46 | bauzas | in devstack, we name them nova_api and nova_cell1 | |
| 13:34:26 | bauzas | + there is a specific nova_cell0 DB which is exactly like nova_cell1 but where only have NoValidHosts instances | |
| 13:46:22 | opendevreview | Danylo Vodopianov proposed openstack/nova master: Packed virtqueue support was added. https://review.opendev.org/c/openstack/nova/+/876075 | |
| 14:08:49 | sean-k-mooney | dvo-plv: cool ill take a look shortly | |
| 14:11:03 | sean-k-mooney | dvo-plv: fyi the funtional tests are not run as part of the py38 target | |
| 14:11:14 | sean-k-mooney | you need to run tox -e functional | |
| 14:11:25 | sean-k-mooney | optionally wiht -py38 at the end | |
| 14:17:29 | sahid | bauzas: we are running ussuri :D | |
| 14:21:50 | sahid | in all cases I don't get what is happening with this script | |
| 14:45:24 | plibeau | hello guys if you have time to review: https://review.opendev.org/c/openstack/nova/+/861172 thx | |
| 14:52:22 | bauzas | sahid: do you have all the right cells ? | |
| 14:56:31 | sahid | I don't know I think there is something I don't get | |
| 14:57:03 | sahid | when I list cells, I have 2 rows, one for rabbit without name and one cell0 | |
| 14:57:44 | sahid | the database connection for cell0 looks to use nova_cell0 dbname | |
| 14:58:27 | sahid | but in nova_cell0 db, there are nothing I mean If I count the instances fro; this database (nova_cell0), there are nothing | |
| 14:59:13 | sahid | the right database used is the dbname called 'nova' | |
| 14:59:22 | sahid | I don't know if that is normal... | |
| 15:07:12 | bauzas | sahid: can you list them in an etherpad ? | |
| 15:07:37 | bauzas | sahid: cell0 is just a DB for instances that were not accepted | |
| 15:08:21 | bauzas | I mean, when an instance doesn't have a specific host because of some problem (scheduling in general), then we create the record in cell0 | |
| 15:08:39 | bauzas | instead of recording the instance in the cell DB related to the host | |
| 15:09:02 | bauzas | so you should have 3 DBs : one for the API, one for cell0 and one (at least) for *one* cell | |
| 15:09:25 | bauzas | you should have *at least* 3 (I mean) | |
| 15:10:48 | sahid | nova, nova_api and nova_cell0 | |
| 15:11:17 | sahid | https://etherpad.opendev.org/p/cells | |
| 15:12:20 | sahid | bauzas: the issue we have is that, we are trying to archive deleted instances | |
| 15:12:33 | sahid | but when executing the script they stay | |
| 15:12:57 | sahid | and nothing in shadow tables | |
| 15:16:05 | sean-k-mooney | the nova database there is the nova cell 1 database | |
| 15:16:32 | sean-k-mooney | sahid: some tools did that but we recommend actully not calling it nova but nova_cell1 | |
| 15:16:47 | sean-k-mooney | basicly it simpliced upgrades so they kept the old name | |
| 15:16:57 | bauzas | sahid: sean-k-mooney: mmm, I wonder why the cell1 doesn't have a name | |
| 15:17:28 | sean-k-mooney | the actuall db schame name is not used for anything | |
| 15:17:54 | sean-k-mooney | when we splict the monolitic nova db into the celldb and api db | |
| 15:18:13 | sean-k-mooney | most instaler did that by extracting the api tables into a differnt schema | |
| 15:18:24 | sean-k-mooney | and reusing the existing db as the cell1 db | |
| 15:19:27 | bauzas | ok, but then I wonder why nova-manage db archive_deleted_rows doesn't work | |
| 15:22:24 | sahid | bauzas, sean-k-mooney I should perhaps try to name cell1 the cell with empty name? | |
| 15:22:53 | sean-k-mooney | it should wrok waht is the error | |
| 15:23:09 | sean-k-mooney | where was it run | |
| 15:23:11 | sahid | no errors are reported | |
| 15:23:18 | sahid | but nothing change | |
| 15:23:23 | sean-k-mooney | it gets the info form the nova.conf | |
| 15:23:54 | sean-k-mooney | does the nova.conf have the api and cell db set | |
| 15:25:31 | sahid | it has section [database] that mentions as connection = .../nova_cell0 | |
| 15:25:34 | sahid | and .. | |
| 15:26:02 | sahid | [api_database] that mentions as connection = .../nova_api | |
| 15:26:30 | sahid | (for both of the section we have a slave_connection) | |
| 15:27:55 | bauzas | sahid: you shouldn't mention the cell0 connection string in the nova.conf | |
| 15:28:23 | bauzas | [database] should refer to to cell connection it's attached to | |
| 15:28:50 | sean-k-mooney | it depends on which service that nova.confi is for | |
| 15:29:16 | sean-k-mooney | if it was the api's nova.conf or sheduler then pointing at cell0 is fine | |
| 15:29:49 | sahid | it is for nova-api | |
| 15:30:18 | sean-k-mooney | ok so what command are you trying to run exactly | |
| 15:31:30 | bauzas | sean-k-mooney: correct, I was assuming sahid was referring to the compute's nova.conf | |
| 15:31:33 | sahid | I have tried several ways but basically, nova-manage archive_deleted_rows --all-cells --unit-complete --verbose | |
| 15:32:10 | sean-k-mooney | ack so provided you have the cells mapped properly that should lookup the other cells db via the api db | |
| 15:32:34 | sean-k-mooney | nomally we woudl run this using the conducotrs nova.conf | |
| 15:33:13 | bauzas | https://docs.openstack.org/nova/latest/cli/nova-manage.html#db-archive-deleted-rows | |
| 15:33:13 | bauzas | sahid: you need to have a right nova.conf that gives the right connection for API DB | |
| 15:33:15 | sean-k-mooney | sahid: are you using templated URLs | |
| 15:33:21 | bauzas | " Move deleted rows from production tables to shadow tables. Note that the corresponding rows in the instance_mappings, request_specs and instance_group_member tables of the API database are purged when instance records are archived and thus, api_database.connection is required in the config file." | |
| 15:33:38 | elodilles | i know that nova meeting will have start in less than 30 mins, but i have a question about nova gate: | |
| 15:33:49 | sean-k-mooney | sure | |
| 15:34:01 | elodilles | is there a reported bug about the nova-lvm gate issue? https://zuul.opendev.org/t/openstack/builds?job_name=nova-lvm&skip=0 | |
| 15:34:26 | sean-k-mooney | i was not aware there was one but apprelty so | |
| 15:34:28 | elodilles | it's 100% failing on master and newer stable branches | |
| 15:35:15 | sean-k-mooney | ok so it looks like | |
| 15:35:16 | elodilles | the issue is mostly the same as it was last week with the ceph job | |
| 15:35:20 | sean-k-mooney | the tempest test is not using sshable | |
| 15:35:22 | elodilles | but it's different | |
| 15:35:25 | sean-k-mooney | and we dont have that configued in the job | |
| 15:35:32 | sean-k-mooney | tempest.lib.exceptions.InvalidParam: Invalid Parameter passed: When validation.connect_method equals floating, validation_resources cannot be None | |
| 15:35:38 | sean-k-mooney | test_rebuild_volume_backed_server | |
| 15:35:46 | sean-k-mooney | so that should be a simple fix | |