| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-07-18 | |||
| 19:55:54 | openstackgerrit | Jackie Truong proposed openstack/nova master: [WIP] Add trusted certificates to InstanceExtras https://review.openstack.org/457711 | |
| 19:57:35 | openstackgerrit | Jackie Truong proposed openstack/nova master: [WIP] Add trusted certificates to InstanceExtras https://review.openstack.org/457711 | |
| 20:02:45 | oorgeron | Hi folks, I'm working on a patch to enable nova to use mysql cluster. In tests, I'm running into errors that seem to indicate that sqlite is being used in the sqlalchemy migrations. Michael Bayer said that this has happened before in the nova gate, anyone know how to fix that? https://review.openstack.org/#/c/446643/ | |
| 20:06:08 | melwitt | oorgeron: the unit tests intentionally use sqlite | |
| 20:06:27 | melwitt | and functional tests | |
| 20:08:06 | melwitt | that said, I don't know how you could get things working with your change. probably need to ask someone like sdague | |
| 20:12:31 | oorgeron | melwitt: I'll reach out to sdague. | |
| 20:13:14 | openstackgerrit | Dan Smith proposed openstack/nova master: Consider instance flavor resource overrides in allocations https://review.openstack.org/484935 | |
| 20:13:30 | dansmith | mriedem: edleafe ^.. I still need to figure out what gabbit is failing on one of the functional tests | |
| 20:13:50 | mriedem | oorgeron: you can run mysql and postgresql opportunistic tests locally if you do some setup | |
| 20:14:32 | sdague | oorgeron: yeh, everything there looks like unit testable failures because of an assumption about an index | |
| 20:14:51 | mriedem | there should be a script in nova/tools to set this stuff up | |
| 20:14:58 | sdague | mriedem: it's not even that | |
| 20:15:20 | mriedem | https://github.com/openstack/nova/blob/master/tools/test-setup.sh | |
| 20:15:30 | sdague | http://logs.openstack.org/43/446643/6/check/gate-nova-python27-ubuntu-xenial/6e79a39/console.html.gz#_2017-07-14_19_47_37_197086 | |
| 20:15:39 | sdague | these are just run of the mill, non mysql / pg tests failing | |
| 20:16:01 | oorgeron | sdague: yeah the test is failing with a sqlite error with dropping an index. The scripts work fine with innodb and ndb as the database. | |
| 20:16:13 | sdague | oorgeron: right, they have to work with sqlite as well | |
| 20:16:18 | sdague | that's our unit testing environment | |
| 20:16:37 | mriedem | yeah...you can't just do stuff like this https://review.openstack.org/#/c/446643/6/nova/db/sqlalchemy/migrate_repo/versions/249_remove_duplicate_index.py | |
| 20:17:50 | mriedem | and this scares me https://review.openstack.org/#/c/446643/6/nova/db/sqlalchemy/models.py | |
| 20:18:21 | oorgeron | mriedmen: A lot of the other services are just doing tests against mysql and maybe postgresql. I could use another method to drop and add the foreignkey and not use the migrate method for doing that. | |
| 20:18:40 | mriedem | oorgeron: everyone is using sqlite for unit tests | |
| 20:18:48 | mriedem | we test with mysql in the devstack jobs in the gate | |
| 20:18:52 | mriedem | for integration testing | |
| 20:19:07 | mriedem | the unit tests also run with the opportunistic mysql and postgresql backends in the gate | |
| 20:19:09 | oorgeron | mriedem: I can drop foreign keys like this in other services and the tests pass. | |
| 20:19:14 | mriedem | but locally you have to set that all up optionally | |
| 20:19:23 | mriedem | oorgeron: that doesn't make it right | |
| 20:19:34 | mriedem | "hey look i deleted a bunch of code and nothing blew up!" | |
| 20:19:39 | mriedem | that probably just means they aren't doing testing | |
| 20:20:31 | mriedem | the model changes here worry me that the schema will change for other backends, like postgresql | |
| 20:20:50 | oorgeron | mriedem: first off, when you drop an index, you supposed to remove the foreign keys, if we're talking about the right order of sql steps. Innodb just happens to be very forgiving on that, doesn't make it right either. | |
| 20:21:13 | oorgeron | mriedem: also, if you look at that, it adds the key back after the index is dropped, which is the right way to do this in sql | |
| 20:21:46 | jaypipes | mriedem: how may I be of assistance? | |
| 20:21:49 | oorgeron | mriedem: But I'll try a dumber wait to drop and recreate the foreign key, instead of using the migrate method here. | |
| 20:21:57 | oorgeron | way | |
| 20:23:27 | oorgeron | mriedem: the changes to model will only apply when ndb is turned on. The ndb function calls there do the right thing automatically. | |
| 20:24:16 | mriedem | so in https://review.openstack.org/#/c/446643/6/nova/db/sqlalchemy/migrate_repo/versions/249_remove_duplicate_index.py you'd have to handle the index not existing before trying to drop it | |
| 20:25:02 | mriedem | jaypipes: half joking | |
| 20:25:08 | mriedem | i unsummon you | |
| 20:25:13 | mriedem | and banish you back to placement land | |
| 20:25:17 | mriedem | *placement island | |
| 20:26:22 | oorgeron | mriedem: besides, 292_drop_nova_volumes_tables.py uses the same method and is not part of my patch :) | |
| 20:28:33 | openstackgerrit | Jan Gutter proposed openstack/os-vif master: Add support for VIFPortProfileOVSRepresentor https://review.openstack.org/483921 | |
| 20:30:51 | openstackgerrit | Jackie Truong proposed openstack/nova master: [WIP] Implement certificate_utils https://review.openstack.org/479949 | |
| 20:49:01 | oorgeron | sdague: is there a way to run the tests just against sqlite? | |
| 20:54:45 | openstackgerrit | Jackie Truong proposed openstack/nova master: Add trusted certificates to InstanceExtras https://review.openstack.org/457711 | |
| 20:57:20 | openstackgerrit | Jan Gutter proposed openstack/os-vif master: Add support for VIFPortProfileOVSRepresentor https://review.openstack.org/483921 | |
| 21:06:25 | openstackgerrit | Ed Leafe proposed openstack/nova master: WIP - Migrate Ironic Flavors https://review.openstack.org/484949 | |
| 21:06:36 | edleafe | dansmith: ^^ Strawman code that has the inventory/allocation changes | |
| 21:06:55 | edleafe | dansmith: I had already written that before you said not to :) | |
| 21:08:13 | mriedem | dansmith: here it comes | |
| 21:08:16 | mriedem | prepare | |
| 21:08:27 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Handle uuids in os-hypervisors API https://review.openstack.org/471899 | |
| 21:08:28 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Use uuid for id in os-services API https://review.openstack.org/464280 | |
| 21:14:45 | dansmith | edleafe: remember how I asked a few times if anyone was working on that? | |
| 21:17:59 | edleafe | dansmith: remember how I said that I was adding that code to the ironic driver? | |
| 21:18:46 | openstackgerrit | Matt Riedemann proposed openstack/nova master: api-ref: fix max_version for deprecated os-quota-class-sets parameters https://review.openstack.org/484952 | |
| 21:18:48 | mriedem | alex_xu: gmann: ^ we got confused | |
| 21:23:12 | openstackgerrit | Octave Orgeron proposed openstack/nova master: Enables MySQL Cluster Support for Nova https://review.openstack.org/446643 | |
| 21:24:07 | mriedem | edleafe: dansmith: now now now, you're BOTH wrong! :) | |
| 21:24:17 | mriedem | edleafe: you shouldn't be doing the placement stuff in the init_host | |
| 21:24:19 | mriedem | -1 for that | |
| 21:24:24 | mriedem | dan's change builds on top of yours | |
| 21:24:47 | mriedem | edleafe: just need you to update the existing instance flavors on init_host, | |
| 21:24:57 | mriedem | update_available_resource will pick up on that (via dan's change) and adjust the allocations | |
| 21:25:00 | mriedem | automagically | |
| 21:25:30 | openstackgerrit | Octave Orgeron proposed openstack/nova master: Enables MySQL Cluster Support for Nova https://review.openstack.org/446643 | |
| 21:29:51 | openstackgerrit | Matt Riedemann proposed openstack/nova master: placement: alloc candidates only shared resources https://review.openstack.org/484900 | |
| 21:30:01 | mriedem | jaypipes: +2 | |
| 21:30:14 | jaypipes | mriedem: danke | |
| 21:30:29 | mriedem | kein Problem | |
| 21:31:12 | openstackgerrit | Jackie Truong proposed openstack/nova master: Implement certificate_utils https://review.openstack.org/479949 | |
| 21:32:40 | openstackgerrit | Jackie Truong proposed openstack/nova master: Implement certificate_utils https://review.openstack.org/479949 | |
| 21:34:39 | edleafe | mriedem: like I said, I had most of that done before our little conversation earlier | |
| 21:35:05 | edleafe | I just pushed it before I gut it :) | |
| 21:39:11 | openstackgerrit | Ed Leafe proposed openstack/nova master: WIP - Migrate Ironic Flavors https://review.openstack.org/484949 | |
| 21:39:19 | openstackgerrit | Matt Riedemann proposed openstack/nova master: api-ref: fix description of service_disable_reason https://review.openstack.org/484955 | |
| 21:39:34 | edleafe | dansmith: mriedem: rebased it on dansmith's change and deleted the allocation stuff | |
| 21:39:36 | openstackgerrit | Rick Bartra proposed openstack/nova master: Add policy granularity to the Flavors API https://review.openstack.org/449288 | |
| 21:39:41 | edleafe | still needs tests, though | |
| 21:39:51 | edleafe | but they will have to wait until tomorrow | |
| 21:39:54 | mriedem | edleafe: ok, that's the wrong order really, but ok | |
| 21:40:24 | edleafe | mriedem: no, because without his you could get errors | |
| 21:40:31 | edleafe | with mine | |
| 21:40:51 | edleafe | with his first nothing will happen until mine | |
| 21:41:13 | mriedem | i don't understand how dan's change should impact yours | |
| 21:41:28 | mriedem | they could be done in parallel actually | |
| 21:41:30 | mriedem | but dan's builds on yours | |
| 21:41:55 | edleafe | well, maybe it only throws errors with the init_host updating inventory | |
| 21:42:05 | edleafe | didn't get to test that scenario | |
| 21:42:40 | melwitt | dansmith: I'm looking at the fleetify patch to figure out how to build upon it for per cell console proxies. it looks like it sets up only one cell? https://review.openstack.org/#/c/477556 | |
| 21:43:13 | openstackgerrit | Rick Bartra proposed openstack/nova master: Add policy granularity to the Flavors API https://review.openstack.org/449288 | |
| 21:43:38 | mriedem | melwitt: you have to look at the multinode job | |
| 21:43:48 | mriedem | http://logs.openstack.org/56/477556/5/check/gate-tempest-dsvm-neutron-multinode-full-ubuntu-xenial-nv/aed9486/ | |
| 21:44:03 | melwitt | mriedem: I did, I only see super conductor and one conductor process and NOVA_NUM_CELLS is 1? | |
| 21:44:29 | mriedem | hmm http://logs.openstack.org/56/477556/5/check/gate-tempest-dsvm-neutron-multinode-full-ubuntu-xenial-nv/aed9486/logs/devstack-gate-discover-hosts.txt.gz | |
| 21:44:48 | mriedem | yeah i see 2 computes, 2 cell mappings, but really only 1 cell | |
| 21:45:00 | dansmith | melwitt: it treats cell0 as the main cell by pointing everything to it, so cell1 is the different cell, | |