| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-05-24 | |||
| 19:33:28 | ckopper | that adds a flavor extra_spec `hw:hide_kvm` to set the correct options through libvirt and was wondering what the correct procedure to get this proposed for upstream was. | |
| 19:33:28 | ckopper | Hi! I am new to nova and openstack development, so I apologize in advance if this is the wrong place to ask or if there is a certain procedure for this. I have a local use case for OpenStack where we want to use PCI passthrough for consumer GPUs which works well in principle. However the NVidia driver checks if it is run within a VM and refuses to load for consumer cards if you don't set a certain KVM/qemu flag (`kvm=off`). I have a local patch | |
| 19:33:33 | ckopper | The patch itself is pretty trivial (mostly tests): https://gist.github.com/claudiok/5dd9fd01b9628e35f0b48884471e58cf | |
| 19:34:02 | ckopper | I assume this needs a spec and blueprint and then submission to gerrit for review but I thought I would check here first | |
| 19:34:16 | openstackgerrit | Sean Dague proposed openstack/nova master: WIP: request_id passing to neutron https://review.openstack.org/467242 | |
| 19:38:07 | mriedem | ckopper: lucky for you someone is already working on that https://review.openstack.org/#/c/459753/ | |
| 19:38:55 | ckopper | Haha - I have to admit I searched for this but apparently missed this review | |
| 19:39:08 | mriedem | sdague: yup, i have a functional test but b/c of sqlite, doing a GET /servers/id with upper-case returns 404 :( | |
| 19:39:35 | sdague | mriedem: yep | |
| 19:39:46 | mriedem | so, i'm not sure how to hack this | |
| 19:40:22 | sdague | so, there is the right way, which is figure out how to be able to run some tests against a mysql backend directly | |
| 19:40:32 | mriedem | i could stub the pull from the db to change the case on what is returned from the db so what's in the cache is different from what's in the request | |
| 19:40:40 | sdague | right, please don't do that | |
| 19:41:02 | sdague | honestly, any energy doing crazy stuff like that should be put into real mysql support in the test suite | |
| 19:41:23 | sdague | the biggest issue was the way the magic oslo.db stuff worked for that iirc | |
| 19:41:30 | mriedem | i'll poke | |
| 19:41:33 | sdague | it's honestly I think better to just ship the patch | |
| 19:41:40 | mriedem | maybe i can just extend test_base.MySQLOpportunisticTestCase | |
| 19:41:59 | sdague | mriedem: there was something weird about that when I tried | |
| 19:42:08 | melwitt | too bad there's not a sqlite pragma for that | |
| 19:43:32 | sdague | mriedem: https://review.openstack.org/#/c/283364/ that's a restore of an attempt to go down that path | |
| 19:43:39 | sdague | maybe it was working, but the fix never landed | |
| 19:44:04 | sdague | mriedem: the issue was about tricking engine fascade I think | |
| 19:44:39 | mriedem | trying it | |
| 19:44:42 | sdague | mriedem: anyway, that might be useful to build the pattern from | |
| 19:45:13 | mriedem | which mysql package do we need? | |
| 19:45:16 | mriedem | mysql-server? | |
| 19:46:10 | mriedem | bindep has both the server and client | |
| 19:46:16 | mriedem | but for some reason they fail to install on 16.04 | |
| 19:46:50 | mriedem | ah here we go | |
| 19:49:01 | ckopper | mriedem: concerning the img_hide_hypervisor_id feature: there one more flag that needs to be set to hide the hypervisor for Windows VMs (not necessary for Linux). (<hyperv><vendor_id state="on" value="FakeID"/></hyperv>). Is this something that should go into a separate patch or would it be appropriate to add this as a comment to the current code review? | |
| 19:49:18 | ckopper | (thanks again for pointing me to it!) | |
| 19:49:21 | mriedem | ckopper: please comment on the review | |
| 19:49:25 | ckopper | thanks | |
| 19:52:52 | mriedem | sdague: do you remember if you had to do any special creds tweaking with the mysql server or anything locally to get it to connect? | |
| 19:53:05 | mriedem | i thought there was a special userid/password or something | |
| 19:54:01 | mriedem | with user 'openstack_citest' and password 'openstack_citest' on localhost. | |
| 19:54:01 | mriedem | For the opportunistic testing you need to set up db's named 'openstack_citest' | |
| 19:54:01 | mriedem | ah | |
| 19:55:42 | sdague | yep, that | |
| 19:55:50 | mriedem | also, tools/test-setup.sh | |
| 19:58:28 | cdent | mriedem, sdague, melwitt (whatever other cores are around): this https://review.openstack.org/#/c/456198/ is a fix to tooling used by the placement api-ref that makes the tooling work with multiple inc files instead of just one monolithic file. getting it merged will allow some placement api-ref work to proceed more easily without needing to be in the same stack | |
| 19:59:08 | sdague | +2 | |
| 19:59:19 | cdent | thanks | |
| 20:01:11 | cdent | thanks melwitt and dansmith | |
| 20:01:30 | dansmith | two +Ws means you skip the gate I think | |
| 20:01:35 | cdent | \o/ | |
| 20:01:43 | melwitt | heh, secret code | |
| 20:08:28 | mriedem | sdague: EngineFacadeFixture isn't in nova, is that part of your patch series somewhere? | |
| 20:09:26 | mriedem | huh removed here | |
| 20:09:27 | mriedem | https://review.openstack.org/#/c/434843/2 | |
| 20:10:45 | sdague | mriedem: yeh, so I guess there is an oslo.db solution now? | |
| 20:10:51 | mriedem | yeah looking at ReplaceEngineFacadeFixture | |
| 20:12:54 | mriedem | it at least connected to my local mysql server which is progress | |
| 20:15:41 | melwitt | dansmith: is this true that online_data_migrations runs in the background after the command exits? someone working on a man page entry https://review.openstack.org/#/c/464634 | |
| 20:15:54 | dansmith | no | |
| 20:16:20 | melwitt | okay, I didn't think so | |
| 20:16:21 | mriedem | there is no scheduling | |
| 20:16:24 | mriedem | or queueing | |
| 20:17:13 | dansmith | which is kinda the point | |
| 20:18:01 | openstackgerrit | Kevin Benton proposed openstack/nova master: WIP: Bulk query floating IPs in _nw_info_get_ips https://review.openstack.org/465792 | |
| 20:21:14 | mriedem | i'm lost as to how to use the oslo.db opportunistic test fixture stuff | |
| 20:24:17 | melwitt | this is all I found about it https://specs.openstack.org/openstack/oslo-specs/specs/kilo/long-lived-transactionalized-db-fixtures.html#opportunistic-urls | |
| 20:26:04 | sdague | mriedem: if you put up a partial patch, I might play in the morning | |
| 20:27:30 | mriedem | ok yeah b/c i'm going to have to give up on this soon | |
| 20:28:17 | sdague | that's fine, having something to chew on over coffee in the morning is usually useful | |
| 20:42:58 | mriedem | zzzeek: i'm strugglying trying to use OpportunisticDBTestMixin to get a mysql-backed test running | |
| 20:43:19 | mriedem | i've got a test extending OpportunisticDBTestMixin and defining FIXTURE = test_fixtures.MySQLOpportunisticFixture | |
| 20:43:26 | zzzeek | mriedem: OK...are there no existing tests using it? | |
| 20:43:28 | mriedem | but teardown complains about not having a schema attribute on the test | |
| 20:43:30 | mriedem | nope | |
| 20:43:33 | mriedem | well, | |
| 20:43:48 | mriedem | our db schema migration tests do some stuff, but i can't make heads or tails of those | |
| 20:44:13 | mriedem | going through the setup code in oslo.db totally gets me lost | |
| 20:44:29 | zzzeek | mriedem: yeah | |
| 20:44:42 | zzzeek | oslo_test_base.BaseTestCase ? | |
| 20:44:42 | zzzeek | mriedem: are you mixing in test_fixtures.OpportunisticDBTestMixin, | |
| 20:44:57 | mriedem | not the BaseTestCase | |
| 20:45:25 | zzzeek | mriedem: try sticking "SCHEMA_SCOPE=None" on your test class ? | |
| 20:46:15 | mriedem | trying it | |
| 20:46:36 | zzzeek | mriedem: hmmm that's not it | |
| 20:46:40 | zzzeek | mriedem: stack trace ? | |
| 20:48:13 | zzzeek | mriedem: oh. this is all deprecated stuff anyway. you on newish oslo.db ? | |
| 20:48:21 | mriedem | yeah i'm trying to use the fixtures | |
| 20:48:22 | zzzeek | mriedem: e.g. dont need to backport to mitaka ? | |
| 20:48:33 | mriedem | with our nova base TestCase code | |
| 20:48:36 | mriedem | which has it's own DatabaseFixture | |
| 20:48:58 | mriedem | i started copying this https://review.openstack.org/#/c/283364/6/nova/tests/functional/regressions/test_bug_1538011.py but that's extending a deprecated test class in oslo.db | |
| 20:49:08 | zzzeek | mriedem: yeah i was goign to go in there and change it all a while back, a lot of this fixture stuff had nova in mind. | |
| 20:50:01 | mriedem | getting this back to where i started and then i can push it up | |
| 20:51:09 | jose-phillips | hey | |
| 20:51:27 | jose-phillips | someone now how to set a metadata with python novaclient | |
| 20:51:29 | zzzeek | mriedem: yeah i need stack trace to work w/ this here | |
| 20:51:38 | zzzeek | mriedem: need to see where it looks for .schema | |
| 20:51:51 | jose-phillips | should be nova.aggregates.set_metadata (openstackId,metadata) ? | |
| 20:51:59 | zzzeek | mriedem: id guess your fixture is mixing in stuff from your existing fixture stuff ? | |
| 20:52:19 | mriedem | zzzeek: yeah | |
| 20:52:50 | mriedem | hmm, maybe this worked for https://review.openstack.org/#/c/283364/6/nova/tests/functional/regressions/test_bug_1538011.py because it wasn't pulling in our TestCase default setup in nova | |
| 20:53:03 | zzzeek | mriedem: you'd need to start totally clean first. i had a plan to totally rewrork nova's fixtures at some point | |
| 20:53:05 | mriedem | which means i'd have to reinvent some things in my test | |