| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2021-06-15 | |||
| 15:12:19 | sean-k-mooney | or vt-x | |
| 15:13:59 | sean-k-mooney | there is a script to check that | |
| 15:14:22 | pjakuszew | ubuntu has one iirc | |
| 15:14:28 | sean-k-mooney | virt-host-validate | |
| 15:14:37 | sean-k-mooney | or you can use kvm_ok on ubuntu | |
| 15:14:59 | sean-k-mooney | virt-host-validate is what you use on fedora/centos/rhel | |
| 15:15:06 | sean-k-mooney | although it might be on ubuntu also | |
| 15:15:26 | sean-k-mooney | i think its actully part of libvirt where as kvm_ok is part of cpu-checker i think | |
| 15:15:34 | sean-k-mooney | its not actully part of libvirt | |
| 15:17:18 | spatel | sean-k-mooney here its - http://paste.openstack.org/show/806634/ | |
| 15:17:20 | spatel | kvm-ok | |
| 15:17:45 | sean-k-mooney | yep so it deteced the problem | |
| 15:17:53 | sean-k-mooney | do you happen to have virt-host-validate also | |
| 15:18:12 | sean-k-mooney | it check more things but kvm_ok is nice for a quick check | |
| 15:18:42 | spatel | yes i do have virt-host-validate, it has nice output.. i didn't know about this | |
| 15:19:01 | spatel | i will add in my ansible playbook to validate VT before start deploying code | |
| 15:19:10 | sean-k-mooney | kvm-ok was what i always used to use only found out about virt-host-validate about a year ago | |
| 15:19:52 | pjakuszew | spatel: the hostname from your pastebin suggests that you probably will be interested in vt-d and intel_iommu=on as well | |
| 15:19:55 | sean-k-mooney | virt-host-validate will check if you can use sriov and a few other things also | |
| 15:20:29 | sean-k-mooney | intel_iommu=on iommu=pt most likely | |
| 15:20:30 | spatel | pjakuszew my all computes running on sriov | |
| 15:20:45 | sean-k-mooney | iommu=pt is not strictly requried if its only sriov and not dpdk | |
| 15:20:59 | spatel | i have added both anyway | |
| 15:22:59 | spatel | sean-k-mooney nowadays i am playing with OVN, it looks complicated but promissing | |
| 15:23:11 | spatel | promising* | |
| 15:23:29 | sean-k-mooney | it is on both counts | |
| 15:23:36 | sean-k-mooney | simpler then odl or onos | |
| 15:23:52 | sean-k-mooney | but more complex the simple sriov | |
| 15:24:20 | spatel | complex with sriov? | |
| 15:25:02 | spatel | i haven't started playing with OVN + SR-IOV yet. I am planning to use dpdk instead of sriov this time | |
| 15:26:41 | sean-k-mooney | no i ment sriov is relitvly simple in that its just a vlan on a vf in most case with simpel vlan networking | |
| 15:26:57 | spatel | oh! | |
| 15:27:37 | sean-k-mooney | ovn is more complex because of the need to under stand how the logical switchs in ovn mape too the neutron network and how ovn translattes the configuretion into openflow rules and the complexites of geneve tunneling | |
| 15:27:59 | sean-k-mooney | once you wrap your head around it its fine but there is a lot more context to understand with ovn | |
| 15:29:07 | spatel | no doubt about complexity. it has lots of moving components in single piece. l2/l3/firewall/dhcp so it looks simple but kinda hard to troubleshoot. | |
| 15:30:11 | spatel | sometime its very hard to read flow table for troubleshooting, specially when it has hundreds of entries. | |
| 15:31:42 | spatel | but i like all the feature it providing like DVR etc.. which is very easy and it won't waste of public IPs unlike legacy DVR | |
| 15:35:51 | gibi | fyi I've opened a new gate-failure bug https://bugs.launchpad.net/nova/+bug/1932043 nova-ceph-multinode test_resize_server_revert fails with rbd.ReadOnlyImage: [errno 30] RBD read-only image | |
| 15:35:53 | melwitt | elodilles: if I could trouble you for review on https://review.opendev.org/c/openstack/nova/+/794186 and https://review.opendev.org/c/openstack/nova/+/791577 please if you get a chance | |
| 15:36:49 | sean-k-mooney | gibi: strange was the a new cpeh release or soemthign | |
| 15:37:01 | sean-k-mooney | e.g. did somethign cause a change in the behaivor | |
| 15:37:25 | gibi | sean-k-mooney: I saw it 3 times in the last couple of days, so it is not a total blocker | |
| 15:37:27 | sean-k-mooney | im not sure why the image would be ReadOnlyImage | |
| 15:37:44 | sean-k-mooney | ok | |
| 15:37:54 | melwitt | lyarwood: and if I could trouble you for review on https://review.opendev.org/c/openstack/nova/+/791577 please if you get a chance | |
| 15:44:38 | gibi | sean-k-mooney: I don't see ceph version differences between failed and successfull jobs both using nautilus | |
| 15:45:56 | sean-k-mooney | ack do you know why an image would be read only? | |
| 15:46:21 | sean-k-mooney | i can only assume its a permissions issue but i woudl expect taht to affect all vms in the same tempest run | |
| 15:48:03 | gibi | sean-k-mooney: no idea. it happens during a same host resize | |
| 15:49:54 | sean-k-mooney | that does a direct snapshot of the volumn in ceph | |
| 15:50:17 | sean-k-mooney | at least when using the rbd image backend in nova | |
| 15:50:30 | gibi | FYI nova meeting starts in 10 minutes _in this channel_ | |
| 15:50:38 | sean-k-mooney | :) | |
| 15:51:14 | gibi | sean-k-mooney: yes it does a snapshot during finish_migration and that fails | |
| 15:52:06 | pjakuszew | I have a question about extra_specs mangling but I guess I will ask about it after the meeting :) | |
| 15:52:33 | sean-k-mooney | pjakuszew: we dont actully mangel extra specs in the db as far as im aware | |
| 15:52:55 | sean-k-mooney | we do prefix image properties with image_ in the system metadata table | |
| 15:53:05 | pjakuszew | uh I phrased it wrong | |
| 15:53:34 | pjakuszew | this is something I mentioned ~last week here | |
| 15:53:50 | pjakuszew | "regenerate extra_specs in requestspec" topic | |
| 15:53:51 | sean-k-mooney | regenerating the request spec? | |
| 15:53:53 | pjakuszew | yeah | |
| 15:54:14 | sean-k-mooney | we store an copy of the entire flavor as a json bob i think | |
| 15:54:24 | pjakuszew | so, I noticed that Instance.flavor seems to be a different entity stored somewhere else | |
| 15:54:30 | pjakuszew | http://paste.openstack.org/show/806638/ | |
| 15:54:32 | sean-k-mooney | in instnace_extra maybe havent checked in a while | |
| 15:55:14 | pjakuszew | and I wonder where Instance.flavor is actually stored, if it is not looked up from either RequestSpec, or the actual Flavor | |
| 15:55:40 | sean-k-mooney | pjakuszew: yes when we create an instance we make a copy of the flavor and extra specs an store a copy to use for the lifetime of the instace | |
| 15:56:00 | sean-k-mooney | to intentionally not pick up any change to extra specs that operators do after the vm is created | |
| 15:56:33 | sean-k-mooney | https://github.com/openstack/nova/blob/master/nova/db/sqlalchemy/models.py#L389 | |
| 15:56:43 | opendevreview | Stephen Finucane proposed openstack/nova master: db: Remove dead code https://review.opendev.org/c/openstack/nova/+/786291 | |
| 15:56:46 | sean-k-mooney | its sotred in the instance_extra db tables as a json blob | |
| 15:56:47 | opendevreview | Stephen Finucane proposed openstack/nova master: db: Remove 'nova.db.sqlalchemy.utils' https://review.opendev.org/c/openstack/nova/+/786292 | |
| 15:56:50 | opendevreview | Stephen Finucane proposed openstack/nova master: db: Remove unused DB methods https://review.opendev.org/c/openstack/nova/+/786293 | |
| 15:56:52 | opendevreview | Stephen Finucane proposed openstack/nova master: db: Use module-level imports for sqlalchemy https://review.opendev.org/c/openstack/nova/+/786295 | |
| 15:56:54 | opendevreview | Stephen Finucane proposed openstack/nova master: db: Fold in indexes https://review.opendev.org/c/openstack/nova/+/786296 | |
| 15:56:57 | opendevreview | Stephen Finucane proposed openstack/nova master: db: Fold in ForeignKey constraints https://review.opendev.org/c/openstack/nova/+/786297 | |
| 15:57:01 | opendevreview | Stephen Finucane proposed openstack/nova master: db: Remove 'nova.db.base' module https://review.opendev.org/c/openstack/nova/+/786298 | |
| 15:57:07 | opendevreview | Stephen Finucane proposed openstack/nova master: db: Copy docs from 'nova.db.*' to 'nova.db.sqlalchemy.*' https://review.opendev.org/c/openstack/nova/+/786299 | |
| 15:57:11 | opendevreview | Stephen Finucane proposed openstack/nova master: db: Synchronize function signatures https://review.opendev.org/c/openstack/nova/+/786300 | |
| 15:57:18 | opendevreview | Stephen Finucane proposed openstack/nova master: db: Clean up migration code https://review.opendev.org/c/openstack/nova/+/786301 | |
| 15:57:21 | opendevreview | Stephen Finucane proposed openstack/nova master: db: Use module-level imports for sqlalchemy (for real) https://review.opendev.org/c/openstack/nova/+/796519 | |
| 15:57:33 | sean-k-mooney | lol i just realise that that is one drawback of having the meeting here | |
| 15:57:40 | pjakuszew | sean-k-mooney: huh ok, I thought that InstanceExtra was the RequestSpec actually | |
| 15:57:51 | sean-k-mooney | pjakuszew: no that is in the api db | |
| 15:58:01 | sean-k-mooney | instance extra is in the cell db | |
| 15:58:27 | sean-k-mooney | https://github.com/openstack/nova/blob/master/nova/db/sqlalchemy/api_models.py#L169-L181 | |
| 15:58:44 | pjakuszew | hm, okay, and what is taken into consideration when a migration is requested? | |
| 15:58:48 | pjakuszew | the cell db, or api db? | |
| 15:58:57 | sean-k-mooney | both | |
| 15:59:10 | sean-k-mooney | the flavor will come form the instance_extra table | |
| 15:59:26 | sean-k-mooney | schduler hints and other thing like az will come form teh request spec | |
| 15:59:30 | pjakuszew | right, so if I want to make it behave like new instances, should I change both objects? | |
| 15:59:33 | sean-k-mooney | but lets pick this up after the meeting | |
| 15:59:36 | pjakuszew | sure thing | |
| 16:00:03 | gibi | #startmeeting nova | |
| 16:00:03 | opendevmeet | Meeting started Tue Jun 15 16:00:03 2021 UTC and is due to finish in 60 minutes. The chair is gibi. Information about MeetBot at http://wiki.debian.org/MeetBot. | |
| 16:00:03 | opendevmeet | Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. | |
| 16:00:03 | opendevmeet | The meeting name has been set to 'nova' | |
| 16:00:13 | gibi | o/ | |