| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2021-09-03 | |||
| 08:13:53 | gibi | bauzas: yeah, that too | |
| 08:13:54 | bauzas | gibi: but I'm not sure I'll have time for this today | |
| 08:14:39 | gibi | me neighter | |
| 11:06:44 | ralonsoh | hi folks, I don't see a better channel to make this question | |
| 11:06:48 | ralonsoh | this is about ceph | |
| 11:06:55 | ralonsoh | (because I want live migration) | |
| 11:07:23 | ralonsoh | I have a node with a ceph cluster | |
| 11:07:49 | ralonsoh | when I try, from the other node, to execute "sudo ceph -c /etc/ceph/ceph.conf auth get-key client.cinder" | |
| 11:07:53 | ralonsoh | I receive | |
| 11:07:56 | ralonsoh | 2021-09-03T11:04:46.531+0000 7f8c47fff700 -1 monclient(hunting): handle_auth_bad_method server allowed_methods [2] but i only support [2] | |
| 11:08:01 | sean-k-mooney2 | ceph isnet needed for live migration but if you have it help avoid copying the disk | |
| 11:08:20 | ralonsoh | well, that was the reason | |
| 11:08:30 | ralonsoh | in any case, if I can't make it work, I'll skip that part | |
| 11:08:43 | sean-k-mooney2 | we spoke about this before | |
| 11:08:46 | ralonsoh | I had the same config as when you helped me last time | |
| 11:08:48 | ralonsoh | yes | |
| 11:08:59 | sean-k-mooney2 | have you disable the cve fix | |
| 11:09:11 | ralonsoh | do you have a link?? | |
| 11:09:16 | sean-k-mooney2 | for this https://docs.ceph.com/en/latest/security/CVE-2021-20288/#cve-2021-20288 | |
| 11:09:23 | sean-k-mooney2 | ill get the link to the workaround | |
| 11:09:36 | ralonsoh | yes, I did that | |
| 11:09:59 | sean-k-mooney | and you check the keyring file is not empty this time | |
| 11:10:17 | ralonsoh | yes, I have ceph.client.admin.keyring | |
| 11:10:22 | ralonsoh | and ceph.client.cinder.keyring | |
| 11:10:28 | ralonsoh | in /etc/ceph | |
| 11:11:07 | ralonsoh | sean-k-mooney, should I copy this files from the master one? | |
| 11:11:23 | sean-k-mooney | they should have the same content | |
| 11:11:27 | ralonsoh | ooook | |
| 11:11:43 | sean-k-mooney | i assume you have configured nova with the cinder keyring too | |
| 11:12:02 | ralonsoh | I used the devstack plugin | |
| 11:12:06 | ralonsoh | I think it is | |
| 11:12:20 | ralonsoh | sean-k-mooney, ok ok | |
| 11:12:23 | ralonsoh | I think that was the problem | |
| 11:12:30 | ralonsoh | (I don't know what I did before) | |
| 11:12:39 | ralonsoh | the keyrings where different... again | |
| 11:12:45 | ralonsoh | sean-k-mooney, again, thanks a lot!! | |
| 11:12:52 | sean-k-mooney | no worries | |
| 11:13:18 | sean-k-mooney | i dont use ceph often but the auth can be a bit of a pain to get working initally | |
| 11:13:41 | ralonsoh | sean-k-mooney, how do I configure Nova not to use ceph? | |
| 11:13:53 | ralonsoh | because if I remember correctly, this is required | |
| 11:14:01 | ralonsoh | for live-migration | |
| 11:14:02 | sean-k-mooney | you need to set the libvirt secrete | |
| 11:14:26 | ralonsoh | ah yes... do you have a link for this? | |
| 11:14:34 | ralonsoh | just to configure it myself | |
| 11:14:37 | sean-k-mooney | https://github.com/openstack/devstack-plugin-ceph/blob/master/devstack/lib/ceph#L242-L259 | |
| 11:14:46 | ralonsoh | ahhh yes | |
| 11:14:54 | ralonsoh | good stuff | |
| 11:15:02 | sean-k-mooney | the devstack plugin will do it for you but im guessing you only included that on the contoler | |
| 11:15:09 | sean-k-mooney | so you will have ot do that on the rest of the nodes | |
| 11:15:35 | ralonsoh | yes, I'll need to do it manually (just one compute, not too much hehehe) | |
| 11:15:38 | sean-k-mooney | ralonsoh: other then that you have to set the images_backend to rbd in the nova.conf | |
| 11:15:52 | ralonsoh | what is that? | |
| 11:16:15 | sean-k-mooney | https://github.com/openstack/devstack-plugin-ceph/blob/master/devstack/lib/ceph#L832-L839 | |
| 11:16:45 | sean-k-mooney | so set [libvirt]/images_type=rbd and configure the pool and ceph conf | |
| 11:16:53 | sean-k-mooney | you can copy that form the contoler nova.conf | |
| 11:17:08 | ralonsoh | but this should be done by the ceph plugin | |
| 11:17:16 | ralonsoh | I have it in the controller and the compute node | |
| 11:17:24 | sean-k-mooney | oh ok then yes it will be | |
| 11:17:35 | sean-k-mooney | i tought you had it only on the contoler | |
| 11:17:51 | ralonsoh | I think both nodes are online now, I'll test it | |
| 11:19:01 | ralonsoh | ok no | |
| 11:19:02 | ralonsoh | Sep 03 11:17:59 u20ovn2 nova-compute[887609]: ERROR nova.compute.manager nova.exception.StorageError: Storage error: Could not determine disk usage | |
| 11:19:28 | ralonsoh | I'll need to check that in the compute (controller does not list the compute as hypervisor yet) | |
| 11:19:49 | sean-k-mooney | yep same issue you had before | |
| 11:20:02 | sean-k-mooney | if nova cant talk to ceph the compute agent will fail to start properly | |
| 11:20:15 | ralonsoh | right | |
| 11:20:31 | sean-k-mooney | maybe just restart the agent | |
| 11:20:42 | sean-k-mooney | if the keyrings were wrong that would break it | |
| 11:21:28 | ralonsoh | and again something is overwritting the keyrings | |
| 11:21:36 | ralonsoh | ahh maybe the ceph plugin | |
| 11:22:00 | sean-k-mooney | ya i dont thin kthe ceph plugin has supprot for being used on the compute node | |
| 11:22:13 | ralonsoh | and here we are! all is working now | |
| 11:22:28 | sean-k-mooney | you shoudl be able to copy them and do "sudo systemctl restart devstack@n-cpu " | |
| 11:22:34 | sean-k-mooney | cool | |
| 12:45:27 | opendevreview | Merged openstack/nova master: Add --sleep option for archive_deleted_rows --until-complete https://review.opendev.org/c/openstack/nova/+/771749 | |
| 13:55:31 | stephenfin | gibi, lyarwood: Are we expecting to land https://review.opendev.org/c/openstack/nova/+/802060/ today? | |
| 13:55:46 | stephenfin | If not, that reno needs to come back and merge before we start branching | |
| 13:55:49 | gibi | stephenfin: I'm fine in both ways | |
| 13:55:56 | gibi | stephenfin: sure I will move the renoe | |
| 13:55:57 | gibi | reno | |
| 13:56:05 | gibi | or basically create a separate reno | |
| 13:56:11 | gibi | for the things that is landed | |
| 13:56:49 | stephenfin | Whatever suits. Depends on whether bauzas, lyarwood or melwitt will have time to review that nova-manage patch today | |
| 13:56:51 | gibi | I need to make a note about the fact the the neutron extension is not landed in Xena so this feature is not really usable | |
| 13:57:00 | stephenfin | The reno is noddy enough | |
| 13:57:03 | stephenfin | Ah, pity | |
| 13:57:20 | stephenfin | Hopefully that will happen early in Yoga | |
| 13:57:26 | gibi | yepp that is the plan now | |
| 13:57:29 | bauzas | that reminds me I accepted to provide a reno prelude :) | |
| 13:57:37 | gibi | bauzas: yepp :) | |
| 13:57:38 | stephenfin | In theory one could deploy Yoga neutron with Xena nova, right? | |
| 13:57:48 | gibi | but renos and the prelude can land until RC1 | |
| 13:58:01 | gibi | stephenfin: we made that combination working in this feature | |
| 13:58:16 | gibi | I have no info about the generic support of that combination | |
| 14:00:08 | gibi | I will do the reno reshuffling on monday | |
| 15:03:15 | sean-k-mooney | stephenfin: yes you can have newwer or older neutron with nova | |
| 15:03:50 | sean-k-mooney | so upstream at least we could have merged gibi pps qos work in nova in xena | |
| 15:04:01 | sean-k-mooney | and you could have used that with a yoga neutron | |
| 15:04:12 | sean-k-mooney | provide nothing changed in the neutron implemation | |
| 15:04:25 | sean-k-mooney | but we generally dont do that incase the api does change | |