| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2022-02-04 | |||
| 12:24:40 | sean-k-mooney | i was just looking for it for the topic/commit message | |
| 12:35:57 | opendevreview | sean mooney proposed openstack/nova master: skip test_tagged_attachment in nova-next https://review.opendev.org/c/openstack/nova/+/827851 | |
| 12:36:37 | sean-k-mooney | ok ya i think ^ is all that is needed | |
| 12:42:49 | opendevreview | Merged openstack/nova master: Reproduce bug 1952941 https://review.opendev.org/c/openstack/nova/+/820121 | |
| 12:44:12 | gibi | sean-k-mooney: thanks | |
| 12:44:28 | gibi | sean-k-mooney: I will fast approve it as soon as zuul is green on the patch | |
| 12:52:50 | artom_ | gibi, sean-k-mooney, I hadn't, so thanks! | |
| 12:58:13 | opendevreview | Attila Fazekas proposed openstack/nova master: Document the floating ip behavior in short https://review.opendev.org/c/openstack/nova/+/827856 | |
| 13:28:47 | dmitriis | gibi: looking | |
| 13:31:06 | tobias-urdin | i have a weird issue, back in december 2021 a resize that failed, also failed to cleanup /var/lib/instances/ |
|
| 13:31:06 | tobias-urdin | logs https://paste.opendev.org/show/812522/ | |
| 13:31:33 | tobias-urdin | the first issue was because of a libvirt issue (out of fd and socket issues) | |
| 13:32:33 | tobias-urdin | don't have logs to understand if for example SELinux blocked the file deletion, does it make sense to insert something that would catch the issue and remove the _resize directory before trying the rename? since the API would ensure there can't be an race condition for resizes | |
| 13:34:34 | gibi | tobias-urdin: do I understand correctly that the original delete of the dir faild and this now prevents the retry of the resize? I think we can robustify nova by retry the delete during the next resize if it sees the stale dir there | |
| 13:51:42 | tobias-urdin | gibi: yes, resize failed because of leftover <uuid>_resize directory in instances directory | |
| 14:02:59 | opendevreview | Tobias Urdin proposed openstack/nova master: Cleanup _resize directory in instances dir on resize https://review.opendev.org/c/openstack/nova/+/827865 | |
| 14:03:15 | tobias-urdin | gibi: i guess it could be as simple as ^ | |
| 14:12:42 | gibi | tobias-urdin: that could be it, but would be nice to add a test around it | |
| 14:13:14 | gibi | tobias-urdin: and bug report for tracking purposes :) | |
| 14:16:13 | tobias-urdin | yeah i'll fix everything up if it feels acceptable (bug report, test, release note etc) - should be very low risk backport as well | |
| 14:16:28 | gibi | tobias-urdin: I agree this feels a safe thing to do | |
| 14:17:04 | sean-k-mooney | i have not beeing following but you will need to ensuer its not a shared file system | |
| 14:17:12 | sean-k-mooney | otherwise the cleanup would delete the vm | |
| 14:17:45 | gibi | sean-k-mooney: it is the _resize dir I think that is being created there, so deleting the dir before we creating it feels save | |
| 14:17:48 | gibi | safe | |
| 14:18:10 | opendevreview | Merged openstack/nova master: Migrate RequestSpec.numa_topology to use pcpuset https://review.opendev.org/c/openstack/nova/+/820153 | |
| 14:18:17 | sean-k-mooney | ah ok we have a parallel dir that we then move into the final location later? | |
| 14:18:51 | sean-k-mooney | os.rename(inst_base, inst_base_resize) | |
| 14:19:03 | sean-k-mooney | so that is moving inst_base to inst_base_resize | |
| 14:19:17 | gibi | yepp, to save it if we want to roll back | |
| 14:19:38 | sean-k-mooney | ok so ya that should be safe | |
| 14:19:59 | gibi | tobias-urdin: thanks for taking care of | |
| 14:20:02 | gibi | it | |
| 14:30:55 | opendevreview | Balazs Gibizer proposed openstack/nova stable/xena: Reproduce bug 1952941 https://review.opendev.org/c/openstack/nova/+/827868 | |
| 14:30:56 | opendevreview | Balazs Gibizer proposed openstack/nova stable/xena: Migrate RequestSpec.numa_topology to use pcpuset https://review.opendev.org/c/openstack/nova/+/827869 | |
| 14:34:11 | opendevreview | Balazs Gibizer proposed openstack/nova stable/wallaby: Reproduce bug 1952941 https://review.opendev.org/c/openstack/nova/+/827870 | |
| 14:34:12 | opendevreview | Balazs Gibizer proposed openstack/nova stable/wallaby: Migrate RequestSpec.numa_topology to use pcpuset https://review.opendev.org/c/openstack/nova/+/827871 | |
| 14:37:14 | opendevreview | Balazs Gibizer proposed openstack/nova stable/victoria: Reproduce bug 1952941 https://review.opendev.org/c/openstack/nova/+/827872 | |
| 14:37:15 | opendevreview | Balazs Gibizer proposed openstack/nova stable/victoria: Migrate RequestSpec.numa_topology to use pcpuset https://review.opendev.org/c/openstack/nova/+/827873 | |
| 15:23:30 | opendevreview | Merged openstack/nova master: Revert "Revert resize: wait for events according to hybrid plug" https://review.opendev.org/c/openstack/nova/+/819494 | |
| 15:43:14 | opendevreview | Merged openstack/nova stable/wallaby: [rt] Apply migration context for incoming migrations https://review.opendev.org/c/openstack/nova/+/820555 | |
| 16:08:39 | opendevreview | Dmitrii Shcherbakov proposed openstack/nova master: Add supports_remote_managed_ports capability https://review.opendev.org/c/openstack/nova/+/827839 | |
| 16:08:39 | opendevreview | Dmitrii Shcherbakov proposed openstack/nova master: [yoga] Add support for VNIC_TYPE_SMARTNIC https://review.opendev.org/c/openstack/nova/+/824835 | |
| 16:08:40 | opendevreview | Dmitrii Shcherbakov proposed openstack/nova master: Filter computes without remote-managed ports early https://review.opendev.org/c/openstack/nova/+/812111 | |
| 16:17:01 | dmitriis | gibi: re-uploaded with the cases mentioned covered | |
| 16:17:27 | dmitriis | gibi: hopefully the approach in general is OK | |
| 16:36:20 | gibi | dmitriis: thanks. I will check it on monday | |
| 16:36:23 | gibi | have a nice weekend | |
| 16:37:14 | dmitriis | gibi: same to you, cheers o/ | |
| 16:37:31 | gibi | o/ | |
| 17:42:56 | opendevreview | Federico Ressi proposed openstack/nova master: Debug Nova APIs call failures https://review.opendev.org/c/openstack/nova/+/806683 | |
| 18:32:55 | chateaulav | has anyone run into issues with libvirt in rgards to `error: failed to get CPU model names` | |
| 18:37:47 | sean-k-mooney | i hit it yesterday when i set an incorrct value on an arm system | |
| 18:38:08 | sean-k-mooney | i put virt for [libvirt]/cpu_model | |
| 18:38:20 | sean-k-mooney | it should have been max or host | |
| 18:38:41 | sean-k-mooney | so if you use virt-type=qemu and cpu_model=virt it will raise that error | |
| 18:39:43 | sean-k-mooney | i shoudl have had virt_type=qemu, cpu_mode=custom and cpu_model=max with machine_type=aarch64:virt or set on the image | |
| 18:39:53 | chateaulav | sean-k-mooney: it relates to riscv currently. I can run emulation using the qemu binary directly; however, run into issue because virsh doesnt see any available cpu models | |
| 18:40:33 | chateaulav | https://www.irccloud.com/pastebin/SoDG0HXa/error-risc-v | |
| 18:42:01 | sean-k-mooney | if you do qemu-system-riscv64 --cpu ? | |
| 18:42:08 | sean-k-mooney | it will list the supported ones | |
| 18:42:43 | sean-k-mooney | home ok libvirt may not have support for that yet | |
| 18:42:53 | sean-k-mooney | you could try rv64 | |
| 18:43:06 | sean-k-mooney | for the cpu_model | |
| 18:43:13 | sean-k-mooney | and see if it work if you spcify one | |
| 18:43:20 | chateaulav | yeah no luck, been going back and forth and though i may have missed something. | |
| 18:43:47 | sean-k-mooney | no it sound like libvirt is missing supprt | |
| 18:45:07 | chateaulav | yeah thats what i was thinking to | |
| 18:45:12 | sean-k-mooney | it might work in a later libvirt | |
| 18:45:16 | sean-k-mooney | what version are you using | |
| 18:45:27 | sean-k-mooney | the most recent release is 8.0.0 i belive | |
| 18:45:55 | chateaulav | libvirtd (libvirt) 6.0.0 | |
| 18:46:33 | sean-k-mooney | i mean its ment to have supprot | |
| 18:46:54 | sean-k-mooney | i see stuff as far back as 4.7.0 | |
| 18:46:57 | sean-k-mooney | in the release notes | |
| 18:47:08 | chateaulav | it seems like partial support, gonna test along that route | |
| 18:47:50 | sean-k-mooney | maybe they dont support emulation with riscv | |
| 18:47:56 | sean-k-mooney | just native supprot? | |
| 18:49:47 | chateaulav | I feel that as well, a couple last checks and then ill put a nail in it | |
| 19:56:23 | opendevreview | Artom Lifshitz proposed openstack/nova master: DNM: Run Nova gate on Neutron notifier patch https://review.opendev.org/c/openstack/nova/+/827917 | |
| 19:59:09 | opendevreview | Ade Lee proposed openstack/nova stable/xena: Add check job for FIPS https://review.opendev.org/c/openstack/nova/+/827895 | |
| 20:00:39 | opendevreview | Ade Lee proposed openstack/nova stable/xena: Add check job for FIPS https://review.opendev.org/c/openstack/nova/+/827895 | |
| 20:01:28 | opendevreview | Ade Lee proposed openstack/nova stable/wallaby: Add check job for FIPS https://review.opendev.org/c/openstack/nova/+/827896 | |
| 20:03:10 | melwitt | dansmith: hi :D I have been working on removing tracebacks in the logs in nova-multi-cell that happen when a neutron network-changed event is received after the instance has moved to another cell during a cross-cell migration, | |
| 20:04:32 | melwitt | this seemed simple at first until the fact that the exception is being raised from the db access on nova-conductor but the tracing is happening in nova-compute, | |
| 20:05:24 | sean-k-mooney | it helps if you enable forwarding on the host if you want nat to work for your vms.... sudo sysctl net.ipv4.conf.all.forwarding=1 | |
| 20:05:44 | dansmith | melwitt: because of indirection I assume | |
| 20:06:08 | melwitt | I'm seeing tracing still from the rpc messaging layer, so I tried this https://review.opendev.org/c/openstack/nova/+/826942/5/nova/objects/instance_info_cache.py#77 but it doesn't work as expected. I get oslo_messaging.rpc.client.RemoteError: Remote error: ExpectedException on the client side instead of InstanceNotFound | |
| 20:06:58 | melwitt | I had thought when oslo.messaging receives ExpectedException on the client side it translates it into the original exception type that it wrapped on the server side, but that is not occurring | |
| 20:07:56 | dansmith | no, I think that just avoids the traceback on the conductor side | |
| 20:08:14 | melwitt | hm ok | |
| 20:08:21 | sean-k-mooney | melwitt: i assume you are not checkign if the instance exists then updating because it could get deleted in between | |
| 20:08:28 | sean-k-mooney | which is why you are trying to catch the excpiton | |
| 20:08:34 | melwitt | sean-k-mooney: yeah | |
| 20:09:06 | melwitt | well, I'm catching DBReferenceError and then raising InstanceNotFound to be handled by upper layers | |
| 20:09:49 | sean-k-mooney | do the upper layers need to do anything with the not found | |
| 20:09:57 | sean-k-mooney | the instance has been moved to the other cell yes | |
| 20:10:50 | melwitt | most things handle not found, like "the instance disappeared while <whatever>" | |
| 20:11:07 | melwitt | if someone deletes while it's in the middle of <whatever> | |
| 20:11:36 | melwitt | it's working fine except I'm still getting tracebacks in nova-compute coming from oslo.messaging and I can't get it to stop | |