| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2023-01-25 | |||
| 15:39:55 | bauzas | artom: I'm glad that the litterally first message you write here today is an emoji | |
| 15:40:13 | bauzas | artom: and I'm happy this isn't a poop one | |
| 15:40:57 | artom | I'm like emoji batman | |
| 15:43:11 | bauzas | a wealthy man that hides his emitions behind a mask and wears underwear on top of his pants ? | |
| 15:44:12 | artom | Well, one of those is true | |
| 15:45:16 | bauzas | I hope it's the former, I'm afraid it could be the latter | |
| 15:46:43 | gibi | too much details :D | |
| 15:47:05 | bauzas | yay | |
| 15:47:17 | bauzas | gibi: you were right, it sounds we have a problem with grenade on xena | |
| 15:47:38 | gibi | bauzas: did it fail again? | |
| 15:47:41 | bauzas | yup | |
| 15:47:46 | gibi | with the same fastener dep issue? | |
| 15:47:56 | bauzas | https://zuul.opendev.org/t/openstack/build/d8e11efbdc6d43e7b9e27273cf038786 | |
| 15:48:50 | gibi | /o\ I have a hunch | |
| 15:49:31 | gibi | https://review.opendev.org/c/openstack/tempest/+/821732/21/requirements.txt yeah we landed this | |
| 15:49:45 | bauzas | 2023-01-25 14:45:10.253 | ERROR: Could not find a version that satisfies the requirement fasteners>=0.16.0 | |
| 15:49:55 | bauzas | https://00f8d73ac2d869c11924-90ff9157beec64657d8a46242c5af814.ssl.cf1.rackcdn.com/871557/3/check/nova-grenade-multinode/d8e11ef/controller/logs/grenade.sh_log.txt | |
| 15:50:28 | bauzas | and I was wrong, this is on wallaby, not xena | |
| 15:50:43 | bauzas | -ETOOMANYBRANCHESANDPATCHESTOTRACK | |
| 15:51:04 | gibi | what is the last stable branch tempest master supports? | |
| 15:51:23 | gibi | I feel like wallaby is at the boundary | |
| 15:51:29 | bauzas | possibly | |
| 15:51:30 | gibi | gmann: ^^ | |
| 15:51:42 | bauzas | but we also have problem with ceph-multistore on yoga | |
| 15:51:50 | bauzas | changing my focus | |
| 15:51:57 | bauzas | https://review.opendev.org/c/openstack/nova/+/871624 failed again | |
| 15:52:15 | gibi | gmann: it seems https://review.opendev.org/c/openstack/tempest/+/821732 affects stable/wallaby grenade runs | |
| 15:52:46 | gmann | gibi tempest master stopped wallaby support recently. it is stable/xena the last stable it support | |
| 15:53:15 | gmann | gibi: but I have not pinned stable/wallaby with old compatible tempest which I should do | |
| 15:53:39 | gibi | gmann: I see so stable/wallaby runs with master tempest and has https://review.opendev.org/c/openstack/tempest/+/821732 but it shoudl not run with master tempest any more | |
| 15:53:41 | gmann | till now it was running fine but if it is breaking its time to pin tempest there | |
| 15:53:58 | gibi | gmann: yes, it breaks now on the fasteners >= 0.16 dependency | |
| 15:54:06 | gibi | that https://review.opendev.org/c/openstack/tempest/+/821732 introduced | |
| 15:54:09 | gmann | gibi: yeah. I will pin it today | |
| 15:54:16 | gibi | gmann: thank you! | |
| 16:07:08 | opendevreview | Merged openstack/nova stable/zed: [stable-only][cve] Check VMDK create-type against an allowed list https://review.opendev.org/c/openstack/nova/+/871616 | |
| 16:25:29 | bauzas | sean-k-mooney: +2d sahid's implementation of stopping evacuated instances | |
| 16:35:00 | opendevreview | Balazs Gibizer proposed openstack/nova stable/wallaby: Reproduce bug 1981813 in func env https://review.opendev.org/c/openstack/nova/+/859320 | |
| 16:35:01 | opendevreview | Balazs Gibizer proposed openstack/nova stable/wallaby: Gracefully ERROR in _init_instance if vnic_type changed https://review.opendev.org/c/openstack/nova/+/859321 | |
| 16:38:54 | opendevreview | Balazs Gibizer proposed openstack/nova stable/victoria: Reproduce bug 1981813 in func env https://review.opendev.org/c/openstack/nova/+/869583 | |
| 16:38:55 | opendevreview | Balazs Gibizer proposed openstack/nova stable/victoria: Gracefully ERROR in _init_instance if vnic_type changed https://review.opendev.org/c/openstack/nova/+/869584 | |
| 16:42:33 | opendevreview | Balazs Gibizer proposed openstack/nova stable/ussuri: Reproduce bug 1981813 in func env https://review.opendev.org/c/openstack/nova/+/869585 | |
| 16:42:34 | opendevreview | Balazs Gibizer proposed openstack/nova stable/ussuri: Gracefully ERROR in _init_instance if vnic_type changed https://review.opendev.org/c/openstack/nova/+/869586 | |
| 16:52:01 | sahid | thank you bauzas ++ | |
| 16:59:22 | opendevreview | Balazs Gibizer proposed openstack/nova stable/train: Reproduce bug 1981813 in func env https://review.opendev.org/c/openstack/nova/+/869673 | |
| 16:59:23 | opendevreview | Balazs Gibizer proposed openstack/nova stable/train: Gracefully ERROR in _init_instance if vnic_type changed https://review.opendev.org/c/openstack/nova/+/869674 | |
| 17:00:08 | bauzas | sahid: I'm really sorry, but I forgot to look at your dependent patch and I found something :( | |
| 17:00:15 | bauzas | sahid: https://review.opendev.org/c/openstack/nova/+/858383/25 | |
| 17:01:07 | bauzas | sahid: tl,dr: you return an exception if a caller asks for a target_state parameter that the compute doesn't know | |
| 17:02:02 | bauzas | sahid: thinking out loud, I think this would be better to just *not* provide the target_state parameter if the compute is old | |
| 17:02:28 | gibi | the vmdk cv victoria patch https://review.opendev.org/c/openstack/nova/+/871699/ will be get kicked out of the gate as the commit message has a hash but that hash is not laneded yet https://zuul.opendev.org/t/openstack/build/0e2475a0312d4cdaa0774a0fc20c42ce/log/job-output.txt#1552 it seems the [stable-only] tag only disables the hash check if there is no hash in the commit message | |
| 17:02:32 | bauzas | this shouldn't be arriving, since you verify that all computes are upgraded, but I'd prefer us to make it clear | |
| 17:02:43 | bauzas | gibi: ack | |
| 17:04:18 | gibi | I will go and remove the hash from the commit message to keep landing the fixes in parallel | |
| 17:06:45 | opendevreview | Balazs Gibizer proposed openstack/nova stable/wallaby: [stable-only][cve] Check VMDK create-type against an allowed list https://review.opendev.org/c/openstack/nova/+/871557 | |
| 17:07:01 | opendevreview | Balazs Gibizer proposed openstack/nova stable/victoria: [stable-only][cve] Check VMDK create-type against an allowed list https://review.opendev.org/c/openstack/nova/+/871699 | |
| 17:07:19 | opendevreview | Balazs Gibizer proposed openstack/nova stable/ussuri: [stable-only][cve] Check VMDK create-type against an allowed list https://review.opendev.org/c/openstack/nova/+/871702 | |
| 17:08:45 | bauzas | gibi: sean-k-mooney: I'm actually surprised to see some RPC pattern returning an exception if a compute is too old, instead of just remove the parameter from the call we do | |
| 17:09:20 | bauzas | if we really want to have RPC backwards compat, the RPC client needs to adapt to what the manager supports | |
| 17:12:51 | sean-k-mooney | if you request something at the api that requries a new rpc version we shoudl not back levle | |
| 17:12:56 | sean-k-mooney | that should be an api error | |
| 17:13:23 | sean-k-mooney | we normally use a compute service bump to allow use to detect this in the api | |
| 17:13:28 | sean-k-mooney | before getting to the rpc code | |
| 17:13:32 | bauzas | and that's what sahid does | |
| 17:13:55 | bauzas | but I don't really like us returning exceptions we don't really manage upside | |
| 17:14:03 | dansmith | making a call, getting an exception and making it again with different stuff is wasteful *and* wrong, | |
| 17:14:17 | sean-k-mooney | right and we are not doing that | |
| 17:14:25 | dansmith | okay | |
| 17:14:33 | sean-k-mooney | where it can be backleveled we do prepare/version check | |
| 17:14:50 | bauzas | dansmith: tl;dr sahid is adding a service check that verifies all computes are upgraded before adding a parameter | |
| 17:14:58 | sean-k-mooney | but if you use the new parmater then its not valide to remove it | |
| 17:15:00 | dansmith | ack | |
| 17:15:44 | bauzas | by default the param is set to None on the API method | |
| 17:15:45 | sean-k-mooney | bauzas: yep i would break our api microversioning to backlevel in this case | |
| 17:15:55 | bauzas | then it calls the conductor and then the compute | |
| 17:16:01 | sean-k-mooney | bauzas: only for the new microversion no? | |
| 17:16:14 | sean-k-mooney | we backlevel if its actully None | |
| 17:16:22 | sean-k-mooney | i need to pull up the patch again | |
| 17:16:32 | tobias-urdin | is the uuid of a mdev just a random uuidutils.uuid() or is it tracked somewhere in placement? | |
| 17:16:45 | sean-k-mooney | tobias-urdin: totally random | |
| 17:17:01 | tobias-urdin | sean-k-mooney: ack ty | |
| 17:17:33 | bauzas | sean-k-mooney: technically with sahid's proposal, we backlevel by the if condition | |
| 17:18:31 | bauzas | but ok, I see my mistake, I'll clarify my second comment | |
| 17:18:46 | bauzas | either way, my first comment remains valid, we lack a negative test | |
| 17:18:49 | sean-k-mooney | we do it by if not client.can_send_version(version): | |
| 17:19:49 | sean-k-mooney | so if we cant send the version and target_state is not None we raise | |
| 17:19:54 | sean-k-mooney | https://review.opendev.org/c/openstack/nova/+/858383/25/nova/compute/rpcapi.py#1106 | |
| 17:20:01 | sean-k-mooney | this is what you are askign about yes | |
| 17:20:48 | bauzas | yep, you convinced me on the RPC contracty | |
| 17:21:00 | bauzas | my second comment is not valid | |
| 17:21:29 | bauzas | that being said, I'm not super happy with those generic exceptions being raised without being properly captured at the API side | |
| 17:21:37 | bauzas | but that's not worth a -1 | |
| 17:21:50 | bauzas | my -1 is just about missing unittests on the RPC checks | |
| 17:21:51 | sean-k-mooney | im ok with useing pop to do the removal but we woudl need to check the result to determin if we raise or reducet the verison | |
| 17:22:15 | bauzas | sean-k-mooney: we're on agreement | |
| 17:22:29 | bauzas | that's what I mean by the RPC contract | |
| 17:22:48 | bauzas | if this parameter is set to something, we can't backlevel | |
| 17:22:54 | sean-k-mooney | oh its missing a negitive test | |
| 17:22:57 | bauzas | you convinced me | |
| 17:23:04 | bauzas | sean-k-mooney: yup, that's why I -1 | |