| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2024-01-31 | |||
| 14:05:44 | bauzas | Uggla: are you OK ? | |
| 14:06:09 | bauzas | no, we don't mount the share on the instance dir | |
| 14:06:31 | bauzas | but we have share mappings telling which share to use with what instance | |
| 14:06:33 | sean-k-mooney | it should be either configurabel or under the instace dir | |
| 14:06:45 | bauzas | there is a spec | |
| 14:06:53 | bauzas | and within this spec, this is explained | |
| 14:07:01 | Uggla | bauzas, yep for me it is ok. I have my answers | |
| 14:07:15 | sean-k-mooney | bauzas: im pretty sure the spec did not say its oke to defer the unmoutn to a perodic | |
| 14:08:31 | bauzas | that, indeed, I was just saying about the share path | |
| 14:08:53 | bauzas | anyway, /me going back to work given Uggla got his answers | |
| 14:09:33 | sean-k-mooney | Uggla: so your going to do the unmount when deleting the isntance as part of delete with a fallback to a perodic | |
| 14:10:08 | Uggla | sean-k-mooney, yep but atm I have not periodic. | |
| 14:10:08 | sean-k-mooney | and not have manilla block the delete if unlocking fails? | |
| 14:10:39 | sean-k-mooney | right so without that we would have to make the failure to unlock put the instnace in error | |
| 14:10:49 | sean-k-mooney | gibi: how do you feel about that ^ | |
| 14:12:48 | sean-k-mooney | i do not have the full context so im mostly relying on you adn bauzas to ensure the correct behavior. if we just ignroed the failure to unlock the manilla share with no recovery method i woudl think that deserves a -2 | |
| 14:14:59 | gibi | The cheaper way is to put the instance in ERROR during delete. I suggest to check what we do with cinder volumes during delete if cinder is unavailable | |
| 14:15:08 | sean-k-mooney | we need to be careful of two things: 1.) we are not locking the share indefintly if we delete the isntnce and the manial call fails, that has billing implciations. 2.) we are not exporting the sahre to host it shoudl not be on, the has security implcitations, and possible regelarotry complicne issues. | |
| 14:15:34 | gibi | I agree about these goals ^^ | |
| 14:16:21 | sean-k-mooney | i would have been ok with puting the instnace in error | |
| 14:16:28 | sean-k-mooney | so you can just try again | |
| 14:16:33 | sean-k-mooney | when manilla is fixed | |
| 14:17:07 | sean-k-mooney | and i suspect that is what would happen for a BFV guest with delete on terminate but im not sure | |
| 14:18:08 | sean-k-mooney | it looks like in terminate we do self._cleanup_volumes(context, instance, bdms, | |
| 14:18:10 | sean-k-mooney | raise_exc=False, detach=False) | |
| 14:18:59 | sean-k-mooney | that imples we allwo the delete to proceed | |
| 14:19:18 | bauzas | I also agree with the two bullet points (just lurking) | |
| 14:20:30 | bauzas | tbh, I don't really like us copying the cinder workflow | |
| 14:20:36 | sean-k-mooney | i am about to enter 3 hours of meetings in 3.5 hours | |
| 14:20:45 | bauzas | recent issues led me think there are some corner cases I'd like to avoid :) | |
| 14:20:47 | sean-k-mooney | so ill also be "steppign way shortly" | |
| 14:21:18 | bauzas | but agreed on the | |
| 14:21:23 | sean-k-mooney | right i think gibi and i wourl prefer to be stircter and fail the delte | |
| 14:21:25 | bauzas | DONTs and the DOES | |
| 14:21:50 | sean-k-mooney | but lets loop back to this after Uggla has updated things | |
| 14:21:53 | sean-k-mooney | and we have more time | |
| 14:22:05 | bauzas | I think we said we should never try to fail on an instance delete | |
| 14:22:15 | bauzas | as there are also some capacity and billing implications | |
| 14:22:25 | sean-k-mooney | we can fail on isntance delete for other reasons im pretty sure | |
| 14:22:45 | sean-k-mooney | but you are correct on the capsity and billing implciations if we do too | |
| 14:22:55 | gibi | it is a trade off if we let the VM delete to happen but leak resources in the meantime you still get billing issues :D | |
| 14:23:12 | bauzas | as an operator, I'd be more angry if my instance would continue to exist because of a cheap open file not deleted | |
| 14:23:37 | bauzas | but I could be wrong | |
| 14:23:50 | gibi | bauzas: e.g. we delete your VM happily but fail to unlock the share in manila. So you are billed for a share you cannot use :) | |
| 14:24:03 | bauzas | hence the periodic :) | |
| 14:24:27 | gibi | which can be turned off : | |
| 14:24:28 | gibi | :) | |
| 14:24:34 | bauzas | I'm not saying we should leave the pipe leaking | |
| 14:24:53 | bauzas | what I'm saying is that there are bigger things to do than fixing the pipe | |
| 14:24:54 | sean-k-mooney | bauzas: would that same operator be ok if custoemr data was leaked via presitent memroy, or if vgpus were never made avaiable again or pci devices | |
| 14:24:58 | gibi | at least if my VM is in ERROR and I cannot delete it I will notice and raise a ticket | |
| 14:25:18 | sean-k-mooney | so there is a way to fix this | |
| 14:25:40 | sean-k-mooney | we coudl ask manilla to aloow the share delteion if the nova instnace nolonger exists | |
| 14:25:56 | sean-k-mooney | that coudl be a followup enhacnment on there side | |
| 14:26:24 | sean-k-mooney | so if its locked by nova but nova says the isntnace is gone aloow the user to delelete it even with the lock | |
| 14:26:31 | bauzas | gibi: sean-k-mooney: standing as a user for a sec, | |
| 14:26:42 | bauzas | if I delete my instance and it goes into ERROR | |
| 14:26:53 | bauzas | the second after, I'm gonna try to delete again | |
| 14:27:01 | bauzas | so I could be okay | |
| 14:27:01 | sean-k-mooney | yep which might work | |
| 14:27:04 | gibi | :) | |
| 14:27:21 | bauzas | provided we ensure that the instance deletion path is idempotent | |
| 14:27:38 | sean-k-mooney | yes but we already need to do that | |
| 14:28:05 | sean-k-mooney | i belive one path where instance delete ends with an instance in error | |
| 14:28:10 | sean-k-mooney | is if the connection to libvirt is down | |
| 14:29:23 | sean-k-mooney | or ironic/hyperv/vmware api is not accable | |
| 14:29:45 | sean-k-mooney | assuming the comptue agent is fine it will have to put it in error when it cant talk to the hypervior | |
| 14:30:00 | sean-k-mooney | well that or leave the instnace in active | |
| 14:30:05 | sean-k-mooney | and put the instnace action in errror | |
| 14:30:21 | sean-k-mooney | ok meetings starting | |
| 15:11:52 | bauzas | every time I'm trying to pull a devstack on a RHEL machine, I'm getting funny stories | |
| 15:24:58 | bauzas | Could not find versioned identity endpoints when attempting to authenticate. Please check that your auth_url is correct. Service Unavailable (HTTP 503) | |
| 15:24:58 | bauzas | Failed to discover available identity versions when contacting http://10.73.116.68/identity. Attempting to parse version from URL. | |
| 15:25:15 | bauzas | but when looking at keystone logs in the journal, lgtm | |
| 15:25:31 | bauzas | wtf... | |
| 15:27:04 | bauzas | I did set GLOBAL_VENV to False fwiw | |
| 15:29:56 | sean-k-mooney | well its technially not a supported operating system | |
| 15:30:10 | sean-k-mooney | i can likely take a look in a while if you are still having issues | |
| 15:35:18 | bauzas | SELinux is preventing /usr/sbin/httpd from write access on the sock_file keystone-wsgi-public.socket | |
| 15:35:21 | bauzas | doh | |
| 15:39:19 | sean-k-mooney | devstack disabels it on centos and fedora | |
| 15:39:28 | sean-k-mooney | so it proably is missing a check for rhel | |
| 15:39:43 | sean-k-mooney | that is proably a simple fix in devstack but also you can do it locally | |
| 15:40:21 | bauzas | I think I need to do some kind of install notes for kashyap and others wanting to test vGPUs with devstack | |
| 15:40:40 | bauzas | that's always good to reinstall your devstack | |
| 15:40:59 | bauzas | and I think I got the same problem 6 months ago but I forgot | |
| 15:41:12 | bauzas | so I'm litterally hitting the same bumps without learning | |
| 16:23:01 | melwitt | sean-k-mooney: thanks for the reviews! I will be going through and addressing the comments. and I agree re: functional tests, I have been working on them and other issues | |
| 16:24:10 | sean-k-mooney | ack i was talking to stephen an he made a point that we coudl do the functional tests at the end fo the series | |
| 16:24:36 | sean-k-mooney | i was going to chat to you later and see how you felt about doing it in line or at the end | |
| 16:25:55 | melwitt | I thought about that too and I think ideally I would put them inline (so that one can look at the code alongside the func test showing it's working) | |
| 16:37:04 | opendevreview | Merged openstack/nova master: Revert "[pwmgmt]ignore missin governor when cpu_state used" https://review.opendev.org/c/openstack/nova/+/905671 | |
| 19:16:59 | melwitt | sean-k-mooney: re: your comment about image backends reporting support ... currently I have each patch adding one action (trying to make review not a nightmare) but I'm not flipping the image backend flag to reporting support until the end. in real life it won't be runnable but in func tests it would. are you saying you would prefer to see the flag reporting support from the first patch and guarding all other actions to reject at the | |
| 19:17:01 | melwitt | API instead? and lift each gate with each subsequent patch? | |
| 19:17:47 | sean-k-mooney | i would have done it with the api gating approch and lifted it. | |
| 19:17:54 | sean-k-mooney | but given the code appears to be working | |
| 19:18:02 | sean-k-mooney | and we can test it with functional tests | |
| 19:18:11 | sean-k-mooney | im hesitent to actullly ask you to flip it | |
| 19:18:17 | melwitt | sean-k-mooney: ack. wanted to make sure I understood what you meant | |
| 19:18:47 | sean-k-mooney | so i woudl like to be able to test each "feature" patch either with tempest ro functional ideally | |