| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2023-01-10 | |||
| 10:39:28 | bauzas | and then I can +2 it | |
| 10:47:13 | opendevreview | Balazs Gibizer proposed openstack/nova stable/train: Reproduce bug 1981813 in func env https://review.opendev.org/c/openstack/nova/+/869673 | |
| 10:47:14 | 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 | |
| 10:52:51 | Uggla | bauzas, looking at it | |
| 11:32:32 | sean-k-mooney | bauzas: traits never enable features so we need an image proprly or flavor extra spec | |
| 11:32:40 | sean-k-mooney | https://review.opendev.org/c/openstack/nova-specs/+/861881/7/specs/2023.1/approved/virtiofs_scaphandre.rst#127 | |
| 11:49:52 | opendevreview | Pavlo Shchelokovskyy proposed openstack/nova master: Process all nodes when host aggregate changes https://review.opendev.org/c/openstack/nova/+/869687 | |
| 12:03:37 | sean-k-mooney | gibi: the pypy behavior was exactly what i was concerend about | |
| 12:06:29 | gibi | I haven't checked jython and ironpython but I guess they also inherit the gc from the runtime env for java and .net so they are not ref count based | |
| 12:06:49 | sean-k-mooney | yep that is proably | |
| 12:07:09 | sean-k-mooney | althogh as i said im my reply to your comment | |
| 12:07:19 | sean-k-mooney | outside the clients we officaly only support cpython | |
| 12:07:46 | sean-k-mooney | but im not against enabling other interperts for core nova we just dont supprot that officaly or test it | |
| 12:09:12 | sean-k-mooney | i mostly cargo cult the "with open ..." pattern because that is waht i learned to do year ago and the implict understandign that when used as a context manager the file like object retruned by open would close the underlying stream/socket/filedescriptor | |
| 12:10:11 | sean-k-mooney | with statements dont actully form scopes so the closign is delayed until the exit of the function in either case | |
| 12:10:49 | sean-k-mooney | well without the with i gues it when the gc runs ranter then the exit of the fuction scope | |
| 12:12:24 | kashyap | gibi: sean-k-mooney: The functional test failures here seem unrelated to my patch: https://review.opendev.org/c/openstack/nova/+/869587 | |
| 12:12:40 | kashyap | Am I reading it right? Example failure from py38 job: https://zuul.opendev.org/t/openstack/build/21785a8539e34c65aad0038a6cae0cde | |
| 12:13:07 | kashyap | "failed with could not find python interpreter matching any of the specs functional-py38" | |
| 12:13:08 | sean-k-mooney | this is the gate blocker we discussed at the start of the tech call yesterday | |
| 12:13:18 | sean-k-mooney | so not related to your patch | |
| 12:13:25 | kashyap | Ah, I missed it. | |
| 12:13:27 | sean-k-mooney | its related to tox 4 | |
| 12:24:25 | kashyap | sean-k-mooney: Thanks! (My keyboard got locked accidentally, and had to reboot) | |
| 12:25:04 | kashyap | I guess it's related to this patch from gibi -- https://review.opendev.org/c/openstack/placement/+/868418 ("Make tox.ini tox 4.0.0 compatible") | |
| 12:26:44 | opendevreview | Rajesh Tailor proposed openstack/nova master: Fix huge-page doc https://review.opendev.org/c/openstack/nova/+/869689 | |
| 12:28:27 | gibi | kashyap: yes | |
| 12:28:50 | sean-k-mooney | kashyap: well tehre is a nova version of that too | |
| 12:28:57 | sean-k-mooney | i think from gmann ? | |
| 12:29:28 | gibi | https://review.opendev.org/c/openstack/nova/+/869545 this is the nova one | |
| 12:29:48 | gibi | sean-k-mooney: I think the with statement itself make sure that after the with block the fd is closed. it might not free up any memory but it closed the file descriptor | |
| 12:30:14 | sean-k-mooney | ah right the stable branch ones are from gmann to pin and the master one is form you | |
| 12:31:00 | sean-k-mooney | stephenfin: just hit the nova patch so that should merge soon | |
| 12:31:11 | gibi | cool | |
| 12:31:31 | kashyap | Thanks for the link! | |
| 12:31:46 | sean-k-mooney | gibi: out of interest what does the -I do in python -I -m ... | |
| 12:32:16 | stephenfin | sean-k-mooney: -I : isolate Python from the user's environment (implies -E and -s) | |
| 12:32:26 | stephenfin | -E : ignore PYTHON* environment variables (such as PYTHONPATH) | |
| 12:32:31 | gibi | yepp | |
| 12:32:32 | sean-k-mooney | ack just checked the help text | |
| 12:32:35 | gibi | I had to look that up too | |
| 12:32:39 | gibi | tox uses -I | |
| 12:33:24 | sean-k-mooney | this ensure that it uses only pacakes in teh venv | |
| 12:36:29 | sean-k-mooney | seams to be workign on my local system so fine by me | |
| 12:39:15 | sean-k-mooney | well mostly given i dont have all the dep to powervm on my laptop | |
| 12:39:48 | sean-k-mooney | if im not mistaken once that merges we can resume merging the pci serise yes | |
| 12:39:55 | sean-k-mooney | that is the last patch blocking that? | |
| 12:40:35 | gibi | sean-k-mooney: that is my hope | |
| 12:40:47 | gibi | I will recheck the bottom of the PCI series when the tox patch lands | |
| 12:41:40 | sean-k-mooney | ack | |
| 12:54:59 | kashyap | Thanks for unblocking the Gate, gibi and gmann! | |
| 12:56:19 | gibi | no worries, I have my incentives to do it :) I want to land the PCI series | |
| 12:58:33 | kashyap | :) | |
| 13:03:33 | stephenfin | gibi: btw, filed https://github.com/pypa/pip/issues/11718 | |
| 13:04:33 | stephenfin | Hopefully we can get ahead of it. This will be a big issue once pip switched to the PEP-517/PEP-660 flow by default | |
| 13:17:49 | sean-k-mooney | stephenfin: nice bug report | |
| 13:18:00 | sean-k-mooney | also usign os-vif sicne its small is a good choice | |
| 13:24:57 | gibi | stephenfin: thanks that is a really good summar of the issue | |
| 13:36:50 | stephenfin | gibi: Just replied there. Turns out this happens without '--use-pip517' or '-e': tox 3 never used 'install_command' when installing the package under test \o/ | |
| 13:39:03 | stephenfin | so perhaps the "correct" thing to do is to not override 'install_command' at all | |
| 13:44:35 | bauzas | sean-k-mooney: about Uggla's spec, OK, you don't want to use the trait in the flavor extraspec | |
| 13:45:14 | sean-k-mooney | well we can have a required trait but htat required trait cannot change the guest xml | |
| 13:45:23 | sean-k-mooney | traits never enable a feature | |
| 13:45:31 | sean-k-mooney | they only find a host where the feature can be enebaled | |
| 13:45:49 | bauzas | sean-k-mooney: I'm OK with your specific image property : "hw_power_metrics" | |
| 13:46:00 | bauzas | without virtiofs | |
| 13:46:05 | sean-k-mooney | sure | |
| 13:47:47 | bauzas | I wrote a comment | |
| 13:47:58 | bauzas | just b/c virtiofs is related to a specific virt driver | |
| 13:48:05 | bauzas | but meh | |
| 13:48:21 | bauzas | Uggla: would you be able to provide a new revision today then ? | |
| 13:48:35 | bauzas | I guess sean-k-mooney and me agree | |
| 13:49:10 | sean-k-mooney | yes one thing im thinking about is in the future we might need a second property | |
| 13:49:13 | bauzas | Uggla: about the instance UUID, don't be afraid | |
| 13:49:31 | sean-k-mooney | hw_power_metrics_interface=virtiofs|qemu-channel | |
| 13:49:40 | bauzas | Uggla: Scaphandre can know about the instance name | |
| 13:49:50 | sean-k-mooney | ie if qemu gains the ablity to supprot scaphadre without virtio fs | |
| 13:49:55 | bauzas | by looking at the metadata API | |
| 13:49:56 | sean-k-mooney | but i think we can evolved to that | |
| 13:50:19 | bauzas | sean-k-mooney: good point | |
| 13:51:38 | bauzas | https://github.com/openstack/nova/blob/master/nova/api/metadata/base.py#L165 | |
| 13:51:46 | bauzas | Uggla: ^ | |
| 13:52:00 | Uggla | bauzas, yes I can probably send a new rev today. | |
| 13:53:11 | gibi | stephenfin: I'm confused :) | |
| 13:53:33 | bauzas | Uggla: https://docs.openstack.org/nova/latest/user/metadata.html#ec2-compatible-metadata | |
| 13:53:38 | gibi | stephenfin: so in tox3 we never constrained the editable package install? | |
| 13:53:41 | stephenfin | gibi: https://github.com/pypa/pip/issues/11718#issuecomment-1377291170 | |
| 13:53:45 | gibi | stephenfin: evne with skipdist = True? | |
| 13:54:12 | stephenfin | yup | |
| 13:55:22 | gibi | so either we silently had unconstrained deps in the env or we were lucky and those deps were never updated during the package install | |
| 13:56:52 | stephenfin | I suspect the latter | |
| 13:57:12 | Uggla | bauzas, if I understand well you suggest to keep the uuid as dir name, and change scaphandre to get the instance uuid in the metadata ? | |
| 13:57:48 | bauzas | Uggla: problem is, instance hostname is mutable | |
| 13:57:57 | gibi | stephenfin: so we keep os-vif as is | |
| 13:58:15 | gibi | do you want me to change the nova and placement patch and remove the install_command from there?> | |
| 13:58:25 | gibi | it is not merged yet | |
| 13:58:41 | sean-k-mooney | maybe do that as a follow up patch | |
| 13:59:02 | sean-k-mooney | so we can leave them merge and unblock the gates but then decied the best path forward | |
| 13:59:03 | stephenfin | It might be worth it. The install_command approach will only work for service projects now since those are the only ones not listed in upper-constraints | |
| 13:59:10 | stephenfin | but a follow-up is fine | |