Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2025-04-02
14:23:59 Uggla dansmith, gibi, I have a pb with the eventlet. Eventlet "team" is doing a session Tue 14 --> 16h UTC. Will you want to go to this meeting. Do you want to address the "nova" eventlet point within this meeting or keeping it one the nova side ?
14:24:47 gibi Uggla: I would like to go to the global eventlet session to gather input
14:25:36 gibi Uggla: if possible it woudl be nice to have the nova specific eventlet discussion after the global session
14:28:53 Uggla Ok fine by me, if possible I'll try to not set something at the same time so people could go to the eventlet session. But we have a lot of topics, so I'm unsure we can do it yet. So at least you will be our nova representative guy in that session.
14:29:48 gibi Uggla: feel free to schedule things for nova in that timeslot I can add my views as comment to the etherpad for those topics and the team can discuss without me
14:30:24 dansmith Uggla: also, you don't need to ruin anything just to work around my schedule
14:30:51 dansmith things I really need to be part of would be nice to be earlier in the week, but if it's not possible, it's not worth breaking everything
14:30:51 Uggla gibi, sure, but I'd like to assist if possible. TBH that sounds difficult.
14:31:07 dansmith not sure if that's the reason for the scheduling problems or not, but.. it shouldn't be
14:32:04 gibi Uggla: sure appreciate that you try, just not worry too much :)
14:33:21 Uggla I try to adapt to all constraints if possible. If not that's life.
14:34:19 Uggla dansmith, fyi cinder confirms the timeslot for glance/cinder/nova session.
14:34:43 dansmith Uggla: wait, nova/cinder/glance or nova/glance?
14:35:01 Uggla nova/cinder/glance
14:35:57 dansmith okay is that when we're discussing the locations stuff?
14:35:59 Uggla dansmith, Cyril requested cinder too.
14:36:17 dansmith ah okay I'm not sure that's the best plan for that but... okay
14:36:39 Uggla dansmith, you know better than me. :)
14:36:43 dansmith he needs to talk to cinder about the same topic as nova, but we don't all need to take the global lock, AFAIK
14:37:07 dansmith don't change it now, just saying
14:38:29 Uggla dansmith, yep let's keep it like this. As booking is more difficult than releasing people.
14:39:11 dansmith yup
14:49:19 elodilles gibi is right, but we couldn't help releasing / announcing a bit earlier: https://lists.openstack.org/archives/list/openstack-announce@lists.openstack.org/thread/ACVRVHIKT5K2TKZIPVT4P2OGAK3LZVVL/
14:49:33 elodilles thanks everyone :) \o/
14:49:49 elodilles 'OpenStack 2025.1 Epoxy is officially released!'
14:50:10 gibi \o/
14:50:30 Uggla elodilles, thanks \o/
14:51:27 dansmith Uggla: https://review.opendev.org/c/openstack/releases/+/946162
14:51:33 dansmith just so you don't have to do it ^
14:52:15 Uggla dansmith, thx I'll review ASAP.
14:54:06 dansmith aside from the mistake that was splitting out placement, having os-traits a library that requires a commit there, a release, a u-c bump, and requirements change to get a string constant into nova is about the most ridiculous process we could have designed :)
15:09:43 mnaser Has there been any changes in how network_info is generated/cached/etc? I have an 2024.1 environment and it seems that the Ironic driver is getting the following as network_info from Nova inside spawn: https://paste.openstack.org/show/b6wcQ0hWl9JvdpwHEOks/ -- this environment uses routed provider networks
15:10:05 mnaser As a result of this, it naturally generates a metadata of `{'links': [], 'networks': [], 'services': []}` only.
15:10:41 mnaser But after the fact if I check the instance_info_cache, the value is populated correctly, I got that value by adding a bunch of logs, so it tells me that i'm somehow getting an "unresolved" network_info
15:13:04 Uggla gibi, is "Nova needs to adjust its DB sche ma to Glance's regarding ImageProperties." something we could discuss with glance too ?
15:14:01 gibi as far as I understand from the past glance does not really want to change anything so we can bring it up with them if we have time but I think we can handled that as a nova only bug to fix
15:14:14 dansmith agree
15:46:59 Uggla melwitt, is 1h enough to covert your PTG topics (power mgmt & Tempest testing requirements for API changes
16:19:59 melwitt Uggla: I hope so :) it "should" be
16:20:29 Uggla melwitt, I try to put a convenient slot for you.
16:20:41 melwitt thanks!
16:20:45 Uggla not too early.
16:26:26 Uggla Hi all, please have a look at https://etherpad.opendev.org/p/nova-2025.2-ptg — if you spot anything incorrect, feel free to let me know.
17:21:51 opendevreview Dan Smith proposed openstack/nova master: Support "one-time-use" PCI devices https://review.opendev.org/c/openstack/nova/+/943816
17:21:51 opendevreview Dan Smith proposed openstack/nova master: Add one-time-use devices docs and reno https://review.opendev.org/c/openstack/nova/+/944262
17:21:52 opendevreview Dan Smith proposed openstack/nova master: Functional tests for one-time-use devices https://review.opendev.org/c/openstack/nova/+/946065
17:32:21 dansmith gmann: do you know why this fails like this? https://zuul.opendev.org/t/openstack/build/82b63cd467c342fa89cd0c29f599431d
17:32:33 dansmith for this requirements bump: https://review.opendev.org/946177
17:32:50 gmann checking
17:38:09 gmann dansmith: it should pass, i think it is just mirror issue?
17:38:51 dansmith oh is it? it seemed to be complaining about the conflict of versions and not that it couldn't actually find the version requested
17:39:33 gmann but conflict versions are not valid - 3.4.0 is >=3.3.0
17:40:06 dansmith you mean the complaint it has is not valid right?
17:40:30 gmann yeah, I have seen these in past due to mirror issue
17:40:40 dansmith okay, weird
17:40:59 gmann 3.4.0 is just released right?
17:41:13 dansmith related, why can't we set the u-c for things like os-traits to like <4.0.0? It's just a bunch of constants, there's no harm in running with a newer one of the same major version
17:41:27 dansmith having to bump u-c for this is kinda silly
17:41:33 dansmith gmann: yes, minutes before
17:41:58 dansmith gmann: I just didn't even consider it possibly a mirror issue since it's not failing to download, but I'm sure you're right
17:43:06 gmann dansmith: on constraints, I agree that there is no point to pin it in u-c especially we only maintain this lib and we can have (might have) master code testing before any new release and made sure new version will not break
17:43:18 dansmith yeah
17:43:32 gmann from testing tools, we consider the same approach and do not have constraint for tempest, hacking or so
17:43:59 gmann and if any project want to pin it can be done in requirements.txt for safer side like we do for hacking version
17:45:07 dansmith yeah, makes sense for nova to require the version with the constants it needs, but u-c seems like too much trouble for this
17:45:56 gmann dansmith: here we can propose to exclude it from u-c, there are many neutron lib in that list https://github.com/openstack/requirements/blob/be97f2507812c9047c5a39ee4a35b88ea7f677d7/denylist.txt#L23
17:46:52 dansmith gmann: so add it to the denylist and remove from u-c.txt?
17:47:24 dansmith I can propose that on top of this bump for discussion
17:47:35 gmann dansmith: once you add it in denylist at same time you can remove it from u-c
17:47:44 dansmith yeah that's what I meant
17:47:54 gmann yeah
17:49:17 Uggla dansmith, fyi added a slot for One-time-use devices (30mn). Is it ok for you ?
17:50:04 dansmith Uggla: yeah tbh I'm not sure there's much discussion needed at this point, but sure
17:50:24 dansmith Uggla: maybe schedule it near something else so we can go straight to the other topic if there are no issues to discuss?
17:51:21 Uggla ok
17:52:14 dansmith gmann: https://review.opendev.org/c/openstack/requirements/+/946183
17:53:21 gmann ++
17:55:12 gmann dansmith: this job failure gives clear reason for failure "ERROR: No matching distribution found for os-traits===3.4.0" https://zuul.opendev.org/t/openstack/build/69b690d9b5744edd8a96a53fa2899193
17:55:58 dansmith gmann: ah indeed
18:45:34 Uggla dansmith, https://review.opendev.org/c/openstack/releases/+/946162 merged and 3.4.0 available on pypi.
18:46:06 Uggla dansmith, fyi ^
18:46:11 dansmith Uggla: yep, thanks I already have the u-c patch up and the nova patches bumped (but I was a few minutes above the mirrors)
18:46:23 dansmith s/above/before/
18:49:35 Uggla cool, have a good day. Time for me to afk.
18:50:17 frickler dansmith: I'd rather go with the bot generated patch that has all the metadata if you don't mind https://review.opendev.org/c/openstack/requirements/+/946181 I'll also leave the decision about removing it from u-c to the "real" reqs ppl (cc tonyb)
18:52:08 dansmith frickler: okay, I've just been blocked for a while and was trying to get this moving
18:53:42 frickler dansmith: sure thing, don't hesitate to ping me if this gets stuck again (or other reqs+release stuff)
18:54:02 opendevreview Dan Smith proposed openstack/nova master: Support "one-time-use" PCI devices https://review.opendev.org/c/openstack/nova/+/943816
18:54:02 opendevreview Dan Smith proposed openstack/nova master: Add one-time-use devices docs and reno https://review.opendev.org/c/openstack/nova/+/944262
18:54:03 opendevreview Dan Smith proposed openstack/nova master: Functional tests for one-time-use devices https://review.opendev.org/c/openstack/nova/+/946065
18:54:12 dansmith rebased to depends-on the bot patch ^
18:54:26 dansmith frickler: once it passes tests can you approve it?
18:54:37 dansmith (the bot requirements patch I mean
18:54:59 frickler I'll approve now, might not be around long enough today
18:55:08 dansmith even better thanks :)
18:56:32 dansmith the bot patch looks like it happened at basically the same time mine did, I thought there was (used to be?) a substantial delay there
19:23:37 frickler there may be delays if the bot job needs to wait for a node, but that should seldom be long. other than that there's only a 10min delay to give the newly uploaded pkg time to appear in the pypi index cached by our mirrors
23:16:39 cardoe mnaser: that paste was an unbound interface which seemed weird.
23:17:23 mnaser cardoe: right, there's routed provider networks here in the equation which only assign a port/subnet until the port is plugged since it doesn't know what subnet it can pick an ip in
23:18:22 mnaser but good point, it seems like it wasn't bound yet in neutron with the baremetal stuff

Earlier   Later