Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2023-04-11
12:50:58 gibi https://review.opendev.org/q/topic:limited-lower-constraints-jobs
12:51:52 bauzas gibi: ack, will look
12:52:29 sean-k-mooney gibi: is there a reason you are stating which packages to restrict
12:52:37 sean-k-mooney gibi: instead of everything in requirements.txt
12:53:15 gibi sean-k-mooney: yes, I wanted to limit the scope of the lower constraint jobs as we know that pulling everythin back can be tricky
12:53:40 gibi we can try to pull back all our direct depts
12:53:42 sean-k-mooney ok so now its only testing other nova deliverables
12:54:03 sean-k-mooney honsetly those are the ones i was the least worried about
12:54:25 gibi funnily in Zed we delivered nova with a wrong os-traits dept
12:54:38 gibi and that triggered my thinking about this job
12:54:52 gibi *wrong minimum os-traits dept
12:55:12 sean-k-mooney for which feature
12:55:20 gibi let me dig it out
12:55:32 sean-k-mooney there was one that i expicty didint want to bump it for shice it was optional
12:55:56 sean-k-mooney although that was os-brick i think not os-traits
12:56:22 gibi sean-k-mooney: this was the fix https://review.opendev.org/c/openstack/nova/+/858280
12:56:59 gibi we released RC2 due to this ^^
12:57:23 gibi and we only detected that in time because zigo was eager to package the RC1
12:57:39 sean-k-mooney did it cause failusre when pci in placment was disabeld
12:57:49 sean-k-mooney becaus if it did not we should not have released an rc2 for that
12:58:30 sean-k-mooney in zed we had already stop including lower constraits in the PTI
12:59:39 zigo gibi: And it's out of luck that I found it (ie: because using sbuild + git-buildpackage with the experimental as overlay only, which takes packages from unstable unless there's a strong hint that the experimental version is needed...)
13:00:00 sean-k-mooney i guess given it broke unit tests it could break packagers workflows
13:00:08 sean-k-mooney but high seams a strech to me
13:01:38 sean-k-mooney gibi: https://github.com/openstack/governance/blob/584e06b0c186d4355d1d51f2d6df96e822253bef/resolutions/20220414-drop-lower-constraints.rst happend in zed
13:02:23 gibi sean-k-mooney: and we dropped lower-constraints as a blanket lower constraints job is very hard to do right do to lack of tooling support
13:02:30 gibi hence my take on a limited scope
13:03:38 sean-k-mooney well to me its value was in direct depencies on things like oslo or any other package we directly use
13:03:58 gibi sean-k-mooney: as I said we can try to extend the current proposal to more direct deps
13:04:22 gibi I'm not sure it will be correct if there as cross dependencies between our direct depts
13:04:29 gibi but we can try
13:04:29 sean-k-mooney well if its not everythign in requiremetns.txt i find it hard to trust that it works
13:04:46 gibi I proved it works for os-traits. see the DNM patch on the top
13:05:05 sean-k-mooney yep but what if we start usign a feature in oslo-utils
13:05:18 sean-k-mooney like the imageinfo json output for the vmdk cve
13:05:57 sean-k-mooney to me the real benifit of lower constraits was to ensure we catch issues when backporting
13:06:00 gibi the possible fault we want to catch with the lower constraint is generic for all our direct (or even indirect) depts. I'm not disagreeing there
13:07:00 sean-k-mooney it also has benifit on master to ensure we are intentional when increaseing version and call it out in release notes
13:07:22 sean-k-mooney well if required like swapping to sqlacmey 2.0
13:11:39 gibi so I would do a blanket lower constraint job if I could. But I know that it is impossible to do that right. So I started from a small set of depts. If you have suggestions what else to add to the set like oslo-utils or sqlachemy. Then we can try. I think blindly add all the direct dept can lead to problems
13:17:13 bauzas gibi: sean-k-mooney: I think we somehow agreed on last weekly meeting that the job should only check the existing packages by pinning their versions
13:17:58 gibi bauzas: "existing packages" <- this is not well defined.
13:19:21 bauzas gibi: I mean, all of the packages from reqs.txt
13:19:43 gibi see above. I think that is bigger than what we can chew. but we can try
13:20:30 gibi as soon as we have cross dependencies between our direct debt the same problem will appert than what we had with the blanket lower constraint job
13:23:21 sean-k-mooney i would almost prefer to take the opisite approch
13:23:27 sean-k-mooney have an exclude list
13:23:39 sean-k-mooney which we filter out
13:23:53 sean-k-mooney so include all the package by default and then exclude ones we have issues with
13:24:03 sean-k-mooney for example cryptography
13:26:07 gibi sean-k-mooney: I have to think about how can we effectively decide what to filter out
13:27:27 sean-k-mooney for me it would be anything that is normally isntalled form the package manager
13:27:41 sean-k-mooney although with tox we might be abel to avoid that
13:28:21 sean-k-mooney i know in devstack cryptography has been problematic in the past due to the fact its normally already installed in teh vm and we cant remove/upgrade it
13:29:04 gibi sean-k-mooney: the problem is that we have no way to force lower constraints transitively with the pip tooling
13:29:29 gibi so as soon as direct dept has cross dependencies we are lost do the the tooling limits
13:32:21 sean-k-mooney pip takes the first constrait for a given package as the constiat too use
13:32:32 sean-k-mooney at least that used to be the behavior
13:32:40 sean-k-mooney i have not looked in a while.
13:33:18 sean-k-mooney but if you include the modifed requirements fiel with -c before uc i think that will make it work for transitive deps
13:33:46 sean-k-mooney of couse if one of our transitive dep increase its min version then we will have to bump ours
13:34:17 sean-k-mooney sorry direct deps
13:34:36 sean-k-mooney it basically gets problematic becasue pip does not support this usecase
13:38:06 gibi yeah
13:38:24 gibi so I don't want to go big and hit the same wall we had before with the lower constraint job
13:39:02 sean-k-mooney i have +2d the relevent patches but i still dont really like this change
13:39:40 gibi this is the smalles, lowest risk change, that can catch the specific issue we had during Zed
13:40:07 gibi I agree that it does not solve every aspect of the generic issue, but I think it is already useful
13:40:35 gibi I cannot promise I can solve the generic issue either
13:49:53 opendevreview Justas Poderys proposed openstack/nova-specs master: Add support for Napatech LinkVirt SmartNICs https://review.opendev.org/c/openstack/nova-specs/+/859290
14:53:29 bauzas elodilles: not sure you added some items in the wikipage yet but auniyal wanted to add a topic
14:55:16 bauzas elodilles: he added https://wiki.openstack.org/w/index.php?title=Meetings/Nova&diff=183013&oldid=182963
14:55:32 bauzas elodilles: so I moved those items into https://wiki.openstack.org/w/index.php?title=Meetings/Nova&diff=next&oldid=183013
15:13:41 elodilles bauzas: not yet, but i'm about to see if anything needs to be added there, will check it now
15:18:03 bauzas elodilles: I'm done with this so you can do it anytime
15:19:16 elodilles ack, just updated it
15:20:15 elodilles i'll update the Xena etherpad soon, i'm just generating its content
15:21:02 elodilles so that we can discuss xena-em transition based on that on the meeting
15:42:30 bauzas reminder: nova meeting in 18 mins
16:00:21 opendevmeet The meeting name has been set to 'nova'
16:00:21 opendevmeet Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:00:21 opendevmeet Meeting started Tue Apr 11 16:00:21 2023 UTC and is due to finish in 60 minutes. The chair is bauzas. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:21 bauzas #startmeeting nova
16:00:31 bauzas heydo heya
16:00:43 bauzas #link https://wiki.openstack.org/wiki/Meetings/Nova#Agenda_for_next_meeting
16:00:49 elodilles o/
16:02:53 gibi o/
16:02:56 bauzas hmmmm,
16:03:07 bauzas okay, let's start
16:03:11 gibi what a crowd
16:03:19 Uggla o/
16:03:22 bauzas thanks Hungary
16:03:23 auniyal o/
16:03:29 bauzas hah, people arrive :)
16:03:39 bauzas #topic Bugs (stuck/critical)
16:03:44 bauzas #info No Critical bug
16:03:48 bauzas #link https://bugs.launchpad.net/nova/+bugs?search=Search&field.status=New 19 new untriaged bugs (-3 since the last meeting)
16:04:09 bauzas unfortunately a new bug just arrived one minute before the meeting started :)
16:04:37 bauzas I haven't created an etherpad since most of the bugs I closed were not really needed to be look
16:04:42 bauzas looked*

Earlier   Later