Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2019-09-06
16:05:14 mriedem efried: +2 on that deps update for the sighup fix https://review.opendev.org/#/c/679974/ - mnaser / dansmith fyi on that
16:05:21 efried thanks mriedem
16:05:23 mriedem definitely rc blocker
16:06:28 dansmith efried: I was thinking you had tested something related to sighup and decided that it was *not* properly restarting us after the fix
16:06:35 dansmith efried: did I misunderstand that or did something else change?
16:07:00 efried dansmith: that was at the point where we only had the oslo.service fix. I dug further and fixed the oslo.privsep issue. With both fixes together, shit works.
16:07:11 dansmith gotcha okay cool
16:07:13 efried links from that patch --^
16:07:54 dansmith efried: yeah okay I didn't get that from the reno when I first read it,
16:08:03 dansmith but I didn't study both bugs and just wanted to make sure
16:08:10 efried ++
16:08:16 melwitt efried, stephenfinI think you're right :/ I didn't notice that. commenting
16:08:37 openstackgerrit sean mooney proposed openstack/nova master: [DNM] testing with old lable https://review.opendev.org/680739
16:08:37 openstackgerrit sean mooney proposed openstack/nova master: [DNM] testing with new lable. https://review.opendev.org/680738
16:08:44 efried melwitt: looking again, I think the one I commented on is okay, but the one a couple lines down should be access_url, not access_url_base.
16:08:46 sean-k-mooney ignore those by the way ^
16:09:04 sean-k-mooney im trying to debug stuff.
16:09:28 openstackgerrit Matt Riedemann proposed openstack/nova master: Find instance in another cell during floating IP re-association https://review.opendev.org/656594
16:09:37 melwitt efried: I think they should both be access_url. the second one works because it's just checking the scheme, which would be the same on both access_url_base and access_url
16:09:56 efried melwitt: access_url is a @property tho, not a field.
16:10:10 melwitt yeah, but I think you need to get the access_url first and if None, raise
16:11:02 melwitt that check is supposed to be checking whether there is an access_url available and the property can return None
16:11:09 stephenfin efried: Yeah, what melwitt is saying. I've replied on the patch
16:11:24 melwitt but I dunno, maybe it doesn't really matter because all that method is doing is validating the scheme
16:11:49 melwitt it's never going to look at anything beyond the "https" or whatever the scheme is
16:12:27 efried ah, that makes sense
16:12:44 efried arguably a bit unrelated, but sokay I guess.
16:13:32 melwitt yeah, I'd agree that using the actual access_url here would be unrelated to what this method is doing
16:14:08 efried melwitt: If you're still good, I'll +A
16:14:12 melwitt (realized that as I typed the next two sentences)
16:15:14 melwitt efried: yeah I think it's fine after thinking through
16:16:10 efried ack, thanks for bearing with me melwitt & stephenfin
16:16:13 melwitt probably did
16:16:20 stephenfin thanks for the review (y)
16:22:59 efried sean-k-mooney: can you please ack this https://review.opendev.org/#/c/680094/ ?
16:23:54 sean-k-mooney oh the os-vif request
16:24:02 sean-k-mooney yes i wanned to double check one thing
16:24:08 sean-k-mooney but i think its fine
16:24:16 efried sean-k-mooney: ...per http://lists.openstack.org/pipermail/openstack-discuss/2019-September/009182.html
16:24:26 efried thank you
16:24:39 sean-k-mooney we did 2 train release already for m1 and m2 so this tecnically is not the first os-vif release of train
16:24:52 sean-k-mooney but i think 1.17.0 still make sense for other reasons
16:25:19 efried sean-k-mooney: right, that was the main thing I was concerned about. It looked like there were two bugfixes since the last release, but one of them seemed like it might be feature-y.
16:25:56 efried I'm sure it's not the end of the world if we go feature bump when we should have gone bugfix bump... is it smcginnis?
16:25:59 sean-k-mooney one of them was a securiy bug which has been backported
16:26:33 sean-k-mooney honestly version number are cheap so i would rather bump more then we neeeded too then not so 1.17 looks good to me
16:27:41 sean-k-mooney efried: no if we bump major it would be an issue but bumping feature should not be
16:27:56 sean-k-mooney stien is on the 1.15.x seriese anyway
16:28:30 efried I guess the only downside could be confusion of having .16 and .17 both in train. But meh.
16:28:39 smcginnis efried: So that one wasn't a new feature? I didn't go much further than the commit message.
16:29:15 efried smcginnis: the commit message sounded feature-y for sure, but there was a bug (not a bp) associated. This is what I wanted sean-k-mooney to confirm for us.
16:29:31 efried I could look at the code, but...
16:29:38 efried jangutter: you have an opinion here?
16:30:29 efried specifically: is this https://review.opendev.org/#/c/658786/ a "feature" or a "bugfix", for purposes of release numbering?
16:31:16 jangutter efried: it leans more to "feature" than bugfix, especially since the original idea was "feature".
16:31:31 efried jangutter: Okay, good enough for me, thanks.
16:31:37 efried sean-k-mooney, smcginnis: ^ cool?
16:31:48 smcginnis Works for me.
16:32:11 jangutter efried: os-vif uses bugs to track features (like neutron)
16:32:39 efried smcginnis: acked, thanks for your patience.
16:32:53 efried jangutter: okay, gtk, then I'm going to have this question every release :P
16:33:53 jangutter efried, sean-k-mooney: is there a way to distinguish between [RFE] bugs and your run of the mill annoyances?
16:33:58 smcginnis Thanks for following up on that.
16:34:39 efried jangutter: With careful use of tagging, I'm sure there is. You may have even had that in place in this case - I admit I didn't look.
16:34:45 smcginnis It's been pretty common in my experience that new "features" also fix bugs, so I've never really used BPs vs bugs as an indicator.
16:35:03 efried But that would also require an understanding of whatever tag scheme by whoever's doing the looking
16:35:11 efried and yeah, more broadly, what smcginnis says.
16:36:13 efried it's on a team by team basis, so really we need team members to make this call one way or another (either by proposing the release patches themselves and noting this reasoning in the commit message, or by being pestered like this when releases are proposed "automatically" or otherwise externally to the team)
16:37:15 sean-k-mooney jangutter: we use RFE bugs in os-vif to track feature
16:37:26 sean-k-mooney jangutter: because we dont want to have to use specs or blueprints
16:37:40 sean-k-mooney so RFE bugs cant be backported as they are features
16:37:43 sean-k-mooney other then that no
16:40:04 efried aspiers: I'm looking at https://review.opendev.org/#/c/680065/8/nova/tests/unit/scheduler/test_utils.py -- I was digging for this answer myself, but am being pulled in multiple directions -- can you confirm that there are other tests covering the behavior when hw_machine_type is absent (which is to allow it through) ?
16:40:19 efried I think tests above those are doing it, yah?
16:40:25 aspiers efried: checking
16:41:03 efried iow I wanted to say "hey, instead of doing this to all three, just do it to one or two, so it proves the other thing" -- but don't want to block on that if other tests are already covering that.
16:41:06 aspiers efried: testing hw_machine_type being absent was missing until very recently, I just added it like 2 days ago
16:41:15 aspiers let me find where
16:41:23 aspiers it's mentioned in a commit message somewhere
16:41:46 aspiers oh yeah it's this review
16:42:07 aspiers let me find the lines
16:42:35 aspiers efried: https://review.opendev.org/#/c/680065/8/nova/tests/unit/virt/test_hardware.py@3837
16:43:09 efried that's asserting failure due to no uefi...
16:43:14 aspiers doh
16:43:19 aspiers brain fried
16:43:23 efried where's a case that succeeds with no mach type
16:43:31 aspiers ah OK
16:43:45 aspiers checking
16:43:55 efried perhaps L3926 could have a None in it too?
16:44:09 aspiers ah that has to be tested in the driver
16:44:20 aspiers since with no machine type in the image, it relies on pulling it from utils
16:44:28 aspiers which reads nova.conf on compute node
16:44:36 aspiers I think it's maybe in a later patch
16:44:39 aspiers checking ...
16:46:58 aspiers efried: https://review.opendev.org/#/c/644565/52/nova/tests/unit/virt/libvirt/test_driver.py@2700
16:47:21 aspiers although maybe there is a test missing for when CONF.libvirt.hw_machine_type is set to x86_64=q35
16:47:58 efried aspiers: that one is *also* a failure scenario
16:48:05 aspiers yeah that's what I'm saying
16:48:23 aspiers CONF.libvirt.hw_machine_type set to x86_64=q35 would be the non-failure scenario
16:48:27 aspiers I can add that

Earlier   Later