| *** tetsuro has joined #openstack-placement | 00:12 | |
| *** takashin has joined #openstack-placement | 00:16 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: Add a placement-manage CLI https://review.openstack.org/600161 | 01:22 |
|---|---|---|
| cdent | smh | 01:23 |
| efried | cdent: Why, oh why are you awake? | 01:28 |
| cdent | because that deep hole I've been in since last saturday with regard to the alembic stack is really really deep | 01:28 |
| cdent | but I may have finally landed | 01:29 |
| efried | cdent: Came across something really nice a couple hours ago. | 01:29 |
| efried | Start zoomed into this: https://review.openstack.org/#/c/615705/1/nova/compute/resource_tracker.py | 01:29 |
| efried | And then zoom out :) | 01:29 |
| cdent | if the landed thing is really landed than about 20 patches can finally merge (alembic migrations, devstack, grenade, integration jobs) | 01:30 |
| * cdent clicks | 01:30 | |
| cdent | oh nice, that is nice | 01:31 |
| efried | If I can pull this off, I'll be expecting a Harrods hamper from jaypipes | 01:31 |
| cdent | depending on what you like, you may prefer Fortnum and Mason | 01:32 |
| cdent | a bit more specialised in the hamper direction | 01:32 |
| efried | I'll settle for a Tesla. | 01:32 |
| efried | anyway, I'm burned up for the night. ttfn | 01:33 |
| cdent | I too am all used up | 01:33 |
| openstackgerrit | Merged openstack/nova-specs master: High Precision Event Timer (HPET) on x86 guests https://review.openstack.org/607989 | 01:34 |
| cdent | jaypipes: if you got cycles to drive through everything above https://review.openstack.org/#/c/614307/ (which you've already +2d) that would fantabulous. efried has already expressed a fair amount of "these things are not my ken" | 01:35 |
| cdent | tetsuro: you too ^ | 01:35 |
| cdent | good night all | 01:35 |
| *** cdent has quit IRC | 01:35 | |
| *** tetsuro has quit IRC | 01:39 | |
| *** tetsuro has joined #openstack-placement | 01:40 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: Add a placement-manage CLI https://review.openstack.org/600161 | 01:44 |
| *** e0ne has joined #openstack-placement | 06:24 | |
| *** e0ne has quit IRC | 06:44 | |
| *** e0ne has joined #openstack-placement | 06:48 | |
| *** e0ne has quit IRC | 07:07 | |
| *** tetsuro has quit IRC | 07:35 | |
| *** tetsuro has joined #openstack-placement | 08:13 | |
| *** helenafm has joined #openstack-placement | 08:15 | |
| *** tssurya has joined #openstack-placement | 08:18 | |
| *** lei-zh has joined #openstack-placement | 08:36 | |
| *** helenafm has quit IRC | 08:42 | |
| *** helenafm has joined #openstack-placement | 08:49 | |
| *** cdent has joined #openstack-placement | 09:22 | |
| *** lei-zh has quit IRC | 09:30 | |
| *** ttsiouts has joined #openstack-placement | 10:07 | |
| *** ttsiouts has quit IRC | 10:18 | |
| *** takashin has left #openstack-placement | 10:39 | |
| *** s10 has joined #openstack-placement | 10:57 | |
| *** e0ne has joined #openstack-placement | 11:11 | |
| openstackgerrit | Brin Zhang proposed openstack/nova-specs master: Support deleting data volume when destroy instance https://review.openstack.org/580336 | 11:49 |
| cdent | gibi: you have some cycles to explore https://review.openstack.org/#/c/614307/ and its children? Is all the database migration related stuff | 12:18 |
| gibi | cdent: I will try. In the last week or so I was pretty lost in the preparation for my summit talk. | 12:19 |
| cdent | gibi: thank. no worries if you can't get to it. I was hoping to have it all merged before summit but if not it isn't a huge deal | 12:20 |
| *** e0ne has quit IRC | 12:29 | |
| *** e0ne has joined #openstack-placement | 12:43 | |
| *** tetsuro has quit IRC | 12:55 | |
| openstackgerrit | Merged openstack/placement master: Move ensure_consumer to a new placement.handlers.util https://review.openstack.org/614307 | 13:09 |
| gibi | cdent: I left some comment in https://review.openstack.org/#/c/612444 I won't block on it I just wanted to note that I feel this decorator does a lot more than what we need | 13:15 |
| cdent | gibi: if you look at some of the discussion on its original in nova it's trying to be forward thinking | 13:16 |
| cdent | and in a patch later on I use the reset functionality | 13:17 |
| cdent | https://review.openstack.org/#/c/614350/13/placement/tests/fixtures.py | 13:17 |
| gibi | cdent: lookinh | 13:18 |
| cdent | that ^ link is because we need to be able to reconfigure an engine facade from sqlite to mysql to postgresql in the same process | 13:19 |
| gibi | cdent: OK, I see the usage of the reset now | 13:20 |
| gibi | cdent: do you have a usage for cleanup as well? | 13:21 |
| cdent | If I was feeling clever I could have used cleanup for the fixtures.reset() method but I decided not to be clever | 13:22 |
| cdent | but something similar to that is the idea | 13:22 |
| cdent | the end goals is that the tripleo people want it to be possible to reconfigure the app (including database settings) on a hup or something similar, with the cleanup function being able to reset the necessary globals and other state | 13:23 |
| gibi | so what is in the fixture.reset() is actually needs to always run together with the configure.reset() call? | 13:23 |
| cdent | I agree that it is probably overkill for the immediate case, but the dicussion on the nova review and the irc discussion associated with it was rather complete | 13:24 |
| cdent | I believe mdbooth may have more info if you're still concerned | 13:24 |
| gibi | cdent: OK, I'm reading the nova review noq | 13:24 |
| gibi | now | 13:24 |
| cdent | the stuff in the fixtures.reset() is not what would need to happen for any configure.reset(), because some of the globals there are related to just the fixture | 13:26 |
| cdent | but things like the pool being disposed and resetting _started could have be in cleanup instead | 13:26 |
| cdent | but I decided not to be clever, at least for now | 13:27 |
| gibi | cdent: I see, thanks | 13:28 |
| cdent | I'm going to go find some lunch. thanks for looking at that stuff. | 13:30 |
| *** cdent has quit IRC | 13:31 | |
| gibi | cdent: enjoy your lunch | 13:42 |
| gibi | cdent: I left a real -1 on the nova patch due to not considering exceptions from func() https://review.openstack.org/#/c/610034 | 13:43 |
| openstackgerrit | sean mooney proposed openstack/nova-specs master: Add spec for sriov live migration https://review.openstack.org/605116 | 13:56 |
| gibi | jaypipes: I'd like to ask a quick review from you on https://review.openstack.org/#/c/570847 about the resource class re-name to NET_BW_EGR_KILOBIT_PER_SEC | 13:59 |
| gibi | efried: from you too, please ^^ | 14:03 |
| *** mriedem has joined #openstack-placement | 14:11 | |
| openstackgerrit | John Garbutt proposed openstack/nova-specs master: Enable rebuild for instances in cell0 https://review.openstack.org/554218 | 14:16 |
| jaypipes | gibi: done. +W. | 14:17 |
| openstackgerrit | John Garbutt proposed openstack/nova-specs master: Add PENDING vm state https://review.openstack.org/554212 | 14:18 |
| openstackgerrit | John Garbutt proposed openstack/nova-specs master: Add PENDING vm state https://review.openstack.org/554212 | 14:19 |
| gibi | jaypipes: cool, thanks | 14:25 |
| openstackgerrit | Merged openstack/nova-specs master: Add support for emulated virtual TPM https://review.openstack.org/571111 | 14:51 |
| openstackgerrit | Merged openstack/nova-specs master: Switch to stestr https://review.openstack.org/581284 | 14:52 |
| *** s10 has quit IRC | 15:04 | |
| *** e0ne has quit IRC | 15:06 | |
| tssurya | efried: I was looking into the tracker’s code to follow what improvisations you are doing in https://review.openstack.org/#/c/614886/ and had a question. | 15:08 |
| tssurya | are we updating prov_tree’s inventory based on both placement and virt driver during every periodic update? | 15:08 |
| efried | tssurya: Yes, but in an as-needed fashion. update_from_provider_tree sees if anything has changed, and only hits placement (and updates the cache) if something has. | 15:09 |
| efried | tssurya: And only for the providers and aspects thereof that have changed | 15:10 |
| efried | tssurya: So for example, if you had a whole tree and sharing provider setup, and you only changed an aggregate association on one provider, we would only do one call to placement. | 15:10 |
| efried | tssurya: The vast majority of the time, nothing has changed, and we'll do zero calls to placement. (Currently we still do the initial rp GET with ?in_tree and the inventory GET every time, even if you have the interval set high. Those are going away with my patches.) | 15:11 |
| tssurya | efried: okay.. so for me what's confusing is https://github.com/openstack/nova/blob/master/nova/compute/resource_tracker.py#L928 versus https://github.com/openstack/nova/blob/90b96170d3f269165f649e8b61739cf31ffb78b8/nova/scheduler/client/report.py#L992 | 15:12 |
| tssurya | like leaving the perforamance POV out of the discussion, | 15:12 |
| tssurya | what is the source of truth here ? we first compare the prov_tree's inventory cache with the GET result from placement and see if something has changed | 15:13 |
| tssurya | and then we do the same with the virt driver ? | 15:13 |
| efried | tssurya: The flow goes like this: | 15:15 |
| efried | 1) The resource tracker uses get_provider_tree_and_ensure_root to make sure it has the latest and greatest representation of the provider data. (This is what my code series is looking to optimize to where it's almost always a no-op because we already had the right picture.) | 15:16 |
| efried | 2) The resource tracker passes that information to the virt driver's update_provider_tree method. This is where the virt driver gets to decide if anything about it should change. | 15:16 |
| tssurya | ah okay so finally virt driver is the source of truth right ? | 15:17 |
| efried | 3) The resource tracker passes the possibly-modified-by-virt provider tree object to update_from_provider_tree (confusing naming, but note the *from* in that one). ufpt is what compares that provider tree with what's in the cache (I should have mentioned that get_provider_tree_and_ensure_root is returning a *copy*, so ufpt still has the baseline to compare from) and flushes any changes back to placement. | 15:17 |
| efried | tssurya: Yes, correct, the virt driver is the source of truth. | 15:18 |
| tssurya | efried: okay so got it, and I also now udnerstand your improvisations.. | 15:18 |
| tssurya | I have something here: https://docs.google.com/document/d/1FWdwQK0gCdXt7LlDI52K5R2j8wbdUCYG8OdAn_h5FgU/edit?usp=sharing | 15:18 |
| tssurya | can you verify this ? | 15:18 |
| tssurya | its basically that we are trying to see what all syncs we can safely disable | 15:19 |
| tssurya | for now | 15:19 |
| efried | tssurya: I'm hoping to have this series in a place where you could just pull it down and try it soon, like by tomorrow. I may not have all the unit/func tests polished up, but if CI passes, we can be confident it at least doesn't break the world too badly. | 15:22 |
| efried | reading... | 15:23 |
| tssurya | also is there anyway we could optimize doing both https://github.com/openstack/nova/blob/90b96170d3f269165f649e8b61739cf31ffb78b8/nova/scheduler/client/report.py#L698 and https://github.com/openstack/nova/blob/90b96170d3f269165f649e8b61739cf31ffb78b8/nova/scheduler/client/report.py#L1022 check | 15:23 |
| efried | tssurya: Oh, you may also want to be looking at the new series where I'm trying to do all of this in pieces: https://review.openstack.org/#/c/615606/ | 15:24 |
| tssurya | probably the lists of RP's for which you do these are different, and they must be necessary | 15:24 |
| tssurya | efried: ack, looking | 15:24 |
| *** ttsiouts has joined #openstack-placement | 15:25 | |
| *** cdent has joined #openstack-placement | 15:36 | |
| cdent | stephenfin: this https://review.openstack.org/#/c/611719/ and this https://review.openstack.org/#/c/613118/ are potentially in your wheelhouse | 15:42 |
| *** ttsiouts has quit IRC | 15:46 | |
| *** ttsiouts has joined #openstack-placement | 15:47 | |
| cdent | mriedem: you have any ideas who might like to look at your grenade change? | 15:49 |
| mriedem | dansmith | 15:50 |
| *** ttsiouts has quit IRC | 15:50 | |
| cdent | Does dansmith still exist? I thought he moved away to a comfortable island where nobody gets on his lawn. | 15:50 |
| *** ttsiouts has joined #openstack-placement | 15:50 | |
| mriedem | he does | 15:51 |
| mriedem | also probably busy preparing for summit talk and such | 15:51 |
| cdent | edleafe: you been able to catch up yet, or is other stuff holding you in its steely grasp? | 15:55 |
| mriedem | fwiw, i plan to go through the alembic series at some point here, but need to get my cross-cell resize spec wip started before the summit and need to wrap up some other stuff this week | 15:56 |
| cdent | mriedem: yeah, I think everybody is in that need/want to get X done before next week | 15:58 |
| dansmith | yeah I'm just overbooked with things | 16:00 |
| stephenfin | cdent: done | 16:13 |
| edleafe | cdent: sorry, was afk | 16:14 |
| edleafe | cdent: yeah, I'm just catching up. Amazing that one day off puts you so far behind | 16:14 |
| cdent | stephenfin: thanks, don't feel bad about -1ing, that's what review is for | 16:14 |
| edleafe | cdent: Anything in particular I need to know? | 16:14 |
| cdent | edleafe: well, I did work 8 hours on saturday and 17 hours yesterday on that stuff ... | 16:14 |
| edleafe | slacker | 16:14 |
| cdent | just that it perhaps finally works, but you might want to compare with some ps version that you're familiar with and if there are changes that make no sense, ask me, so at least one more person than me has some insight | 16:16 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Clean up and clarify tox.ini https://review.openstack.org/611719 | 16:19 |
| edleafe | cdent: roger that | 16:20 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Make tox -ereleasenotes work https://review.openstack.org/613118 | 16:36 |
| *** ttsiouts has quit IRC | 16:50 | |
| *** helenafm has quit IRC | 16:53 | |
| cdent | stephenfin: you probably gone now (good for you) but here's another tox-related thing that might intrigue: https://review.openstack.org/#/c/614559/ | 17:23 |
| cdent | edleafe: did you get a chance to do/review https://review.openstack.org/#/c/613343/ ? | 17:25 |
| cdent | efried: you were previous +2 on https://review.openstack.org/#/c/611719/ and it needs a final kick and this one is nearby: https://review.openstack.org/#/c/613118/ | 17:26 |
| efried | cdent: Yeah, that one has changed a lot since my +2. I had intended to pull it down and try it again, just haven't gotten to it. | 17:27 |
| efried | urgency? | 17:27 |
| edleafe | cdent: not yet. I started trying to recreate your setup, but there were a bunch of requirements that I was missing, and never got to finish | 17:27 |
| edleafe | cdent: that said, I'm always in favor of incremental improvements to docs rather than insisting that they be perfect from the start | 17:28 |
| cdent | efried: not so much urgency as it's been sat there for a long time and it would be nice to get some stuff trundling in | 17:28 |
| efried | yup, agree. A lot of stuff fits that description unfortunately. | 17:28 |
| cdent | edleafe: do you recall which requirements? things I didn't mention or stuff I did and you didn't have? | 17:29 |
| cdent | efried: true stories | 17:29 |
| edleafe | cdent: no, which is why I wanted to start from scratch with a fresh VM and document them | 17:29 |
| edleafe | I tried it on my dev system, which didn't even have a database installed | 17:30 |
| * cdent nods | 17:31 | |
| *** e0ne has joined #openstack-placement | 17:32 | |
| *** e0ne has quit IRC | 18:03 | |
| *** e0ne has joined #openstack-placement | 18:06 | |
| *** e0ne has quit IRC | 18:24 | |
| *** tssurya has quit IRC | 18:26 | |
| edleafe | cdent: kudos for unraveling the Opportunistic stuff. It was very much a black box of confusion. | 18:45 |
| cdent | still is! | 18:46 |
| edleafe | :) | 18:46 |
| cdent | I'd get really close and then either find and edge case in local testing or in the gate | 18:46 |
| cdent | luckily it was saturday and the gate was fast | 18:47 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Correct lower-constraints.txt and the related tox job https://review.openstack.org/614559 | 19:05 |
| cdent | magic is my enemy, this time pbr | 19:06 |
| cdent | I've lost something like 20 hours in the last few days to openstack's predilection to make things more "convenient" for me without actually making it obvious | 19:07 |
| *** mriedem is now known as mriedem_vote | 19:18 | |
| cdent | bbl | 19:23 |
| * cdent waves | 19:23 | |
| *** cdent has quit IRC | 19:23 | |
| *** mriedem_vote is now known as mriedem | 20:20 | |
| openstackgerrit | Adam Spiers proposed openstack/nova-specs master: Add spec for libvirt driver launching AMD SEV-encrypted instances https://review.openstack.org/609779 | 21:59 |
| openstackgerrit | Sundar Nadathur proposed openstack/nova-specs master: Nova Cyborg interaction specification. https://review.openstack.org/603955 | 23:14 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!