| openstackgerrit | Dan Smith proposed openstack/nova master: Batch results per cell when doing cross-cell listing https://review.openstack.org/592698 | 00:15 |
|---|---|---|
| openstackgerrit | Dan Smith proposed openstack/nova master: List instances from all cells explicitly https://review.openstack.org/593717 | 00:15 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Make instance_list perform per-cell batching https://review.openstack.org/593131 | 00:15 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Record cell success/failure/timeout in CrossCellLister https://review.openstack.org/594265 | 00:15 |
| openstackgerrit | Dan Smith proposed openstack/nova master: WIP: Optimize global marker re-lookup in multi_cell_list https://review.openstack.org/594577 | 00:15 |
| *** lei-zh has joined #openstack-placement | 02:24 | |
| *** lei-zh has quit IRC | 02:33 | |
| *** lei-zh1 has joined #openstack-placement | 02:33 | |
| *** edleafe has quit IRC | 02:46 | |
| *** edleafe has joined #openstack-placement | 02:47 | |
| *** tetsuro_ has joined #openstack-placement | 03:10 | |
| *** tetsuro has quit IRC | 03:10 | |
| *** nicolasbock has quit IRC | 03:32 | |
| *** lei-zh1 has quit IRC | 03:41 | |
| openstackgerrit | Merged openstack/nova master: Making consistent used of GiB and MiB in API ref https://review.openstack.org/559985 | 04:03 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Make monkey patch work in uWSGI mode https://review.openstack.org/592285 | 04:04 |
| openstackgerrit | Sam Morrison proposed openstack/nova master: Allow ability for non admin users to use all filters on server list. https://review.openstack.org/526558 | 04:06 |
| openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Merge extended server attributes extension response https://review.openstack.org/584590 | 04:27 |
| *** alex_xu has joined #openstack-placement | 05:32 | |
| *** lei-zh has joined #openstack-placement | 05:35 | |
| openstackgerrit | Merged openstack/nova master: Merge config drive extension response into server controller https://review.openstack.org/584223 | 05:37 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Make monkey patch work in uWSGI mode https://review.openstack.org/592285 | 06:21 |
| *** tssurya has joined #openstack-placement | 06:47 | |
| openstackgerrit | Merged openstack/nova master: Use CONF.long_rpc_timeout in post_live_migration_at_destination https://review.openstack.org/588668 | 06:51 |
| *** rubasov has joined #openstack-placement | 07:07 | |
| *** e0ne has joined #openstack-placement | 08:01 | |
| *** s10 has joined #openstack-placement | 08:30 | |
| *** cdent has joined #openstack-placement | 08:31 | |
| *** lei-zh has quit IRC | 09:20 | |
| openstackgerrit | Sergii Golovatiuk proposed openstack/nova master: libvirt: Always escape IPv6 addresses when used in migration URI https://review.openstack.org/589548 | 09:26 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: conf: Gather 'live_migration_scheme', 'live_migration_inbound_addr' https://review.openstack.org/456572 | 09:55 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: conf: Convert 'live_migration_inbound_addr' to HostAddressOpt https://review.openstack.org/456573 | 09:55 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: conf: Add three new '[libvirt] live_migration_*' options https://review.openstack.org/456571 | 09:55 |
| *** sean-k-mooney has joined #openstack-placement | 10:00 | |
| openstackgerrit | Chen proposed openstack/nova master: Normalize dashless 'resource provider create' uuid https://review.openstack.org/567191 | 10:04 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: api: Remove unnecessary default parameter https://review.openstack.org/564451 | 10:09 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: objects: Remove legacy '_from_dict' functions https://review.openstack.org/537414 | 10:21 |
| *** lei-zh has joined #openstack-placement | 10:27 | |
| openstackgerrit | Surya Seetharaman proposed openstack/nova master: WIP: Record cell success/failure/timeout in CrossCellLister https://review.openstack.org/594265 | 10:27 |
| openstackgerrit | Surya Seetharaman proposed openstack/nova master: Add get_by_cell_and_project() method to InstanceMappingList https://review.openstack.org/591656 | 10:27 |
| openstackgerrit | Surya Seetharaman proposed openstack/nova master: Return a minimal construct for nova list when a cell is down https://review.openstack.org/567785 | 10:27 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: conf: Use new-style choice values https://review.openstack.org/530924 | 10:34 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: Modify PciDevice.uuid generation code https://review.openstack.org/530487 | 10:37 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: Add an online migration for PciDevice.uuid https://review.openstack.org/530905 | 10:37 |
| *** s10 has quit IRC | 10:42 | |
| *** nicolasbock has joined #openstack-placement | 10:44 | |
| openstackgerrit | Surya Seetharaman proposed openstack/nova master: Record cell success/failure/timeout in CrossCellLister https://review.openstack.org/594265 | 10:45 |
| openstackgerrit | Surya Seetharaman proposed openstack/nova master: WIP: Optimize global marker re-lookup in multi_cell_list https://review.openstack.org/594577 | 10:45 |
| *** lei-zh has quit IRC | 10:46 | |
| *** gibi_off is now known as gibi | 11:40 | |
| * gibi waves | 11:40 | |
| cdent | welcome back gibi | 11:41 |
| *** mriedem_away is now known as mriedem | 11:54 | |
| *** mriedem has left #openstack-placement | 11:57 | |
| *** mriedem has joined #openstack-placement | 11:57 | |
| *** mriedem has quit IRC | 12:01 | |
| *** mriedem has joined #openstack-placement | 12:01 | |
| openstackgerrit | Sergii Golovatiuk proposed openstack/nova master: libvirt: Always escape IPv6 addresses when used in migration URI https://review.openstack.org/589548 | 12:08 |
| openstackgerrit | Surya Seetharaman proposed openstack/nova master: [WIP] Add scatter_gather_single_cell utility https://review.openstack.org/594947 | 12:35 |
| openstackgerrit | Surya Seetharaman proposed openstack/nova master: [WIP] Add scatter_gather_single_cell utility https://review.openstack.org/594947 | 12:37 |
| efried | Hi gibi! | 12:46 |
| *** tbarron has joined #openstack-placement | 12:55 | |
| efried | gibi: I'm sure you'll be catching up for a while, but the reshaper series is (at least as of 14h ago) freshly rebased, comment-and-nit-addressed, and ready for review | 12:58 |
| efried | cdent: assume you spy'd that last night too? | 12:58 |
| cdent | efried: i did yes, and mean to get to it, but got pulled into an internal impromptu "explain how placement works" session | 12:59 |
| cdent | which, now complete, I should be able to carry on | 13:00 |
| efried | cool beans | 13:00 |
| efried | alex_xu: https://review.openstack.org/#/c/591863/ got a minor tweak to be rebased to the *bottom* of the series (it was on top before). Could you re+A at your convenience please? | 13:00 |
| edleafe | Ugh - started the run_filter.sh script yesterday at 4:30pm. and 15 hours later it's still running | 13:02 |
| openstackgerrit | Matt Riedemann proposed openstack/nova master: libvirt: Always escape IPv6 addresses when used in migration URI https://review.openstack.org/589548 | 13:02 |
| efried | edleafe: That sounds... not good. What's run_filter.sh? | 13:03 |
| edleafe | It's the script for extraction. You pass it a bunch of files you want to keep in a repo, and it removes the rest, along with every bit of history that doesn't include the chosen files | 13:04 |
| edleafe | Sorry, run_filter.sh is my local script with all the files | 13:04 |
| gibi | efried: ack, try to reach that point, but first a sync with the my internal team | 13:04 |
| edleafe | The actual script is filter_git_history.sh | 13:04 |
| cdent | efried: https://git.openstack.org/cgit/openstack/oslo.tools/tree/filter_git_history.sh | 13:05 |
| edleafe | efried: the script: http://paste.openstack.org/show/728591/ | 13:05 |
| efried | gibi: Thanks. I think Jay gets back tomorrow. | 13:06 |
| alex_xu | efried: checking now | 13:06 |
| gibi | efried: that is a good target date for me :) | 13:06 |
| efried | edleafe: Do you have a sense of how far along it is? | 13:08 |
| cdent | edleafe: is git parellelizing at all, or is it doing all that stuff with one process and/or thread? | 13:08 |
| cdent | because if it is, I can try it on a big vm with many cores if yours is leeetl | 13:09 |
| edleafe | cdent: looks like a single process | 13:09 |
| cdent | feh | 13:10 |
| efried | edleafe: Is it in the merge-base loop or stuck on the filter-branch step? | 13:10 |
| edleafe | how would I tell? | 13:11 |
| efried | edleafe: Well, the filter_git_history.sh has set -x in it, so the output should be tracing, nah? | 13:12 |
| cdent | if it turns out it is going to take for ever, another option would be to make a shallow clone that only goes back to the first placement commit, so less history would need to be edited | 13:12 |
| cdent | efried: you're misreading the code | 13:13 |
| cdent | the script is building a series of shell sript fragments which git filter-branch then runs. once. | 13:13 |
| alex_xu | efried: is there any purpose for missing this two https://review.openstack.org/#/c/591863/6/nova/api/openstack/placement/schemas/allocation.py ? | 13:13 |
| cdent | well, more specifically, git filter-branch runs once, running those fragments many times | 13:13 |
| efried | cdent: swhy I'm asking whether the script is still spinning in the merge-base loop, which you should be able to tell if it's spewing a bunch of git merge-base traces, versus in the filter-branch part, which you would see by the output being static and the last thing being git filter-branch <arglist-of-hugeness> | 13:14 |
| cdent | and what I'm saying is that's not how it works | 13:15 |
| cdent | (as I've tried it)( | 13:15 |
| cdent | look at how $root $set_roots and $skip_empty are used: in filter-branch | 13:16 |
| cdent | now, if you could watch top while it is running, you should see a bunch of various git commands, but they are all subprocesses of filter-brach | 13:16 |
| efried | I understand. I think you're answering my question by implying that L26-40 is already done. | 13:17 |
| openstackgerrit | Merged openstack/nova stable/rocky: Update UPPER_CONSTRAINTS_FILE for stable/rocky https://review.openstack.org/590731 | 13:18 |
| edleafe | cdent: efried: This is what the monitor looks like: https://www.youtube.com/watch?v=0048pPbPPnw | 13:19 |
| efried | cdent, edleafe: You could look at pstree on the script process to see the git subprocesses. If it's doing things in some kind of deterministic order, you might be able to tell how far along it is. | 13:19 |
| efried | wow | 13:20 |
| efried | okay | 13:20 |
| efried | gotta go do a cat thing, bbiab | 13:20 |
| edleafe | meow | 13:20 |
| openstackgerrit | Merged openstack/osc-placement stable/rocky: Update .gitreview for stable/rocky https://review.openstack.org/589773 | 13:21 |
| *** bodgix has joined #openstack-placement | 13:21 | |
| edleafe | efried: pstree shows a single process | 13:23 |
| cdent | I think it just takes a long time... | 13:24 |
| cdent | but i'm poking locally to see if I can see some options | 13:24 |
| efried | renice? | 13:26 |
| openstackgerrit | Merged openstack/osc-placement stable/rocky: Update UPPER_CONSTRAINTS_FILE for stable/rocky https://review.openstack.org/589774 | 13:27 |
| cdent | actually, edleafe I think it _might_ be being broken | 13:28 |
| cdent | meh, dunno, still reading | 13:30 |
| * cdent runs out of disk space | 13:33 | |
| edleafe | I think it's broken | 13:34 |
| edleafe | I looked at a frame from the video where it flashes, and the output is a git merge-base command, followed by a "usage:" listing for that command | 13:35 |
| cdent | it's producing the same ids for roots | 13:35 |
| cdent | yeah, I'm getting that too, but I was assuming it was because I was in a shallow clone | 13:36 |
| cdent | i've created a log, I'm going to try to idenitfy which file started that cascade | 13:36 |
| edleafe | I'm killing it | 13:36 |
| edleafe | the matrix-y output is just the same hash repeated over and over | 13:37 |
| cdent | edleafe: it's a typo in your list of files | 13:37 |
| edleafe | doh! | 13:37 |
| edleafe | where? | 13:37 |
| cdent | there is no nova/db/api directory | 13:38 |
| edleafe | huh | 13:39 |
| edleafe | those entries were all taken from grep output, with the irrelevant files removed | 13:40 |
| efried | So it *was* stuck in the merge-base loop. | 13:40 |
| cdent | looks like you've mixed and matched with sqlalchemy and api in different places. the constants file: nova/db/constants.py | 13:40 |
| edleafe | IOW, I didn't type that in | 13:40 |
| cdent | efried: yes, you can have today's I told you so cookie | 13:40 |
| efried | nomnomnom | 13:41 |
| * edleafe sends efried a cookie | 13:41 | |
| * cdent finds a suitable emoji somewhere | 13:41 | |
| * dims launches a MITM attack for cookies | 13:41 | |
| cdent | edleafe: there are some other errors in filenames to | 13:43 |
| cdent | i'll make a list | 13:43 |
| edleafe | yeah, I'm checking them all now | 13:43 |
| cdent | a lot of them have prefixes they shouldn't: | 13:44 |
| cdent | http://paste.openstack.org/show/728596/ | 13:44 |
| cdent | those are the wrong ones (not -r ), these are the right: http://paste.openstack.org/show/728597/ | 13:46 |
| cdent | edleafe: I'm guess you did a filter or grep or something from not quite the right place | 13:46 |
| edleafe | yeah, I think I ran grep for some of them from the parent directory | 13:46 |
| edleafe | jinxish | 13:46 |
| cdent | cookies for everyone: http://images.cookingforengineers.com/pics2/640/DSC_2782_crop.jpg | 13:47 |
| cdent | because why not? | 13:47 |
| cdent | (by the way, I found that image by searching for: eric's big cookie) | 13:48 |
| edleafe | I might have been afraid to search for that | 13:50 |
| cdent | I was worried, but I decided to be brave | 13:51 |
| edleafe | Fixed the file paths, and now it seems to be running much more smoothly | 13:59 |
| efried | woot | 14:00 |
| cdent | edleafe: has it started reporting seconds at the bottom? | 14:03 |
| *** nicolasbock has quit IRC | 14:04 | |
| *** s10 has joined #openstack-placement | 14:05 | |
| edleafe | yes | 14:08 |
| cdent | how's it seem? | 14:09 |
| edleafe | 680 seconds passed, remaining 1578 predicted | 14:10 |
| *** alex_xu has quit IRC | 14:10 | |
| edleafe | So under a half-hour remaining | 14:11 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Make RecordWrapper record RequestContext and expose cell_uuid https://review.openstack.org/594572 | 14:11 |
| cdent | edleafe: I think I've seen the prediction change when messing before, but who knows. not much time is nice | 14:14 |
| efried | better than 15h | 14:14 |
| cdent | do we need to make a hole for you to push it to? | 14:14 |
| *** efried is now known as efried_goatin | 14:14 | |
| cdent | sometime he be rollin, sometime he be goatin | 14:15 |
| edleafe | cdent: that was the next thing to decide. Should I push it to my personal git repo for initial review, and once it seems close, push it to openstack/placement? | 14:31 |
| *** nicolasbock has joined #openstack-placement | 14:31 | |
| cdent | edleafe: I think it should go to some git repo, your personal github is a fine choice, where we (added as contributors) masage it until it is some semblance of "okay" (tests pass? some other metric? we should decide here) | 14:32 |
| cdent | at which point we seed the git.openstack.org repo with it | 14:32 |
| edleafe | Yeah, I don't want to have a ton of churn history as we figure out what is needed. | 14:33 |
| edleafe | So churn in private, and when it seems to be OK, re-run the extraction on current nova master, and push to the openstack org | 14:33 |
| cdent | a) I don't follow on the re-run the extraction. Do you mean that our churn is a thing which is re-defining the extraction list? Or something else | 14:35 |
| edleafe | redefining | 14:35 |
| edleafe | IOW, we missed something, or included something that didn't belong, etc | 14:35 |
| cdent | b) for me the post extraction massaging is to go from "this looks like nova" to "this looks like placmeent and the tests pass" | 14:36 |
| cdent | c) so if we have to do multiple extractions (which is fine) there will still be a post-extraction, pre-seed period of "arranged things correctly" | 14:36 |
| edleafe | Once we get the files cleanly extracted, we push to openstack/placement, and then remove the nova-isms that are in the files there, as that's legit history to preserve | 14:36 |
| cdent | oh, I was still assuming we'd preserve the re-arranging history, but do it somewhere without a gate | 14:37 |
| cdent | so once we did have git.o.o repo it can (at leastly midly) gate | 14:37 |
| edleafe | oh, I didn't get the gate concern | 14:37 |
| edleafe | that makes sense | 14:37 |
| edleafe | ok, so how about this game plan: | 14:37 |
| edleafe | push the extracted code to my git repo | 14:38 |
| edleafe | we review it there, and find any missing/wrong files | 14:38 |
| edleafe | If needed, we wipe that repo, re-extract with the changes, and push again to a private repo | 14:38 |
| edleafe | Once we're happy with the files, we iterate on the re-arrangements, until we have something where tests are all passing | 14:39 |
| edleafe | ^^ on the private repo | 14:39 |
| edleafe | When that's done, push that to an openstack org repo | 14:40 |
| edleafe | Does that sound correct> | 14:40 |
| edleafe | ? | 14:40 |
| cdent | that's aligned with what I was thinking. I suspect we will hit bumps along the way, but that's OK | 14:41 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Batch results per cell when doing cross-cell listing https://review.openstack.org/592698 | 14:41 |
| openstackgerrit | Dan Smith proposed openstack/nova master: List instances from all cells explicitly https://review.openstack.org/593717 | 14:41 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Make instance_list perform per-cell batching https://review.openstack.org/593131 | 14:41 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Record cell success/failure/timeout in CrossCellLister https://review.openstack.org/594265 | 14:41 |
| openstackgerrit | Dan Smith proposed openstack/nova master: WIP: Optimize global marker re-lookup in multi_cell_list https://review.openstack.org/594577 | 14:41 |
| edleafe | cdent: I'm sure there will be bumps. I'd just prefer that for early, major screw-ups, we start clean | 14:42 |
| cdent | yes | 14:42 |
| edleafe | Since we're all n00bs at this git extraction stuff | 14:42 |
| cdent | ayup | 14:43 |
| cdent | how are your seconds seconding now? | 14:43 |
| edleafe | So I've got good new and bad news | 14:47 |
| edleafe | The good news: the script finished | 14:47 |
| edleafe | The bad news: it didn't remove what it should have removed | 14:47 |
| *** s10 has quit IRC | 14:47 | |
| openstackgerrit | Matt Riedemann proposed openstack/nova master: Wait for network-vif-plugged on resize revert https://review.openstack.org/595069 | 14:48 |
| cdent | edleafe: when you've done it before (with a shorter list) did it behave as expected, and now for some reason it is being wrong? | 14:48 |
| edleafe | pretty much | 14:48 |
| cdent | what's an example of something that unexpectedly stayed? | 14:49 |
| edleafe | nova/api/openstack/compute/aggregates.py | 14:49 |
| edleafe | My guess is that it shares commit history with some placement files | 14:49 |
| edleafe | so it was preserved | 14:50 |
| cdent | yeah, which perhaps is just the way it goes | 14:50 |
| cdent | does it seem like everything you expect to be there is there? | 14:50 |
| edleafe | http://paste.openstack.org/show/728605/ | 14:50 |
| cdent | whatever we do there is always going to cleanup and further tweaks to get things in the right places | 14:50 |
| edleafe | The stuff that was saved there were all agg stuff | 14:50 |
| openstackgerrit | Matt Riedemann proposed openstack/nova master: Wait for network-vif-plugged on resize revert https://review.openstack.org/595069 | 14:51 |
| cdent | doing a git log on some of those file will probably provide a clue to what the related placement stuff was | 14:51 |
| cdent | I think it's fine. We clean that stuff up in post. Or whatever we want to call it. I don't think we can get total cleanliness automatically, for exacly the reason you've said: some commits touch too many files | 14:51 |
| cdent | s/too many/a lot of/ | 14:52 |
| edleafe | ugh, just realized I was supposed to be on a call 20 minutes ago | 14:52 |
| cdent | ta ra | 14:53 |
| edleafe | I'll look more after my slew of meetings is over | 14:53 |
| mriedem | efried_goatin: so none of the virt drivers (libvirt/xen) are actually reshaping provider trees yet in that series right? | 14:53 |
| cdent | if you can go ahead and push it somewhere we can poke and prod at our combined leisures | 14:53 |
| edleafe | I'll push something later | 14:53 |
| cdent | mriedem: last I checked that was the case, but that was some days ago | 14:55 |
| cdent | mriedem: as I recall no virtdriver people had chimed in on how that might actually work (in terms of the representation of stuff being inspected by the virt driver). efried_goatin got as far as "here's the interface you must satisfy" | 14:56 |
| *** efried_goatin is now known as efried | 15:06 | |
| efried | mriedem: Correct. The closest thing is one of the test cases that implements an update_provider_tree method that does reshaping in the spirit of what a real driver would do. | 15:07 |
| efried | mriedem: This one: https://review.openstack.org/#/c/576236/30/nova/tests/functional/test_servers.py@1578 (actually used in a few test cases, but test_reshape is the full "good path"). | 15:15 |
| mriedem | ok, hmm | 15:19 |
| mriedem | not sure if bauzas is present these days | 15:20 |
| mriedem | but assume the xen driver people are | 15:20 |
| mriedem | dansmith: do you know if bauzas was going to work on the actual reshaping of nrps in the libvirt driver? | 15:20 |
| dansmith | mriedem: I haven't seen him in quite a while, so I dunno.. but not that I know of | 15:21 |
| mriedem | hmm, ok, will see who is doing the xen stuff | 15:24 |
| mriedem | naichuan sun | 15:25 |
| dansmith | IMHO, seeing a virt driver that supports numa (or similar) reshape actual inventory and allocations across a grenade boundary is the flag indicating to me that this works | 15:26 |
| mriedem | sure, but, | 15:26 |
| mriedem | ... | 15:26 |
| dansmith | I know, just sayin' | 15:26 |
| mriedem | well, i guess i'll send some emails then | 15:27 |
| mriedem | done | 15:31 |
| *** openstackgerrit has quit IRC | 15:31 | |
| efried | dansmith, mriedem: Are we suggesting waiting to land the series until someone is implementing a real reshape? | 15:34 |
| *** efried has quit IRC | 15:36 | |
| *** efried has joined #openstack-placement | 15:36 | |
| *** openstackgerrit has joined #openstack-placement | 15:38 | |
| openstackgerrit | sahid proposed openstack/nova master: compute: reset instance events https://review.openstack.org/420026 | 15:38 |
| mriedem | i haven't suggested anything | 15:38 |
| mriedem | i want to know if/when xenapi/libvirt people that care about vgpu stuff are going to make the chnages | 15:38 |
| mriedem | *changes | 15:38 |
| mriedem | hence my email | 15:38 |
| dansmith | he left, but, | 15:39 |
| mriedem | if it's "i'm busy for a month so can't" | 15:39 |
| mriedem | efried: is here | 15:39 |
| dansmith | oh, I missed that he re-joined | 15:39 |
| dansmith | I'm not suggesting we not land some or all of that code necessarily, it's just that I don't think it's finished until we can actually use it, | 15:39 |
| dansmith | meaning NRPs (from a nova perspective) | 15:39 |
| dansmith | and I guess vGPU isn't the primary driving force I'm thinking of here, it's NUMA, because I expect vGPU needs NUMA to not have to reshape *again* later | 15:40 |
| mriedem | we can also go through the series and +1/+1.9 with the -2 pin on the api change to start | 15:40 |
| dansmith | which api change? | 15:41 |
| efried | I'll play devil's advocate by saying that getting "all of NRP" working, including actual use by drivers, is going to involve many changes in multiple series, and holding up the entire mesh until it's all proven is a recipe for total deadlock. | 15:41 |
| efried | not to mention rebase hell. | 15:41 |
| efried | dansmith: The API change https://review.openstack.org/#/c/576927/ is -2 pending review up the whole series. | 15:42 |
| mriedem | the reshaper api | 15:42 |
| mriedem | yeah that | 15:42 |
| efried | dansmith: but the series only goes as far as the resource tracker work, not any actual driver impls. | 15:42 |
| dansmith | okay, I guess I thought given the current discussion(s) that we wouldn't hold that up for placement based on nova integration | 15:42 |
| dansmith | it's definitely something I'm concerned about, mind you, | 15:42 |
| dansmith | but few others seem to, so I guess I was expecting we'd just push that in and deal with the "does it work for us" question later as that seems to be the sentiment | 15:43 |
| dansmith | efried: right, I know that it doesn't actually get used | 15:43 |
| efried | I agree with that. | 15:43 |
| *** nicolasbock has quit IRC | 15:43 | |
| efried | push it now, fix it if it breaks when we use it | 15:43 |
| efried | Like, agile or something. | 15:43 |
| mriedem | efried: i assume your functional test is using a similar scenario as libvirt will be in? | 15:44 |
| mriedem | with vgpu inventory on the root and needs to move to a child? | 15:44 |
| efried | mriedem: Well, it's trying to use a kind of representative cross-section of possible moves. | 15:44 |
| efried | let me look at it again... | 15:44 |
| mriedem | https://review.openstack.org/#/c/576236/30/nova/tests/functional/test_servers.py@1840 | 15:44 |
| efried | cause it's been a while since I wrote it :) | 15:45 |
| efried | mriedem: okay, so it's actually moving VCPU and MEMORY_MB to NUMA node RPs, and DISK_GB to a sharing RP. | 15:46 |
| mriedem | and some PF/VFs | 15:46 |
| efried | and creating SRIOV_NET_VF on new PF RPs under the NUMA node RPs. | 15:46 |
| efried | yeah. | 15:46 |
| efried | (dansmith we're looking at the block starting https://review.openstack.org/#/c/576236/30/nova/tests/functional/test_servers.py@1643) | 15:47 |
| efried | mechanically, moving VGPUs from the compute RP to child PGPU RPs would be pretty much identical to what this test is doing with the VCPU/MEMORY_MB - just with different names for stuff. | 15:48 |
| *** nicolasbock has joined #openstack-placement | 15:49 | |
| mriedem | left a couple of comments, | 15:49 |
| mriedem | but couldn't a fake virt driver that does the reshape stuff been written to simulate this? | 15:49 |
| mriedem | ok i see https://review.openstack.org/#/c/576236/30/nova/tests/functional/test_servers.py@1465 | 15:50 |
| mriedem | so, | 15:50 |
| mriedem | this mocks out update_provider_tree in the fake virt driver this test is using | 15:50 |
| efried | yes | 15:50 |
| mriedem | ok | 15:52 |
| dansmith | and just to be clear, nothing else wants to use reshaper, it's just that we want to put it in the tree to avoid having to keep rebasing this yeah? | 15:52 |
| mriedem | not sure what you mean by "wants" | 15:52 |
| *** tssurya has quit IRC | 15:53 | |
| efried | dansmith: There's no patches based on this yet, if that's what you're asking. | 15:53 |
| mriedem | dansmith: can you translate this? "Sorry I'm still on PTO from 2 weeks to the next Wednesday in 8 days." | 15:53 |
| efried | dansmith: So yes, we want to merge it so we don't have to keep rebasing it, and so any consumers can develop against master instead of an in-flight series. | 15:53 |
| dansmith | mriedem: no, but I can look on our pto calendar :) | 15:54 |
| mriedem | i think it means, | 15:54 |
| mriedem | "i've been on PTO for 2 weeks, and will be back next thursday" | 15:54 |
| mriedem | i.e. "i live in france and we have the entire month off" | 15:54 |
| dansmith | he's back on thursday the 30th | 15:54 |
| mriedem | ok so sylvain is at least a week out | 15:54 |
| mriedem | i was right! | 15:54 |
| mriedem | bauzas translator 5000 | 15:55 |
| * cdent is very impressed (seriously) | 15:55 | |
| openstackgerrit | Balazs Gibizer proposed openstack/nova-specs master: Add subtree filter for GET /resource_providers https://review.openstack.org/595236 | 15:55 |
| mriedem | i've gotten used to how he uses "from" | 15:55 |
| mriedem | and "that" | 15:55 |
| mriedem | tip: that -> this | 15:55 |
| efried | and vice versa | 16:00 |
| efried | like I will never understand the rule (if there is one) for using por vs para in Spanish | 16:01 |
| cdent | I used to know that one | 16:01 |
| mriedem | el bano is para pooping | 16:02 |
| mriedem | *es para | 16:02 |
| mriedem | lo siento | 16:02 |
| openstackgerrit | Eric Fried proposed openstack/nova master: Compute: Handle reshaped provider trees https://review.openstack.org/576236 | 16:03 |
| openstackgerrit | Balazs Gibizer proposed openstack/nova-specs master: Placement: support mixing required traits with any traits https://review.openstack.org/565741 | 16:03 |
| efried | mriedem: Good spot on the obsolete condition to handle redundant _update calls, fixed -^ | 16:03 |
| mriedem | \o/ | 16:04 |
| openstackgerrit | Eric Fried proposed openstack/nova master: [placement] Regex consts for placement schema https://review.openstack.org/591863 | 16:09 |
| openstackgerrit | Eric Fried proposed openstack/nova master: [placement] Add /reshaper handler for POST https://review.openstack.org/576927 | 16:09 |
| openstackgerrit | Eric Fried proposed openstack/nova master: reshaper: Look up provider if not in inventories https://review.openstack.org/585033 | 16:09 |
| openstackgerrit | Eric Fried proposed openstack/nova master: Make get_allocations_for_resource_provider raise https://review.openstack.org/584598 | 16:09 |
| openstackgerrit | Eric Fried proposed openstack/nova master: Report client: Real get_allocs_for_consumer https://review.openstack.org/584599 | 16:09 |
| openstackgerrit | Eric Fried proposed openstack/nova master: Report client: get_allocations_for_provider_tree https://review.openstack.org/584648 | 16:09 |
| openstackgerrit | Eric Fried proposed openstack/nova master: Report client: _reshape helper, placement min bump https://review.openstack.org/585034 | 16:09 |
| openstackgerrit | Eric Fried proposed openstack/nova master: Report client: update_from_provider_tree w/reshape https://review.openstack.org/585049 | 16:09 |
| openstackgerrit | Eric Fried proposed openstack/nova master: Compute: Handle reshaped provider trees https://review.openstack.org/576236 | 16:09 |
| openstackgerrit | Balazs Gibizer proposed openstack/nova-specs master: Network bandwidth resource provider https://review.openstack.org/595243 | 16:11 |
| *** efried is now known as efried_rollin | 16:14 | |
| openstackgerrit | Merged openstack/nova master: Handle binding_failed vif plug errors on compute restart https://review.openstack.org/587498 | 16:25 |
| openstackgerrit | Matthew Booth proposed openstack/nova-specs master: Add serial numbers for local disks https://review.openstack.org/595247 | 16:32 |
| openstackgerrit | Matt Riedemann proposed openstack/nova stable/queens: Wait for network-vif-plugged before starting live migration https://review.openstack.org/591735 | 16:37 |
| openstackgerrit | Matt Riedemann proposed openstack/nova stable/queens: DNM: Debug patch to test live migration waiting https://review.openstack.org/591775 | 16:37 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Batch results per cell when doing cross-cell listing https://review.openstack.org/592698 | 16:43 |
| openstackgerrit | Dan Smith proposed openstack/nova master: List instances from all cells explicitly https://review.openstack.org/593717 | 16:43 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Make instance_list perform per-cell batching https://review.openstack.org/593131 | 16:43 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Record cell success/failure/timeout in CrossCellLister https://review.openstack.org/594265 | 16:43 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Optimize global marker re-lookup in multi_cell_list https://review.openstack.org/594577 | 16:43 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Clarify which context is used by do_query() https://review.openstack.org/595250 | 16:43 |
| *** e0ne has quit IRC | 17:07 | |
| openstackgerrit | Merged openstack/nova master: libvirt: Always escape IPv6 addresses when used in migration URI https://review.openstack.org/589548 | 17:16 |
| edleafe | cdent: efried_rollin: So I tracked down at least one place where there was "cross-contaimination" of the repos: commit f722c75b03902b5c6dfabf84cffd680267bd0459 | 17:19 |
| edleafe | It updated the placement doc in doc/source/user/placement, as well as a functional test for nova aggs, and a nova release note | 17:20 |
| cdent | edleafe: my repo seems to think thats a bad object | 17:20 |
| cdent | do you have a change id associated with it? | 17:20 |
| edleafe | cdent: Ibd7aa4f8c4ea787774becece324d9051521c44b6 | 17:21 |
| edleafe | so a 'git show f722c75b03902b5c6dfabf84cffd680267bd0459' doesn't work for your repo? | 17:21 |
| cdent | corret, i have 5eda1fab85e907a59d3be36067bfe25250a7be56 | 17:21 |
| edleafe | hmm - wonder why the extraction would have changed that | 17:22 |
| cdent | (this is my plane jane nova repo) | 17:22 |
| * cdent watches jane fly | 17:22 | |
| cdent | edleafe: lost parent maybe | 17:22 |
| edleafe | You're right - if I switch back to master, that commit has a different hash | 17:22 |
| cdent | in any case I don't reckon this kind of cross contamination is of much concern is it? we just clean up, making a commit that says os | 17:23 |
| cdent | to me that's not bad history | 17:23 |
| cdent | it's just history | 17:23 |
| edleafe | yeah, that was what I was typing when you replied with the "bad object" comment | 17:24 |
| edleafe | It would be nice not to have it, but it can't be avoided | 17:24 |
| edleafe | I was just looking for confirmation that that's why those extra files were there | 17:25 |
| *** N3l1x has quit IRC | 17:25 | |
| *** dims has quit IRC | 17:25 | |
| edleafe | Gimme a few minutes to recover from too many meetings, and I'll push it | 17:25 |
| cdent | huzzah | 17:25 |
| *** dims_ has joined #openstack-placement | 17:36 | |
| *** sean-k-mooney has quit IRC | 17:57 | |
| mriedem | IT IS DONE https://review.openstack.org/#/c/576927/33 | 18:10 |
| *** mriedem is now known as mriedem_lunch | 18:14 | |
| *** ChanServ has quit IRC | 18:16 | |
| cdent | this is why we always need your good input mriedem_lunch | 18:17 |
| *** ChanServ has joined #openstack-placement | 18:21 | |
| *** barjavel.freenode.net sets mode: +o ChanServ | 18:21 | |
| edleafe | cdent: efried_rollin: Should I do the obvious trimming before pushing? | 18:36 |
| edleafe | on the extraction repo | 18:36 |
| cdent | edleafe: not sure it matters, as presumably we'll do this multiple times to iron out the bugs in the process (and catch up with changes) | 18:38 |
| edleafe | I was thinking that doing one commit with that will be the extraction marker when looking back over the history | 18:39 |
| cdent | yeah, that's make some sense | 18:40 |
| *** mriedem_lunch is now known as mriedem | 18:55 | |
| openstackgerrit | Matt Riedemann proposed openstack/nova stable/rocky: Handle binding_failed vif plug errors on compute restart https://review.openstack.org/595317 | 19:05 |
| cdent | efried_rollin: I won't get to the reshaper API fixes tonight, but if you can't or don't want to, I can do it in the morning. leave me some signals what matters | 19:08 |
| *** bodgix has left #openstack-placement | 19:19 | |
| cdent | efried_rollin: I might still be around later, actually, we'll see | 19:26 |
| *** mriedem has quit IRC | 19:35 | |
| *** mriedem has joined #openstack-placement | 19:38 | |
| *** efried_rollin is now known as efried | 19:47 | |
| efried | cdent: ack. Looking over them now. Thanks for the review mriedem. | 19:48 |
| edleafe | cdent: efried: OK, I've pushed the extracted placement repo to my github: https://github.com/EdLeafe/placement | 19:53 |
| edleafe | Please check it out, and make sure that there are no obvious oversights or mistakes. | 19:53 |
| edleafe | When it appears sane enough, we can push it to the openstack repo | 19:54 |
| edleafe | ugh just saw etc/nova - will fix | 19:54 |
| efried | cdent: Isn't there some way you can plug this into (a derivative of) placecat to see if it works? Or are we not trying to get that far yet? | 19:55 |
| cdent | efried: we're quite a distance from that, I would guess | 19:55 |
| efried | k | 19:56 |
| cdent | assuming we want package namespaces to be different | 19:56 |
| cdent | we've still got things like https://github.com/EdLeafe/placement/blob/master/placement/api/placement/handlers/aggregate.py#L20 | 19:56 |
| cdent | but yes, once we get minor tests passing, we ca make placecat talk to it | 19:57 |
| *** e0ne has joined #openstack-placement | 19:57 | |
| cdent | efried: i got far enough along exploring the required policy changes that I'll just do them, cool? | 19:58 |
| efried | cdent: yes please; I wouldn't know how. | 19:59 |
| edleafe | cdent: just fixed that | 19:59 |
| cdent | i mean, at this point, I would assume iterate on getting tests working | 20:00 |
| edleafe | I'm sure a lot of the scaffolding for tests will need attention | 20:00 |
| efried | duh, I guess I don't need to recheck the things above the API patch, cause they're going to get rebased. | 20:20 |
| efried | zu spät | 20:20 |
| cdent | i'm about halfway done | 20:23 |
| cdent | edleafe: less than you might think, I already did quite a lot of it | 20:25 |
| edleafe | \o/ | 20:27 |
| *** e0ne has quit IRC | 20:27 | |
| mriedem | cdent: heh i was going to offer to just do the policy changes myself, | 20:27 |
| mriedem | but they are pretty simple | 20:27 |
| mriedem | copy/paste/tweak | 20:27 |
| mriedem | the one thing i'd ask is a gabbit is added for trying to hit POST /reshaper with x-auth-token: user and assert we get a 403 back | 20:27 |
| cdent | mriedem: yeah, that part is done | 20:28 |
| cdent | already done | 20:28 |
| mriedem | well hells yeah | 20:28 |
| cdent | the thing I'm farting about with now is capturing the right exceptions | 20:28 |
| cdent | it's not as obvious as it should be because some of them are caught, some are caught and reraised, some are never caught | 20:29 |
| mriedem | yeah...it was kind of a truffle hunt, | 20:30 |
| mriedem | i'm pretty sure at least NotFound and ConcurrentUpdateDetected were clearly raised | 20:30 |
| mriedem | NotFound would be caught in the handler.py but returned as a 404 rather than a 400 which is wrong | 20:31 |
| *** tssurya has joined #openstack-placement | 20:31 | |
| cdent | yeah, I have a draft comment pending where I ack that | 20:31 |
| cdent | I'm not sure what I was thinking other than zzzzz | 20:32 |
| cdent | I do recall when first wrote the /reshaper code we were behaving like wanted it yesterday so both the PlacementDirect and Reshaper handler code were written in a big hurry, of the "we'll catch it in review" style | 20:33 |
| cdent | and then | 20:33 |
| *** tssurya has quit IRC | 20:35 | |
| mriedem | yeah | 20:42 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Make CELL_TIMEOUT a constant https://review.openstack.org/594570 | 20:57 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Stash the cell uuid on the context when targeting https://review.openstack.org/594571 | 20:57 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Make RecordWrapper record RequestContext and expose cell_uuid https://review.openstack.org/594572 | 20:57 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Clarify which context is used by do_query() https://review.openstack.org/595250 | 20:57 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Batch results per cell when doing cross-cell listing https://review.openstack.org/592698 | 20:57 |
| openstackgerrit | Dan Smith proposed openstack/nova master: List instances from all cells explicitly https://review.openstack.org/593717 | 20:57 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Make instance_list perform per-cell batching https://review.openstack.org/593131 | 20:57 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Record cell success/failure/timeout in CrossCellLister https://review.openstack.org/594265 | 20:57 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Optimize global marker re-lookup in multi_cell_list https://review.openstack.org/594577 | 20:57 |
| *** dansmith is now known as htimsnad | 21:34 | |
| openstackgerrit | Chris Dent proposed openstack/nova master: [placement] Add /reshaper handler for POST https://review.openstack.org/576927 | 21:41 |
| cdent | efried, mriedem : there we go. as you can see I didn't rebase the rest of the stack, just to there | 21:41 |
| mriedem | ack | 21:42 |
| cdent | edleafe, efried: how do we want to coordinate the work in EdLeafe/placement.git? if at all? | 21:43 |
| efried | what's the strategy for putting that into the openstack project? | 21:44 |
| efried | Get it "working" first, or no? | 21:45 |
| cdent | efried: see http://eavesdrop.openstack.org/irclogs/%23openstack-placement/%23openstack-placement.2018-08-22.log.html#t2018-08-22T14:32:39 for a plan that could work | 21:47 |
| efried | If we've got something prettyclose right now, maybe we propose it as an initial change set to seed the openstack project, then iterate on it in that change set until it's passing tox and merge that. Then propose subsequent change sets until a blank depends-on from Nova passes all its shit. | 21:48 |
| cdent | it's not close | 21:48 |
| cdent | I expect will want to do a few more runs after we find more missing files or concepts | 21:49 |
| cdent | in fact | 21:49 |
| cdent | https://etherpad.openstack.org/p/placement-extraction-file-notes | 21:50 |
| efried | ack | 21:51 |
| efried | so | 21:51 |
| efried | what do you see as the necessary work on the side repo before we propose a change set to the real repo? | 21:51 |
| efried | other than what's in that etherpad so far | 21:52 |
| *** mriedem has quit IRC | 21:55 | |
| *** mriedem1 has joined #openstack-placement | 21:56 | |
| *** mriedem1 is now known as mriedem | 21:59 | |
| efried | actually, the stuff in the etherpad doesn't seem like it should stop us from proposing. | 21:59 |
| efried | perhaps some of it stops us from merging until resolved, but proposing... | 22:00 |
| cdent | I'm not sure what you're trying to accomplish efried? From my standpoint having a repo associated with gerrit and zuul before it can runs its own tests is not good | 22:01 |
| efried | cdent: Right, saying it should do those things before it *merges* for sure. | 22:01 |
| efried | What I'm trying to accomplish is | 22:01 |
| efried | having a thing in gerrit that you could go look at (by going to old patch set) that's as raw as possible - you could draw a straight line from it to the nova commit it came out of | 22:02 |
| efried | and then having the subsequent patch sets on that change set show it being cleaned up and made ready. | 22:02 |
| cdent | that _might_ be good idea once we have a clean recipe for the clean up | 22:03 |
| efried | I guess that history doesn't show up in git, only in gerrit; but even so, seems like it would be nice to have available | 22:03 |
| cdent | we're not going to lose history | 22:04 |
| cdent | all the patches we make in ed's repo will end up in git.o.o. history | 22:04 |
| efried | rightright, but what I'm saying is the history in git is going to jump a pretty big chasm to get from the nova commit to the mergeable placement commit. | 22:04 |
| efried | Having some way to look at incremental points along that parabola would be useful. | 22:05 |
| cdent | it will only be a big chasm if we choose to either make some weird big commit, or squash all the smaller commits we have made to do the cleanup | 22:06 |
| cdent | I think we should do neither | 22:06 |
| cdent | we should make small commits to fix the repo | 22:06 |
| efried | squash the smaller commits... | 22:06 |
| efried | you mean do those commits in the github repo? And then they'll get carried into the gerrit repo? | 22:06 |
| cdent | and those small commits should show up in history | 22:06 |
| cdent | the "get it working commits", yes, I think they should show up in history | 22:07 |
| efried | okay, that's even better. | 22:07 |
| efried | I guess I should learn how to work in github. | 22:07 |
| efried | never had to do it before. | 22:07 |
| cdent | if gerrit workflow is your main exposure to any kind of git, you're going to hate it | 22:07 |
| cdent | but if you had githubbed before gerrit, vice versa | 22:07 |
| efried | orly, githubbers don't just love gerrit off the bat? | 22:08 |
| cdent | maybe tomorrow or friday we should do some kind of group screen share pair thing to make some headway | 22:08 |
| cdent | I hate both github and gerrit. I hope this is not a suprrise. I hate gerrit more. | 22:08 |
| efried | innnnteresting. What is your nirvana code workflow system? | 22:09 |
| efried | or is all workflow tools mostly rapscallions? | 22:09 |
| cdent | I haven't met it yet | 22:11 |
| cdent | If gerrit had a few million dollars spent on human factors research and repeair it has the potential to be good | 22:12 |
| cdent | but it would have to be some pretty good research into the impact it has on psychology of code ownership and the like | 22:12 |
| cdent | (and not just that it has a sucky ui) | 22:13 |
| *** ChanServ has quit IRC | 22:49 | |
| mriedem | you could use the shitty code review system that ibm research took 10 years to develop | 22:53 |
| mriedem | which they told us we had to use *after* being used to gerrit for 3 years | 22:53 |
| mriedem | it involved sit down meetings | 22:53 |
| mriedem | some forms of role playing and games | 22:53 |
| mriedem | it was great | 22:53 |
| openstackgerrit | Merged openstack/nova master: Filter out instances without a host when populating AZ https://review.openstack.org/594050 | 22:53 |
| openstackgerrit | Merged openstack/nova master: Remove ChanceScheduler https://review.openstack.org/576384 | 22:54 |
| cdent | please no | 22:56 |
| efried | I've said it before: I have never used a code+issue tracking system better than CMVC. | 23:00 |
| efried | gerrit is better for review workflow, git is better for source control; but CMVC as an integrated solution was the best. | 23:01 |
| *** ChanServ has joined #openstack-placement | 23:04 | |
| *** barjavel.freenode.net sets mode: +o ChanServ | 23:04 | |
| cdent | efried, edleafe : I'm going to have to stop messing around soon, but I've created a disordered list of things we probably need to do before more massaging of the temp repo (on the etherpad) | 23:07 |
| cdent | I'll add a bit more, but sleep is calling | 23:07 |
| edleafe | cdent: efried: sorry, was fighting puter issues | 23:10 |
| edleafe | cdent: efried: I guess the standard github PR for updates in the repo? That would reduce conflicts | 23:14 |
| cdent | edleafe: yeah, that was what I was assuming as well | 23:15 |
| edleafe | I'm signing off for the day. We can resume tomorrow | 23:17 |
| cdent | 'nigh | 23:19 |
| *** cdent_ has joined #openstack-placement | 23:41 | |
| *** cdent has quit IRC | 23:44 | |
| *** cdent_ is now known as cdent | 23:44 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!
