| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-08-08 | |||
| 18:42:44 | mriedem | but maybe for now, | |
| 18:42:58 | mriedem | we go cheap and easy and throw a bullet in https://docs.openstack.org/nova/latest/user/placement.html#pike-16-0-0 and then reference that from the prelude for 'more info' type stuff for now | |
| 18:43:37 | mriedem | i.e. during scheduling, we get the allocation candidates from placement, and then use those to get the compute nodes from the cells, and iterate the results using the enabled filters | |
| 18:44:14 | mriedem | then we iterate the hosts and make allocation requests for the instance against a given host, retrying as necessary until an allocation is made or all allocations are exhausted, which results in NoValidHost | |
| 18:44:31 | mriedem | for a move operation, allocations are made on the source and dest hosts, | |
| 18:44:39 | mriedem | for a resize to the same host, allocations are summed on the same host | |
| 18:44:49 | mriedem | ^ is that sufficient for a note in https://docs.openstack.org/nova/latest/user/placement.html#pike-16-0-0 ? | |
| 18:44:50 | melwitt | we can just copy-paste this into the doc :) | |
| 18:45:33 | mriedem | i'll just throw something up using that and we can take a look | |
| 18:46:06 | melwitt | ++ | |
| 18:49:52 | jaypipes | dansmith: pls see latest coment on https://review.openstack.org/#/c/491850/1/nova/compute/resource_tracker.py | |
| 18:50:01 | jaypipes | dansmith: line 1060 | |
| 18:50:06 | jaypipes | dansmith: ? for you there. | |
| 18:50:59 | melwitt | guh, looks like it's a PITA to change the resources SmallFakeDriver has | |
| 18:52:24 | mriedem | melwitt: i did something like this, sec | |
| 18:52:36 | mriedem | melwitt: one thing is just using FakeDriver | |
| 18:52:39 | mriedem | it has more resources | |
| 18:52:42 | dansmith | jaypipes: well, you are putting a continue in there that wasn't there before, but also, I think it's actually dead code | |
| 18:52:46 | mriedem | the ServerMovingTests use that | |
| 18:52:48 | mriedem | for the same reason | |
| 18:52:59 | dansmith | jaypipes: L1028 clears the tracked instances, then you roll through and exclude things that are in that set, but it's empty right? | |
| 18:53:08 | melwitt | mriedem: I found some examples but they involve stubbing out the compute driver load | |
| 18:53:47 | jaypipes | dansmith: but the call to _update_usage_from_instance() on line 1042 adds instances back into tracked_instances :) | |
| 18:53:58 | jaypipes | dansmith: perfectly. clear. | |
| 18:54:22 | dansmith | jaypipes: ah right, well then you are changing it | |
| 18:54:37 | jaypipes | dansmith: I know, it's awfulness. | |
| 18:55:36 | mriedem | melwitt: https://github.com/openstack/nova/blob/master/nova/tests/functional/test_servers.py#L1074-L1077 | |
| 18:56:02 | melwitt | mriedem: yesss thank you | |
| 18:56:33 | dansmith | jaypipes: so we only have things in tracked_instances that are not offloaded or deleted, and your assertion is that if we processed them in that thing, added to tracked, and then found an allocation we can ignore entirely | |
| 18:57:07 | jaypipes | dansmith: yup. those represent the "normals" | |
| 18:57:17 | jaypipes | dansmith: and we don't need to remove any allocations for them. | |
| 18:57:25 | jaypipes | dansmith: they're active, stopped, paused, etc | |
| 18:57:43 | jaypipes | dansmith: still consuming resources on the node and that's acceptable. | |
| 18:58:03 | dansmith | jaypipes: yeah okay re-reading my scenario, I see that update would have excluded anything we in those states _and_ things we don't have running on our host I guess, the latter being the critical point | |
| 18:58:17 | jaypipes | right. I was just adding some log statements in there... | |
| 18:59:16 | dansmith | I'm not sure it's less confusing the way you have it written, but don't change it now | |
| 18:59:29 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add track_instance_changes note in disable_group_policy_check_upcall https://review.openstack.org/490627 | |
| 18:59:30 | openstackgerrit | Matt Riedemann proposed openstack/nova master: doc: considerations before deploying multiple cells https://review.openstack.org/491885 | |
| 19:00:50 | cdent | melwitt: you might take a look at gibi’s tests where he’s messing with resize to same host. that’s where a lot of the fiddling and discussion about make tests happy has happened: around ps4 on https://review.openstack.org/#/c/491529/ | |
| 19:00:51 | mriedem | dansmith: posed a question in my own patch ^ for https://review.openstack.org/#/c/491885/ to how we'd best like to communicate the online data migrations not being multi-cell aware | |
| 19:01:26 | melwitt | thanks cdent | |
| 19:02:04 | dansmith | mriedem: I was thinking about this after you brought it up, but people can't have multiple cells worth of things needing migration, so I'm not sure we need to even say anything, right? | |
| 19:02:35 | mriedem | dansmith: because a new cell would be empty and we couldn't fallback to it anyway? | |
| 19:02:38 | mriedem | and new things would be in the api db | |
| 19:02:44 | dansmith | yeah | |
| 19:03:09 | mriedem | not for pike i suppose | |
| 19:03:30 | mriedem | the mention in the code review guide is probably legit though right? https://review.openstack.org/#/c/491885/1/doc/source/contributor/code-review.rst | |
| 19:04:21 | dansmith | mriedem: well, any data migration at all needs to consider multiple cells.. I guess that means we should make the runner thing arrange to run it against all cells | |
| 19:04:22 | dansmith | mriedem: but yeah | |
| 19:04:24 | mriedem | like say you deploy multiple cells in pike, start populating them, and then in queens we have some data migration to move something to the api db | |
| 19:04:38 | dansmith | mriedem: hopefully we're done with that pattern, but yeah | |
| 19:04:54 | mriedem | but i still have to move the bandwidth_usage table to the api db! | |
| 19:05:13 | dansmith | mriedem: redirect to trash can | |
| 19:05:47 | mriedem | ok i'll drop the thing in the layout page | |
| 19:05:47 | openstackgerrit | Merged openstack/nova master: Create reference subpage https://review.openstack.org/490994 | |
| 19:06:38 | openstackgerrit | Merged openstack/nova master: doc: Add additional content to admin guide https://review.openstack.org/490952 | |
| 19:08:24 | openstackgerrit | Matt Riedemann proposed openstack/nova master: doc: code review considerations for online data migrations https://review.openstack.org/491885 | |
| 19:09:24 | openstackgerrit | Jay Pipes proposed openstack/nova master: placement: refactor healing of allocations in RT https://review.openstack.org/491850 | |
| 19:09:24 | openstackgerrit | Jay Pipes proposed openstack/nova master: remove provider allocs in confirm/revert resize https://review.openstack.org/488510 | |
| 19:09:25 | openstackgerrit | Jay Pipes proposed openstack/nova master: Resource tracker compatibility with Ocata and Pike https://review.openstack.org/491012 | |
| 19:09:33 | jaypipes | dansmith, mriedem: k. ^^ | |
| 19:10:26 | toabctl | oomichi, hey. could you have a look at https://review.openstack.org/#/c/398308/ please? | |
| 19:16:47 | openstackgerrit | Sean Dague proposed openstack/nova master: Create For End Users index section https://review.openstack.org/491785 | |
| 19:20:37 | openstackgerrit | melanie witt proposed openstack/nova master: Request zero root disk for boot-from-volume instances https://review.openstack.org/428481 | |
| 19:20:37 | openstackgerrit | melanie witt proposed openstack/nova master: Claim and report zero root disk for boot-from-volume instances https://review.openstack.org/428505 | |
| 19:22:06 | openstackgerrit | Sean Dague proposed openstack/nova master: Create For End Users index section https://review.openstack.org/491785 | |
| 19:22:16 | openstackgerrit | Sean Dague proposed openstack/nova master: rework index intro to describe nova https://review.openstack.org/491834 | |
| 19:22:16 | openstackgerrit | Sean Dague proposed openstack/nova master: Create For End Users index section https://review.openstack.org/491785 | |
| 19:22:17 | openstackgerrit | Sean Dague proposed openstack/nova master: Add For Operators section to front page https://review.openstack.org/491815 | |
| 19:22:50 | sdague | well that was an interesting gerrit edge case | |
| 19:26:43 | melwitt | what happened | |
| 19:27:41 | openstackgerrit | Jackie Truong proposed openstack/nova master: Add trusted_certs to Instance object https://review.openstack.org/489408 | |
| 19:32:20 | melwitt | mriedem: it looks like this is why we shouldn't need to call update_available_resource directly https://github.com/openstack/nova/blob/master/nova/compute/manager.py#L1138-L1144 | |
| 19:33:31 | mriedem | melwitt: but you didn't stop the compute service, you just disabled it | |
| 19:33:35 | mriedem | and then re-enabled it | |
| 19:33:40 | mriedem | honestly i'm not even sure how that test is passing... | |
| 19:34:08 | melwitt | oh, right | |
| 19:34:16 | melwitt | it's magic! | |
| 19:34:33 | mriedem | i probably led you down a dark path | |
| 19:35:18 | melwitt | the path of dark magic | |
| 19:35:34 | mriedem | i'm a 10th level drow elf on the weekends | |
| 19:35:41 | melwitt | lol | |
| 19:37:10 | openstackgerrit | Eric Fried proposed openstack/nova master: Get auth from context for glance endpoint https://review.openstack.org/490057 | |
| 19:37:40 | melwitt | I'm gonna check what happens with this test if I don't re-enable the service. if it doesn't fail then, something has changed | |
| 19:40:00 | melwitt | huh. I think something is afoot here | |
| 19:41:09 | melwitt | disabling the service must not be doing what I think it is | |
| 19:41:31 | dansmith | it does very little.. what do you think it does | |
| 19:41:31 | dansmith | ? | |
| 19:41:53 | melwitt | well, mriedem suggesting doing it to force a local delete. to make the "service.is_up()" return False | |
| 19:41:59 | melwitt | but it seems to be still considered up | |
| 19:42:09 | dansmith | that won't do it | |
| 19:42:15 | melwitt | *suggested. why do I keep using the wrong tense of every word | |
| 19:42:29 | melwitt | well hells bells | |
| 19:42:31 | dansmith | disabled is only checked by a scheduler filter to exclude hosts | |
| 19:42:56 | melwitt | is there a better way other than setting the update interval really low and sleeping? that's what I was doing before and that kinda sucks | |
| 19:43:14 | dansmith | yeah, that's not reasonable, IMHO | |
| 19:43:27 | mriedem | melwitt: i said force-down | |
| 19:43:27 | dansmith | stop the service and tweak the updated_at on it | |
| 19:43:28 | mriedem | not disable | |
| 19:43:43 | mriedem | force-down is what the api will check for is_up | |