| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-05-04 | |||
| 15:55:10 | edleafe | dansmith: I know he was on holiday earlier this week | |
| 15:55:16 | dansmith | edleafe: yeah | |
| 15:55:28 | edleafe | bauzas: still around? | |
| 15:56:15 | mriedem | he's picking up his daugher from school i think | |
| 15:56:20 | edleafe | dansmith: I mean, I'd be happy to work on this, but didn't want to step on bauzas's toes | |
| 15:56:32 | edleafe | mriedem: what, are they walking? :) | |
| 15:56:52 | dansmith | edleafe: yeah, it's been days though so I figured he's just busy, but cool to check | |
| 15:56:54 | mriedem | edleafe: probably an old fashioned bike would be my guess | |
| 15:57:03 | mriedem | with cigs and croissants | |
| 15:57:13 | dansmith | but if you're willing I would think some of it could be shared at least | |
| 15:57:43 | edleafe | dansmith: that's fine with me | |
| 15:58:15 | edleafe | dansmith: I'll start on it after the api-wg meeting, unless bauzas speaks up first | |
| 15:58:38 | nammo | jaypipes: RetryFilter | |
| 15:58:43 | dansmith | edleafe: cool, thanks | |
| 15:59:01 | jaypipes | nammo: that cannot be removed, AFAIK. | |
| 15:59:21 | edleafe | nammo: why would you want to remove that? | |
| 15:59:31 | melwitt | it used to be able to be removed | |
| 15:59:55 | nammo | jaypipes: edleafe Any consecutive request to boot are failing due to RetryFilter | |
| 16:00:52 | edleafe | nammo: RetryFilter only comes into play when a request to build on a host has already failed | |
| 16:01:14 | jaypipes | melwitt: really? TIL... | |
| 16:01:24 | nammo | edleafe: I deleted that VM and tried to boot again | |
| 16:01:56 | melwitt | we had removed it because of the integration with other systems we had done, vast majority of failures were deterministic and retrying a different compute wasn't going to help and would waste time, so we didn't have RetryFilter in our list and set scheduler_max_attempts = 1 | |
| 16:02:47 | nammo | melwitt: Is it intended for me? | |
| 16:03:49 | melwitt | nammo: no. just trivia on one reason someone would remove it | |
| 16:04:09 | dansmith | yay for more reasons to not do scheduler retries | |
| 16:04:40 | nammo | melwitt: scheduler_max_attempts This should be under default section? | |
| 16:05:35 | mtreinish | mriedem: oh I thought the deprecation timer started long enough ago. Well I'll just wip it to block the merge until after the release | |
| 16:06:25 | melwitt | nammo: these days it's [scheduler]/max_attempts. but the fact that you see "Previously tried hosts" in your log still means the RetryFilter is still there, which doesn't make sense if you removed it from enabled_filters and restarted n-sch | |
| 16:07:08 | nammo | melwitt: So is there any other solution to launch VM on the same host? | |
| 16:07:30 | nammo | melwitt: host is up and available | |
| 16:08:34 | melwitt | nammo: what are you trying to do? the RetryFilter just retries on a different host if the selected host fails. you're trying to get multiple VMs to land on the same host? | |
| 16:09:26 | nammo | melwitt: I just have one node that satifies my VM spec | |
| 16:10:03 | nammo | melwitt: I will see the sched logs to see why RetryFilter failed | |
| 16:10:10 | melwitt | that's usually what you want with a vanilla nova deployment, is to have retries. because if the host the scheduler selected fails to spawn the instance, it's usually best to try a different host next | |
| 16:10:16 | mriedem | mtreinish: it's going to be sitting quite awhile if you do that, butok | |
| 16:10:30 | nammo | melwitt: Agree | |
| 16:10:48 | nammo | But for me it just one host that my VM can launch on | |
| 16:11:10 | melwitt | nammo: I see. you should check the sched logs to determine why the first attempt failed. usually means not enough resources or extra_specs didn't match or something like that. if you turn on debug log level for n-sch you will see all the detail of which filters it tried and which one failed | |
| 16:21:52 | nammo | melwitt: http://paste.openstack.org/show/608889/ | |
| 16:25:26 | nammo | melwitt: I see no error here | |
| 16:25:37 | nammo | melwitt: Can you please help? | |
| 16:26:10 | openstackgerrit | Dan Smith proposed openstack/nova master: WIP: Debug race in delete-while-booting test https://review.openstack.org/461513 | |
| 16:26:44 | dansmith | nammo: please see topic, this channel is for development | |
| 16:27:41 | openstackgerrit | Dan Smith proposed openstack/nova master: Move to proper target_cell calling convention https://review.openstack.org/458634 | |
| 16:27:41 | openstackgerrit | Dan Smith proposed openstack/nova master: WIP: De-complicate some of the instance delete path https://review.openstack.org/458636 | |
| 16:27:42 | openstackgerrit | Dan Smith proposed openstack/nova master: DNM: Canary test against multi-cell devstack https://review.openstack.org/458537 | |
| 16:27:42 | openstackgerrit | Dan Smith proposed openstack/nova master: Make target_cell() yield a new context https://review.openstack.org/458635 | |
| 16:28:25 | nammo | dansmith: I see. Thanks | |
| 16:28:44 | dansmith | jaypipes: is the reason that first test worked because of the trait? | |
| 16:29:20 | melwitt | nammo: to close the loop on this, agree there's no problem in n-sch and you need to check your n-cpu logs for why it failed to spawn the instance on the selected host | |
| 16:29:34 | mriedem | dansmith: your access has been...granted | |
| 16:29:38 | dansmith | mriedem: thanks | |
| 16:30:13 | nammo | melwitt: Thanks :) | |
| 16:42:22 | mriedem | bauzas: jaypipes: edleafe: you guys will likely want to see this https://review.openstack.org/#/c/394857/ since it has to do with overhead calculation and overrides in the RT and virt drivers | |
| 16:43:25 | edleafe | mriedem: just reading the commit message on that seems like it's a misunderstanding | |
| 16:44:35 | mriedem | i'm -1ing now | |
| 16:45:49 | mriedem | edleafe: done - you can proofread it | |
| 16:45:50 | mriedem | :) | |
| 16:48:05 | mriedem | bauzas: replied in https://review.openstack.org/#/c/454899/ - that's how the online data migrations were done for adding uuid to aggregates and compute_nodes tables, | |
| 16:48:09 | mriedem | not sure that's worth a -1 | |
| 16:48:16 | mriedem | but maybe others have strong opinions | |
| 16:49:37 | dansmith | mriedem: I kinda like them in the object file, but I don't think you should change it just for that | |
| 16:50:10 | edleafe | mriedem: added my own -1 and explanation | |
| 16:52:33 | jaypipes | dansmith: yes. first test passed because of the trait. | |
| 16:52:47 | dansmith | jaypipes: yeah, ack | |
| 16:58:04 | openstackgerrit | Jackie Truong proposed openstack/nova master: Encode encryption key passed to dmcrypt https://review.openstack.org/462674 | |
| 17:00:53 | openstackgerrit | Jay Pipes proposed openstack/nova master: placement: tests for non-shared with shared https://review.openstack.org/461899 | |
| 17:00:54 | openstackgerrit | Jay Pipes proposed openstack/nova master: placement: test for agg association not sharing https://review.openstack.org/461901 | |
| 17:00:55 | openstackgerrit | Jay Pipes proposed openstack/nova master: placement: shared resources when finding providers https://review.openstack.org/461574 | |
| 17:00:55 | openstackgerrit | Jay Pipes proposed openstack/nova master: placement: test non-shared out of inventory https://review.openstack.org/461900 | |
| 17:00:56 | openstackgerrit | Jay Pipes proposed openstack/nova master: placement: get providers sharing capacity https://review.openstack.org/460798 | |
| 17:01:16 | jaypipes | dansmith, edleafe, cdent: ok, I believe I've responded or fixed all your review comments on above series. | |
| 17:01:30 | jaypipes | other than the moving of the long SQL code comment somewhere... | |
| 17:01:37 | dansmith | jaypipes: gdi jay, you pushed too fast, I wanted to claim that review credit for my commit message nit :P | |
| 17:01:39 | jaypipes | we can bikeshed about where to move that later, ok? | |
| 17:02:00 | jaypipes | heh, sorry dansmith | |
| 17:02:09 | dansmith | *shakes fist* | |
| 17:02:18 | dansmith | mriedem is calling me out BY NAME in public so I have to up my game | |
| 17:02:53 | edleafe | jaypipes: for the record, I like the comment within the method. I just don't like a module-level method in the middle of the object class definitions | |
| 17:03:37 | jaypipes | edleafe: noted. :) | |
| 17:03:50 | cdent | blue | |
| 17:03:57 | dansmith | heh | |
| 17:13:07 | dansmith | jaypipes: in SQLandia, if I to a SELECT * WHERE longstring="uuidthing", | |
| 17:13:21 | dansmith | is that much more efficient than doing an IN check for a list of one? | |
| 17:13:23 | openstackgerrit | Ed Leafe proposed openstack/nova master: Add a convenience attribute for reportclient https://review.openstack.org/462679 | |
| 17:13:23 | openstackgerrit | Ed Leafe proposed openstack/nova master: WIP: Don't delete allocation if instance being scheduled https://review.openstack.org/460177 | |
| 17:13:28 | edleafe | dansmith: ^^ | |
| 17:13:35 | dansmith | meaning SELECT * WHERE longstring IN ("uuidthing") or whatever the syntax is | |
| 17:13:43 | dansmith | edleafe: woot | |
| 17:13:48 | edleafe | dansmith: just addressed the comments. Will work on the rest now | |
| 17:13:50 | mriedem | on an indexed column | |
| 17:13:59 | dansmith | mriedem: ah, yeah, good point the index | |
| 17:14:13 | mriedem | dansmith: well, after https://review.openstack.org/#/c/454887/ that is | |
| 17:14:29 | jaypipes | dansmith: all optimizers I know of will rewrite the latter to the former. | |
| 17:14:44 | dansmith | jaypipes: ah, okay | |
| 17:14:57 | dansmith | actually, I think I remember seeing warnings about doing that | |
| 17:18:31 | mriedem | dansmith: if you see the patch after it, i could have ComputeNode.get_by_uuid call ComputeNodeList.get_all_by_uuids and internally handle the [uuid] and [0], but i just wondered about the IN performance | |
| 17:19:06 | dansmith | mriedem: yeah | |
| 17:20:08 | mriedem | but maybe this is a small enough difference that it doesn't really matter and we could just squash those two changes, and just hide it all in the object code | |
| 17:20:46 | dansmith | yeah, seems like it might be | |