| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-01-26 | |||
| 17:50:28 | bauzas | good points | |
| 17:50:32 | bauzas | I can respin quickly | |
| 17:50:39 | bauzas | for the 2nd test I'm adding | |
| 17:51:10 | dansmith | I would recommend waiting until we get a run so we know if the top one is going to work, but if you have to go before it will finish, then just push | |
| 17:51:15 | bauzas | that is not running init_host, that's rather a defensive patch for making sure anyone can't just change nova.conf.placement opt and add a default without being kicked in the ass | |
| 17:51:34 | bauzas | dansmith: well, I'll get an IndentationError anyway, right? | |
| 17:51:41 | bauzas | dansmith: so I'm pretty sure it would fail anyway | |
| 17:51:52 | bauzas | don't know how I left that FWIW | |
| 17:52:05 | bauzas | probably because I said I love Atom in public | |
| 17:52:12 | dansmith | oh I see, well, might be good to assert the positive case in another test anyway | |
| 17:53:55 | bauzas | dansmith: I can put that elsewhere | |
| 17:54:08 | bauzas | dansmith: if you have ideas | |
| 17:54:20 | bauzas | dansmith: I'm just about uploading a new Ps | |
| 17:54:28 | bauzas | dansmith: unless you disagree | |
| 17:54:39 | dansmith | don't disagree | |
| 17:57:40 | bauzas | k | |
| 17:57:44 | bauzas | shooting, then | |
| 18:02:14 | johnthetubaguy | melwitt: oh, resize is a bit of a bummer in that quota logic. | |
| 18:02:20 | openstackgerrit | Sylvain Bauza proposed openstack/nova: Scheduler calling the Placement API https://review.openstack.org/417961 | |
| 18:02:21 | openstackgerrit | Sylvain Bauza proposed openstack/nova: Block starting compute unless placement conf is provided https://review.openstack.org/425806 | |
| 18:05:12 | melwitt | johnthetubaguy: yeah, it's ugly. it would look a lot better with the check_deltas idea | |
| 18:05:46 | johnthetubaguy | melwitt: I was thinking, its not until resize confim that we update the DB with the new larger flavor so its counted correctly, thats quite a big window | |
| 18:06:09 | dansmith | bauzas: did you not see my comment on the last patch about enabling service version caching? | |
| 18:06:18 | bauzas | dansmith: it seems not | |
| 18:06:23 | dansmith | bauzas: okay, I'll fix | |
| 18:06:25 | bauzas | my bad | |
| 18:06:38 | dansmith | oh, | |
| 18:06:51 | dansmith | I thought you were leaving soon, it's only 1900 your time right? | |
| 18:07:18 | bauzas | dansmith: I'm still here for a couple of hours :) | |
| 18:07:19 | dansmith | either way, let us wait for a run of the top one anyway | |
| 18:07:27 | melwitt | johnthetubaguy: ah, yeah. it's kind of like that now though, since it does a reserve right away, and quota is checked against reserved + in_use. the quota commit is in the confirm_resize but all that does it delete the reservation and increase the in_use | |
| 18:07:27 | dansmith | and if it's after you leave I'll fix it | |
| 18:07:32 | bauzas | dansmith: I just need to bail out for 60 mins | |
| 18:07:36 | dansmith | okay | |
| 18:07:44 | bauzas | dansmith: but yeah, it's 7:07pm | |
| 18:07:59 | bauzas | dansmith: so I should be back around 9pm to midnight | |
| 18:08:05 | dansmith | okay | |
| 18:08:05 | melwitt | johnthetubaguy: oh, you mean with the new way, the in_use doesn't increase until the confirm. yeah | |
| 18:08:11 | bauzas | ie. 2000UTC to 2300UTC | |
| 18:08:24 | johnthetubaguy | melwitt: but the reservations count towards quota, so its a bit different | |
| 18:08:31 | bauzas | dansmith: I can do the fix related to your -1 | |
| 18:08:55 | johnthetubaguy | melwitt: yeah, I think the count instances just needs to use the new flavor if its a pending resize (I kinda hope there is a cleaner way though) | |
| 18:08:56 | bauzas | dansmith: but if you have time for it, go | |
| 18:08:57 | dansmith | bauzas: okay but let the top patch finish a run first | |
| 18:09:08 | bauzas | dansmith: okay, so I leave you implement that | |
| 18:09:12 | melwitt | johnthetubaguy: yeah, I get what you mean now. the old way reflects the increased resource usage immediately whereas the new way defers it until the confirm | |
| 18:09:18 | bauzas | dansmith: because I'll be off for 1.5h-ish | |
| 18:09:30 | openstackgerrit | Eric Fried proposed openstack/nova: Introduce base PowerVM driver https://review.openstack.org/391288 | |
| 18:09:41 | bauzas | now the patches are in the gate, I'm bailing out but I'll be back | |
| 18:09:49 | johnthetubaguy | melwitt: yeah, sorry, thats what I mean | |
| 18:09:52 | bauzas | andreykurilin: where are we with the client side ? | |
| 18:09:53 | openstackgerrit | Eric Fried proposed openstack/nova: Introduce base PowerVM driver (disk drv) https://review.openstack.org/409401 | |
| 18:10:21 | bauzas | andreykurilin: oh, it's missing another +2/+W | |
| 18:10:52 | melwitt | johnthetubaguy: I had been thinking it's good to not reflect the usage until it's really actually used. but, that opens the window for things to get past and go over quota when confirm happens later | |
| 18:10:53 | bauzas | andreykurilin: when are you planning to deliver a new revision ? | |
| 18:11:09 | openstackgerrit | Eric Fried proposed openstack/nova: Introduce base PowerVM driver (console) https://review.openstack.org/409402 | |
| 18:11:21 | openstackgerrit | Eric Fried proposed openstack/nova: Introduce base PowerVM driver (cfg drv) https://review.openstack.org/409404 | |
| 18:11:41 | bauzas | andreykurilin: oh, you're saying all johnthetubaguy's comments can be addressed in a follow-up | |
| 18:11:42 | andreykurilin | mriedem bauzas : let's release novaclient without merging fix for token+cli. It was broken for a year, so it looks like this feature is not in demand, but we need to release new version of novaclient now to release all warning messages which were introduced in 7.0.0 + we need full keystone v3 support | |
| 18:12:29 | bauzas | andreykurilin: the token thing is fixed, that's just the CLI which is not yet fixed, right? | |
| 18:12:45 | bauzas | andreykurilin: unless I'm wrong | |
| 18:13:00 | andreykurilin | bauzas: yes. Token with via python novaclient's API is fixed | |
| 18:13:08 | andreykurilin | *Token auth | |
| 18:13:31 | bauzas | andreykurilin: could you then split your CLI change and just leave the reno file in a separate change ? | |
| 18:13:37 | melwitt | johnthetubaguy: yeah, that would not be great to have to consider pending resize. I wonder if it could be as simple as checking count and quota in the confirm too. and if you are late with your confirm, you may not get to confirm it and have to revert | |
| 18:13:45 | bauzas | andreykurilin: I'd like to include that note for the 7.1.0 release notes | |
| 18:14:00 | bauzas | andreykurilin: and I feel brave enough for a single +2/+W | |
| 18:14:07 | melwitt | but I guess that's not a good experience | |
| 18:14:10 | andreykurilin | bauzas: sorry, I'm not near PC now:( | |
| 18:14:37 | bauzas | andreykurilin: okay, lemme do that then | |
| 18:15:00 | andreykurilin | bauzas Ok, it looks like I'll able to do it in 10-15 mini | |
| 18:15:06 | andreykurilin | *minutes | |
| 18:15:11 | openstackgerrit | Dirk Mueller proposed openstack/nova: [WIP] Handle boot_index in block_device_info also for volumes https://review.openstack.org/425813 | |
| 18:15:14 | bauzas | andreykurilin: I'll need to bail out for 1.5 hours | |
| 18:15:14 | johnthetubaguy | melwitt: I did wonder about that, it just feels like you could do loads of resizes to use too much quota, and just not bother doing and confirms so you get to keep the extra big instances | |
| 18:15:26 | bauzas | andreykurilin: so it's up to you, your call | |
| 18:15:35 | melwitt | johnthetubaguy: oh, heh. yeah | |
| 18:15:36 | johnthetubaguy | melwitt: we do have auto resize confirm to protect against that, but I duno, seems funky | |
| 18:15:44 | andreykurilin | bauzas: oh.. we need to propose a release as soon as possible | |
| 18:15:46 | bauzas | andreykurilin: if you can split that, I can +2 and ask another core to quickly +W the reno change only | |
| 18:16:10 | openstackgerrit | Ed Leafe proposed openstack/nova: placement: RT now adds proper Ironic inventory https://review.openstack.org/404472 | |
| 18:16:12 | bauzas | andreykurilin: I know, but I don't want to ship a release without the proper release note | |
| 18:16:13 | johnthetubaguy | melwitt: I should probably worry that that attack is the first thing that pops into my head! | |
| 18:16:29 | edleafe | jaypipes_: ^^ should fix mriedem's concerns | |
| 18:16:29 | bauzas | and we can still tag the release in the next hours | |
| 18:16:51 | bauzas | there is no rush, the deadline can be very late PST | |
| 18:17:00 | andreykurilin | bauzas maybe johnthetubaguy can find a bit of time to review my answers to his comments? | |
| 18:17:11 | melwitt | johnthetubaguy: haha :) well, it's good you're thinking of those cases | |
| 18:17:12 | bauzas | ask him :) | |
| 18:17:17 | andreykurilin | bauzas g-r will be closed in an hour I think | |
| 18:17:25 | bauzas | what I want is at least the reno file | |
| 18:17:39 | johnthetubaguy | andreykurilin: I just don't understand that keychain stuff, and haven't had chance to read into it :( | |
| 18:17:52 | bauzas | okay, too risky then | |
| 18:17:59 | bauzas | let's just merge the reno file | |
| 18:18:03 | johnthetubaguy | andreykurilin: once you delete that code, I can't see that class being used anywhere | |
| 18:18:30 | johnthetubaguy | so we can always do a stable branch backport for this one | |
| 18:18:34 | johnthetubaguy | for the CLI users | |
| 18:18:55 | melwitt | johnthetubaguy: so on the race prevention thing, you were just thinking of checking the limit twice? once before object create and once after? | |
| 18:19:09 | johnthetubaguy | melwitt: yeah, I mean thats racey, but in a quota goes low sense | |
| 18:19:19 | melwitt | I guess you'd have to redo the count too | |