| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-06-09 | |||
| 09:34:31 | openstackgerrit | Zhenyu Zheng proposed openstack/nova master: Add tags fields for server list detail/show api-ref https://review.openstack.org/465842 | |
| 10:01:56 | openstackgerrit | Merged openstack/nova master: Fix lookup of instance mapping in metadata set-password https://review.openstack.org/472405 | |
| 10:02:45 | openstackgerrit | Merged openstack/nova master: Log a warning if there is only one cell when listing instances https://review.openstack.org/440652 | |
| 10:10:03 | openstackgerrit | John Garbutt proposed openstack/nova master: Update policy descriptions for base. https://review.openstack.org/472622 | |
| 10:24:32 | openstackgerrit | Merged openstack/nova master: Calculate stopped instance's disk sizes for disk_available_least https://review.openstack.org/468269 | |
| 10:40:03 | openstackgerrit | John Garbutt proposed openstack/nova master: Update policy descriptions for base. https://review.openstack.org/472622 | |
| 10:41:31 | openstackgerrit | falseuser proposed openstack/nova master: Clear instance root_device_name in rebuild operation https://review.openstack.org/464991 | |
| 10:43:44 | openstackgerrit | sahid proposed openstack/nova master: libvirt: fix memory locking limit https://review.openstack.org/472633 | |
| 10:55:33 | openstackgerrit | John Garbutt proposed openstack/nova master: Update policy descriptions for base. https://review.openstack.org/472622 | |
| 10:55:34 | openstackgerrit | John Garbutt proposed openstack/nova master: WIP: Use oslo.polcy DocumentedRuleDefault https://review.openstack.org/472634 | |
| 11:19:35 | openstackgerrit | John Garbutt proposed openstack/nova master: Update policy descriptions for base. https://review.openstack.org/472622 | |
| 11:20:34 | openstackgerrit | John Garbutt proposed openstack/nova master: WIP: Use oslo.polcy DocumentedRuleDefault https://review.openstack.org/472634 | |
| 11:29:32 | cdent | bauwser: I changed that bug back to triaged. I agree with you that the solution they've used may not be optimal, but there _is_ a bug | |
| 11:29:52 | cdent | they want to be able to express >1 location of config file, which is a normal thing for something that uses oslo_config | |
| 11:34:44 | sdague | sfinucan: on https://review.openstack.org/#/c/472266 - I think you are right, we can't remove that bit in Pike, it's going to need a db migration | |
| 11:35:24 | sdague | however, this is bad enough that I think we want to get this in and backported asap, as the failure mode is "my guests can't start after upgrade" | |
| 11:35:31 | sdague | which is about as bad a failure mode as we have | |
| 12:32:46 | kashyap | artom: I saw the ping from here, but didn't completely read-through the scrollback | |
| 12:32:49 | kashyap | Was there a question for me? | |
| 12:59:43 | artom | kashyap, err, I think it was last night when we were talking about transient vs. persistent domains | |
| 12:59:55 | artom | Neither of us was 100% sure of what each of those implies | |
| 13:00:03 | kashyap | artom: Yeah, sorry should've mentioned it was last night | |
| 13:00:17 | artom | So we thought of you :) | |
| 13:01:00 | kashyap | For a transient domain, the guest does not stay up across reboots / shutdown | |
| 13:01:17 | kashyap | I.e. you'll lose the guest XML definition | |
| 13:01:23 | kashyap | You might wonder where it would be useful | |
| 13:01:53 | kashyap | artom: An example that comes to mind is the blockCopy() API | |
| 13:02:13 | kashyap | Where you have to temporarily make a guest transient, for it to complete | |
| 13:02:30 | kashyap | Because -- a live block copy job cannot last across reboots | |
| 13:02:35 | kashyap | A simple test -- https://kashyapc.fedorapeople.org/virt/temp/live-blockcopy-quick-test.txt | |
| 13:03:54 | bauwser | sdague: should we maybe stop to ask users to provide a bug report when they have a HTTP500 ? | |
| 13:03:59 | bauwser | for example, see https://bugs.launchpad.net/nova/+bug/1696972 | |
| 13:03:59 | openstack | Launchpad bug 1696972 in OpenStack Compute (nova) "host type 'm1.nano' not found when lanch instance" [Undecided,New] | |
| 13:04:00 | kashyap | artom: Also, from Nova code, in libvirt/driver.py (in _swap_volume(), for blockRebase() which is the same API: | |
| 13:04:03 | kashyap | try: | |
| 13:04:05 | kashyap | # NOTE (rmk): blockRebase cannot be executed on persistent | |
| 13:04:08 | kashyap | # domains, so we need to temporarily undefine it. | |
| 13:04:11 | kashyap | # If any part of this block fails, the domain is | |
| 13:04:13 | kashyap | # re-defined regardless. | |
| 13:04:16 | kashyap | if guest.has_persistent_configuration(): | |
| 13:04:18 | kashyap | support_uefi = self._has_uefi_support() | |
| 13:04:19 | bauwser | sdague: we have lots of bug reports that are possibly a configuration problem | |
| 13:04:21 | kashyap | guest.delete_configuration(support_uefi) | |
| 13:04:50 | sdague | bauwser: sure | |
| 13:05:21 | sdague | bauwser: I'd be fine dropping that message in the 500 payload | |
| 13:05:44 | bauwser | okay | |
| 13:06:03 | sdague | also, someone with traslated logs! | |
| 13:06:10 | sdague | that's the first time I've seen that in practice | |
| 13:06:33 | artom | kashyap, so would I be correct in understanding that transient domains came about more as a technical necessity of long running operations that cannot last across reboots rather than as a user-facing feature? | |
| 13:07:04 | sdague | bauwser: that being said, it would be nice to actually catch these misconfiguration items and be more explicit about them | |
| 13:07:19 | bauwser | sdague: agreed, I'll add a better exception | |
| 13:07:22 | bauwser | for that | |
| 13:07:30 | sdague | bauwser: ok, cool | |
| 13:07:31 | kashyap | artom: Almost correct. It's also useful for "user-facing" feature in the sense that -- if you're testing something w/ libvirt, and don't want to clean up the guests your self, you can use transient domains | |
| 13:07:34 | artom | So what happens if the host dies while such an operation is happening? It aborts, obviously, the transient domain is lost, but the persistent domain, the one without the operation going on, is saved, and comes back after the host recovers? | |
| 13:08:02 | sdague | bauwser: while you are looking at things - https://review.openstack.org/#/c/472266/ really feels like we should get it in soon | |
| 13:08:12 | sdague | that's the fail to start instances after upgrade bug | |
| 13:08:19 | bauwser | sdague: roger. | |
| 13:10:57 | sfinucan | sdague: That sounds good to me. +Wd now | |
| 13:10:59 | openstackgerrit | falseuser proposed openstack/nova master: Clear instance root_device_name in rebuild operation https://review.openstack.org/464991 | |
| 13:11:34 | bauwser | oh man, +W'd duplicated :) | |
| 13:11:43 | bauwser | shit. | |
| 13:40:58 | sdatko | hello | |
| 13:42:01 | sdatko | tiny question: nova.db.sqlalchemy.models.InstanceTypeExtraSpecs vs nova.db.sqlalchemy.api_models.FlavorExtraSpecs | |
| 13:43:31 | openstackgerrit | Sylvain Bauza proposed openstack/nova master: Provide a better log when api_servers is unset https://review.openstack.org/472685 | |
| 13:43:33 | sdatko | for first one there is note about being deprecated; however, when performing context.session.query(models.InstanceTypeExtraSpecs) i can see results, while for context.session.query(api_models.FlavorExtraSpecs) i receive empty array | |
| 13:43:44 | bauwser | sdague: see https://review.openstack.org/472685 for the api_servers opt | |
| 13:44:21 | mriedem | sdatko: the former is in the cell database (nova db), the latter is in the api db (nova_api db), flavors were moved from the cell to the api db a few releases ago | |
| 13:44:24 | mriedem | mitaka i think | |
| 13:44:49 | sdatko | hmm | |
| 13:45:07 | sdatko | ok, to be honest, i just played and created some tests | |
| 13:45:30 | sdatko | in those tests, i used nova.db.sqlalchemy.api.flavor_create() | |
| 13:45:50 | sdatko | this one is outdated? | |
| 13:46:22 | mriedem | you shouldn't be playing around with the db apis directly | |
| 13:46:49 | mriedem | everything goes through nova.objects.flavor.Flavor | |
| 13:47:58 | sdatko | ouuu | |
| 13:48:28 | sdatko | my fault then | |
| 13:48:33 | sdague | bauwser: I would honestly throw an exception there as well | |
| 13:48:44 | sdatko | thanks mriedem | |
| 13:48:45 | sdague | the rest of the code is going to fail with a different exception later | |
| 13:49:04 | bauwser | sdague: fine, I was just thinking of leaving the glanceclient providing its own exception | |
| 13:49:14 | mriedem | sdague: bauwser: here is the ocata backport for that rbd thing https://review.openstack.org/#/c/472687/ | |
| 13:49:20 | mriedem | sdatko: yw | |
| 13:49:43 | bauwser | mriedem: hold my beer | |
| 13:50:04 | sdague | mriedem: probably should trigger a release off of that on stable as well | |
| 13:50:53 | bauwser | mriedem: in other words, ping me when master is merged | |
| 13:51:07 | mriedem | sdague: yeah we have quite a few things to get into stable/ocata for a release | |
| 13:51:25 | bauwser | we could release mid-week | |
| 13:54:24 | sdague | mriedem: meaning there are a bunch of other patches you want in/ | |
| 13:54:25 | sdague | ? | |
| 13:54:31 | sdague | or a bunch that are already? | |
| 13:55:27 | mriedem | sdague: the former | |
| 13:56:00 | mriedem | alex_xu: cdent: on the traits bp, as far as using traits, isn't jay working on some changes in the scheduler to use the misc shared aggregates trait for shared resources (storage)? | |
| 13:56:01 | sdague | mriedem: if you list them, I'll take a look | |
| 13:56:07 | sdague | or just "all of them" | |
| 13:56:18 | cdent | mriedem: yes, but that's just for that one trait | |
| 13:56:19 | mriedem | alex_xu: cdent: i think that's https://blueprints.launchpad.net/nova/+spec/shared-resources-pike | |
| 13:56:26 | cdent | not traits in general | |
| 13:56:39 | mriedem | cdent: sure but i don't think we said we were going to support users requesting traits during server create in pike | |
| 13:56:42 | mriedem | that's a queens thing | |