| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-03-31 | |||
| 14:13:11 | sdague | at the top of it | |
| 14:13:49 | mordred | cool | |
| 14:15:18 | sdague | mordred: if you had to pick one action to be right, which one would you want? I'll write on api-ref patch today on it | |
| 14:15:22 | ZZelle_ | Hi everyone | |
| 14:16:35 | mriedem | mordred: HARD and SOFT are options https://github.com/openstack/nova/blob/master/nova/api/openstack/compute/schemas/servers.py#L239 | |
| 14:16:51 | ZZelle_ | It seems there is two different ways to add a new interface to a VM: using the action AddFixedIp and using os-interface resource ... which one should we prefer? | |
| 14:16:56 | mordred | sdague: neither is super urgent - but mriedem just answered reboot for me, so I'll vote for resize :) | |
| 14:17:09 | sdague | ok | |
| 14:17:33 | mriedem | here is the resize request schema https://github.com/openstack/nova/blob/master/nova/api/openstack/compute/schemas/servers.py#L192 | |
| 14:17:45 | mriedem | flavorRef is the only thing that's required | |
| 14:18:01 | mriedem | i believe the diskConfig stuff is all xen | |
| 14:18:05 | mriedem | all xen, all the time | |
| 14:18:10 | mordred | do we expect reboot will ever get a type other than hard or soft? | |
| 14:18:15 | mordred | yay xen! | |
| 14:18:17 | mriedem | ZZelle_: we're going to deprecate addFixedIP | |
| 14:18:30 | mriedem | mordred: such as?/ | |
| 14:18:55 | ZZelle_ | mriedem, ok so i will use os-interface instead of AddFixedIp in OSC | |
| 14:19:06 | mordred | mriedem: dunno. I was thinking of making the python paramter "hard=True" rather than type="HARD"|"SOFT" | |
| 14:19:21 | mordred | mriedem: but making it a bool flag would get weird if in the future we made a "SQUISHY" or something | |
| 14:20:47 | mriedem | limp? | |
| 14:21:08 | mordred | yah. please do a limp reboot on that server | |
| 14:21:11 | mriedem | limp == reboot the guest, but make sure my network and volumes are no longer connected | |
| 14:21:15 | mriedem | because i like to be dumb | |
| 14:21:18 | mordred | hahahaha | |
| 14:21:27 | mordred | that seems like a good patch for tomorrow | |
| 14:22:35 | mriedem | ZZelle_: fyi https://review.openstack.org/#/c/384261/ | |
| 14:22:53 | ZZelle_ | mriedem, thx | |
| 14:23:12 | mriedem | ZZelle_: also, this is old https://review.openstack.org/#/c/350277/ but i'm not totally sure yet what the long-term plan for os-interface is | |
| 14:23:27 | mriedem | ZZelle_: but for now i think using that for adding an interface is what you want | |
| 14:25:12 | mriedem | that was proposing to deprecate the GET methods, but i think we said we didn't want that, and we should deprecate os-virtual-interfaces, and make os-interfaces work like os-volume_attachments | |
| 14:25:20 | mriedem | but that's probably all queens work at this point | |
| 14:25:20 | ZZelle_ | mriedem, if we deprecate os-interface how do we attach a new interface to a VM :s? | |
| 14:25:29 | ZZelle_ | mriedem, ok | |
| 14:25:32 | mriedem | the spec was for deprecating the GET methods, | |
| 14:25:35 | mriedem | not POST or DELETE | |
| 14:25:41 | ZZelle_ | mriedem, ok | |
| 14:25:46 | mriedem | it's a bit confusing, sorry about that | |
| 14:26:18 | mriedem | anyway, https://developer.openstack.org/api-ref/compute/#port-interfaces-servers-os-interface is what you want | |
| 14:44:55 | mriedem | lyarwood: if you haven't started on it yet, i was going to push the changes to cleanup my nits in the detach refactor patch | |
| 14:46:12 | lyarwood | mriedem: if you already have a commit feel free to push it | |
| 14:50:20 | mdbooth | kaisers: I looked at --scope. I didn't actually test it, but from reading the docs it was my understanding that a --scope group will be terminated automatically with its parent | |
| 14:50:34 | mdbooth | kaisers: I wouldn't bet my house on that, though. | |
| 14:54:50 | dims | dulek : posted a day early? https://review.openstack.org/#/c/452219/ | |
| 14:55:00 | dims | :) | |
| 14:57:36 | cdent | oh no, not that time of year :( | |
| 14:57:50 | lyarwood | melwitt: btw https://review.openstack.org/#/c/451870/ is another trivial docs change if you have time today | |
| 14:58:16 | mriedem | heh | |
| 14:58:19 | mriedem | tweetmq? | |
| 14:58:28 | mriedem | isn't there a forum session about dropping all other MQ backends except rabbit? | |
| 14:58:43 | mriedem | http://forumtopics.openstack.org/cfp/details/8 | |
| 14:58:51 | melwitt | lyarwood: looking | |
| 15:00:15 | melwitt | lol "I would love to see an USPS as a possibility here" | |
| 15:00:57 | mdbooth | kaisers: Hmm, kinda find the reason I thought that now. Could be I was imagining it. Have you tested this manually, btw? | |
| 15:01:13 | mdbooth | kinda? can't | |
| 15:01:57 | openstackgerrit | Sujitha proposed openstack/nova master: Adding tags field to InstancePayload https://review.openstack.org/407228 | |
| 15:02:18 | leakypipes | sfinucan: quick review on this please? https://review.openstack.org/#/c/451635/2 | |
| 15:10:05 | openstackgerrit | Sujitha proposed openstack/nova master: Change tags to default field in Instance object. https://review.openstack.org/415298 | |
| 15:10:18 | openstackgerrit | Sujitha proposed openstack/nova master: Adding auto_disk_config field to InstancePayload https://review.openstack.org/419185 | |
| 15:10:32 | openstackgerrit | Sujitha proposed openstack/nova master: Reduce calls to load_tags() to 0 https://review.openstack.org/435146 | |
| 15:12:42 | openstackgerrit | Sujitha proposed openstack/nova master: Add keypairs field to InstancePayload https://review.openstack.org/419730 | |
| 15:14:39 | cfriesen | are there any qemu-knowledgable folks around? When using qcow2 instance storage I'm wondering why we convert backing file images to raw rather than just leaving them in whatever format they were in (raw, qcow2). | |
| 15:18:15 | kaisers_ | mdbooth: Hi! Yep, i tested this manually. --scope works for me. I actually hacked it into the driver in a test instance and killed the n-cpu process hard, mount point remained active | |
| 15:18:36 | mdbooth | kashyap: That's not quite the same thing | |
| 15:18:37 | kashyap | cfriesen: Good question, I once recall digging it, and now lost it from memory | |
| 15:18:39 | mdbooth | Eurgh | |
| 15:18:45 | mdbooth | kaisers: ^^^ | |
| 15:18:55 | kaisers_ | mdbooth: :) | |
| 15:19:08 | mdbooth | kaisers_: Did you try stopping the nova-compute service? | |
| 15:19:16 | kaisers_ | mdbooth: systemd produces a slice unit who becomes parent of the process | |
| 15:19:27 | kaisers_ | hmm | |
| 15:19:31 | kaisers_ | mdbooth: i can revisit that | |
| 15:19:56 | kaisers_ | mdbooth: but that should behave similarily | |
| 15:20:00 | mdbooth | kaisers_: Having read it again, I suspect it should work | |
| 15:20:14 | kaisers_ | mdbooth: as the bug comes up as soon as the cgroup is closed | |
| 15:20:27 | mdbooth | And frankly this is in your driver, so I'd personally be inclined to take your word for it as a functional test would be logistically difficult | |
| 15:20:33 | kashyap | cfriesen: Are you hitting this? -- "Converting image from qcow2 to raw is slow for images with large virtual size" -- https://bugs.launchpad.net/nova/+bug/1484992 | |
| 15:20:33 | openstack | Launchpad bug 1484992 in OpenStack Compute (nova) "Converting image from qcow2 to raw is slow for images with large virtual size" [Wishlist,Confirmed] | |
| 15:20:54 | kaisers_ | mdbooth: ack | |
| 15:20:54 | mdbooth | kaisers_: However, given the bug you're trying to fix, if it were me I'd want to test this | |
| 15:21:11 | kaisers_ | yep, i'll do that | |
| 15:24:14 | sfinucan | leakypipes: Done | |
| 15:24:23 | leakypipes | sfinucan: thanks! | |
| 15:24:37 | leakypipes | sfinucan: or however you say thanks in Irish :P | |
| 15:25:03 | sfinucan | leakypipes: Go raibh maith agat ;) | |
| 15:25:11 | leakypipes | ah, well, that too! :) | |
| 15:25:42 | sfinucan | leakypipes: This is one to stick in your review queue for today :) https://review.openstack.org/#/c/361140/ | |
| 15:26:21 | sfinucan | ...or for next weeks spec review drive | |
| 15:26:28 | cfriesen | kashyap: yep. and in our case it's made worse by the fact that we do the conversion with ionice and no output caching (necessary to minimize impact of the conversion IO load on the instances using the same disk) | |
| 15:26:39 | kashyap | cfriesen: I'm looking at this function _disk_qcow2_to_raw() | |
| 15:26:48 | kashyap | And it's not being used anywhere from my current `grep`ing | |
| 15:27:23 | kashyap | It was introduced as a helper to be used in nova/virt/libvirt -> finish_migration() | |
| 15:27:24 | cfriesen | kashyap: in our case we're hitting it when downloading the backing image file during a block live migration for a qcow2-based instance | |
| 15:28:35 | kashyap | cfriesen: Do you have a full test case? | |
| 15:29:32 | cfriesen | kashyap: boot instance with qcow2 storage, live-migrate it to a compute node that doesn't have the image cached yet. | |
| 15:29:56 | kashyap | cfriesen: I _think_ I know why we do the conversion: | |
| 15:30:06 | mriedem | dansmith: weird the logs are not found for https://review.openstack.org/#/c/452006/ | |
| 15:30:09 | mriedem | on the failed jobs | |
| 15:30:12 | mriedem | anyway i left some comments | |
| 15:30:15 | kashyap | There's an epic comment from mdbooth in the function finish_migration() | |
| 15:30:27 | kashyap | cfriesen: It's to avoid CVEs from format probing | |