| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-09-22 | |||
| 14:29:32 | finucannot | leakypipes, sahid: That patch has been open for a long time as is a pretty trivial TODO resolution. Could ye take a look at some point | |
| 14:29:43 | finucannot | artom: I didn't try, just in case :) | |
| 14:29:48 | artom | 'cuz finucan't keeps the same number of syllable | |
| 14:29:51 | artom | syllables | |
| 14:29:55 | leakypipes | finucannot: yuppers. just finishing up a review on ralonsoh__'s patch | |
| 14:30:03 | finucannot | leakypipes: (y) | |
| 14:30:31 | finucannot | artom: Hexchat rejects it anyway, so I assume not | |
| 14:30:38 | liuyulong | No. But it is a key data, if rebuilt instance or cloud-platform does not support password. Then the key pair is the only one way to login. | |
| 14:30:51 | artom | finucannot, yeah, I'm pretty sure it's not allowed | |
| 14:31:02 | ralonsoh__ | leakypipes: thanks for the review | |
| 14:31:03 | liuyulong | instance can be running without volume and port. | |
| 14:31:24 | mriedem | liuyulong: Kevin_Zheng's spec provides a way to update the keypair which would still be used when rebuilding | |
| 14:31:25 | liuyulong | but instance can be useless without a way to login. | |
| 14:31:33 | mriedem | and rebooting | |
| 14:31:35 | mriedem | and migrating | |
| 14:31:36 | mriedem | and unshelving | |
| 14:32:31 | Kevin_Zheng | Yeah, the new way will be more generic | |
| 14:32:34 | leakypipes | finucannot: k, on to your patch. swap with you... https://review.openstack.org/#/c/474892/ | |
| 14:32:37 | figleaf | superdan: working on your comments about the Selection object. | |
| 14:32:50 | bauzas | damn Friday gods | |
| 14:32:58 | leakypipes | ralonsoh__: no problemo :) | |
| 14:33:16 | figleaf | superdan: is there a general way of handling the "Cannot load 'node_name' in the base class" type of errors when not every value is passed at creation time? | |
| 14:33:24 | liuyulong | again, its totally two different ways. One API make sense while rebuild. That also make consistent with instance name, image, admin pass. | |
| 14:33:26 | finucannot | leakypipes: I feel this is an unworthy trade, but I like ralonsoh__ so deal ;) | |
| 14:33:52 | superdan | figleaf: everything should be passed at create time for this I would think | |
| 14:33:59 | leakypipes | finucannot: +2 from me on that | |
| 14:34:06 | ralonsoh__ | finucannot: me too hehehe | |
| 14:34:12 | leakypipes | now on to figleaf's patches... | |
| 14:34:28 | figleaf | leakypipes: be aware that I have updates to post soon | |
| 14:34:39 | figleaf | you might want to wait | |
| 14:35:16 | Kevin_Zheng | liuyulong: I think the Matts point is, we don't need two ways, if we open this door, people may ask to add it too to reboot API, etc | |
| 14:36:34 | liuyulong | If new API does not supporting running injection, the DB keypair and guest keypair are not consistent without a rebuiding, reboot, unshelving... This looks not good. | |
| 14:36:43 | leakypipes | figleaf: k. just reading superdan | |
| 14:36:49 | leakypipes | s review of the Selection patch now. | |
| 14:36:57 | leakypipes | figleaf: will hold off on commenting. | |
| 14:37:14 | liuyulong | Kevin_Zheng, no, I'll not agree that. If its not a `recreating` API. | |
| 14:37:43 | Kevin_Zheng | liuyulong: yeah, we will document it well | |
| 14:38:48 | liuyulong | Kevin_Zheng, document can not avoid that inconsistent. Cloud user may not remember what they have done, : ) | |
| 14:39:46 | liuyulong | So, again, IMHO, both specs are OK for now. One step for rebuild API make more reasonable. | |
| 14:43:35 | mriedem | Kevin_Zheng: comments in your spec https://review.openstack.org/#/c/506552/ plus mikal needs to review that | |
| 14:43:55 | takashin | oomichi: Are you around? | |
| 14:44:20 | mriedem | liuyulong: if the user doesn't know what they are doing, they shouldn't be doing it | |
| 14:44:39 | mriedem | one could add layers of orchestration on top of this if needed for simple users, | |
| 14:44:56 | mriedem | like horizon could have a window where you update the keypair on the instance and it asks you to reboot or rebuild the instance | |
| 14:45:08 | mriedem | you select one and horizon does the work of updating the keypair and rebuilding the instance | |
| 14:45:26 | Kevin_Zheng | mriedem: thanks I will check tomorrow, still have to work tomorrow T_T 4th Saturday every month | |
| 14:45:42 | mriedem | Kevin_Zheng: enjoy your time in the salt mines | |
| 14:46:10 | mriedem | i have to go to my daughter's gymnastics class in the morning, | |
| 14:46:13 | mriedem | we all have our burdens | |
| 14:47:25 | Kevin_Zheng | We will have a 8 day holiday though, starting from 1st Oct, nation day | |
| 14:47:37 | finucannot | ralonsoh__: One comment left on https://review.openstack.org/#/c/474892/ | |
| 14:47:45 | mriedem | Kevin_Zheng: you've got me beat there | |
| 14:49:09 | liuyulong | mriedem, your review comments also reminded me. If cloud-init does not running in every boot, the inconsistent can not be removed. To provent that, the entire cloud platform images need updating. | |
| 14:49:50 | openstackgerrit | Rodolfo Alonso Hernandez proposed openstack/nova master: Add datapath type information to OVS vif objects https://review.openstack.org/474892 | |
| 14:49:59 | ralonsoh__ | finucannot: done! | |
| 14:51:00 | finucannot | ralonsoh__: ...and done | |
| 14:54:08 | mriedem | liuyulong: then you rebuild the instance | |
| 14:54:27 | liuyulong | And for the running instance. If user does not change the cloud-init config. They create image and then boot a new VM. New API can not avoid inconsistent in such situation ether. | |
| 14:55:02 | liuyulong | s/ether/either | |
| 14:55:29 | mriedem | liuyulong: if you boot a new vm, you're providing a new keypair anyway | |
| 14:55:58 | mriedem | the user does'nt change the cloud-init config, that's up to the deployer | |
| 14:56:47 | liuyulong | ha, we are talking back to rebuild API again. | |
| 14:57:10 | mriedem | its very simple, if all you want to do is rebuild and have a new keypair, (1) update the keypair on the instance, (2) rebuild the instance | |
| 14:57:28 | mriedem | if (2) is something else, like cold migrate, unshelve, resize, reboot, then you can do that because of (1) | |
| 14:57:58 | liuyulong | What if cloud-init does not running in every boot ? | |
| 14:58:18 | openstackgerrit | Stephen Finucane proposed openstack/nova master: docs: Rename cellsv2_layout -> cellsv2-layout https://review.openstack.org/498821 | |
| 14:58:18 | openstackgerrit | Stephen Finucane proposed openstack/nova master: doc: Cleanup of existing index pages https://review.openstack.org/498819 | |
| 14:58:19 | openstackgerrit | Stephen Finucane proposed openstack/nova master: WIP! doc: Add contents page https://review.openstack.org/498820 | |
| 14:59:07 | liuyulong | rebuild is something like `recreate`, the new VM disk cloud-init may not cache the old instance id. So it can be running. | |
| 14:59:48 | liuyulong | But booting actions like reboot, resize, unshelve do not have such situation. | |
| 15:00:34 | sdague | mriedem: https://review.openstack.org/#/c/454323/ has seen a lot of rechecks / rebases and the full stack testing is looking pretty solid | |
| 15:00:55 | sdague | that's the live snapshot by default | |
| 15:05:35 | mriedem | sdague: there is something new in there | |
| 15:05:37 | mriedem | unless i'm blind | |
| 15:08:15 | liuyulong | Allow me to rephrase: And for the running instance. If user does not change the cloud-init config. They create image and then boot a new VM. Even the entire cloud platform images have been updated, new API can not avoid inconsistent for this new VM. | |
| 15:13:08 | openstackgerrit | konstantin proposed openstack/nova master: don't add device address if there is no any units https://review.openstack.org/506686 | |
| 15:13:09 | openstackgerrit | konstantin proposed openstack/nova master: switch from filesystem to disk for parallels containers https://review.openstack.org/506687 | |
| 15:13:28 | sdague | mriedem: there are test changes | |
| 15:13:38 | sdague | and there is the bit you changed in patch #2 | |
| 15:13:40 | mriedem | no, the PAUSED thing | |
| 15:13:52 | sdague | so PS #2 are your changes | |
| 15:14:01 | mriedem | i don't remember that | |
| 15:14:10 | mriedem | you hacked my account back in april | |
| 15:14:27 | sdague | I can slice them out again | |
| 15:14:36 | mriedem | https://review.openstack.org/#/c/454323/1..2/nova/virt/libvirt/driver.py | |
| 15:14:37 | mriedem | wtf | |
| 15:14:58 | mriedem | it was probably needed for something | |
| 15:16:38 | sdague | here is a version with that deleted back out | |
| 15:16:40 | openstackgerrit | Sean Dague proposed openstack/nova master: Change livesnapshot to true by default https://review.openstack.org/454323 | |
| 15:16:57 | mriedem | suspend does a managedSave | |
| 15:17:00 | mriedem | but pause doesn't | |
| 15:17:00 | sdague | I do think there is a state transition which is wrong in a tempest test | |
| 15:17:08 | sdague | ok | |
| 15:17:38 | sdague | well, I have to go do a preschool pickup, so ponder what's needed there, as you made that slice of changes :) | |
| 15:19:49 | mriedem | yeah idk, don't get it, would probably have to see if tests fail, or ask kashyap | |
| 15:20:20 | mriedem | i think we have a tempest test that will snapshot a paused instance and it will fail, but we'll see | |
| 15:20:43 | mriedem | https://wiki.openstack.org/wiki/Kvm-Pause-Suspend | |
| 15:20:45 | mriedem | oops | |
| 15:20:49 | mriedem | test_create_image_from_paused_server | |
| 15:20:50 | mriedem | that's the one | |