| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-03-09 | |||
| 09:38:33 | finucannot | (rather than posting here) | |
| 09:39:03 | Spaz-Work | Anyone happen to have a map of how the select_destinations codebase works in nova-scheduler? | |
| 09:39:09 | Spaz-Work | I've run into circles about 13 trimes now | |
| 09:40:00 | openstackgerrit | Zhenyu Zheng proposed openstack/nova-specs master: Add host field to instance action event https://review.openstack.org/550411 | |
| 09:41:32 | Spaz-Work | Ah I think I see.. filter_scheduler is what is extending that.. couldn't find what was using it the abstract | |
| 09:42:05 | zigo | finucannot: http://paste.openstack.org/show/696053/ | |
| 09:43:02 | zigo | I really would like to understand better the db thing for cells, does it uses the db of nova, or nova-api, or something else? | |
| 09:43:19 | zigo | What am I supposed to setup? Is there a db-sync somewhere for the cells? | |
| 09:44:51 | zigo | Which of cell0 and cell1 needs a db? Both? With different dbs? | |
| 09:45:01 | zigo | Should they both access rabbit? | |
| 09:48:38 | finucannot | zigo: I think there's should be separate message queue per cell, plus a top-level API one | |
| 09:49:15 | zigo | finucannot: Even in my simple single node setup for package validation? | |
| 09:49:15 | finucannot | They _could_ use the same RabbitMQ instance (is that the correct term?) but it should be some form of logical split nonetheless | |
| 09:49:41 | zigo | Ah, right, like a different rabbit vhost or something... | |
| 09:49:48 | zigo | That's doable. | |
| 09:50:06 | finucannot | I can't say, to be honest. Anything I know about cells v2 was gleaned from a talk dansmith gave in...Sydney? | |
| 09:50:14 | finucannot | vhost...yeah, that sounds familiar | |
| 09:50:32 | zigo | What they call vhost in rabbit is a kind of path. | |
| 09:50:41 | zigo | Like / or /cell0 ... | |
| 09:51:10 | finucannot | zigo: https://www.youtube.com/watch?v=o7jkqzORij8 | |
| 09:51:22 | finucannot | We _surely_ have this in written form somewhere, but lord knows where | |
| 09:51:44 | bauwser | finucannot: zigo: you need to have separate MQs for cells vs. API, but both can be on the same broker | |
| 09:51:57 | bauwser | for AIO at least | |
| 09:52:18 | zigo | Ok. | |
| 09:52:19 | bauwser | https://docs.openstack.org/nova/latest/user/cellsv2-layout.html | |
| 09:52:22 | bauwser | finucannot: ^ | |
| 09:52:25 | finucannot | zigo: This looks somewhere helpful https://docs.openstack.org/nova/latest/user/cellsv2-layout.html Three databases, as suspected | |
| 09:52:31 | finucannot | bauwser: Jinx :) | |
| 09:52:32 | bauwser | jinx ! | |
| 09:53:06 | zigo | I'll read all of this, and will come back here if I don't get it, thanks guys, that's very helpful. | |
| 09:53:11 | bauwser | np | |
| 09:53:17 | bauwser | and sorry for being AWOL | |
| 09:53:31 | bauwser | thank ABRT signal | |
| 09:54:14 | bauwser | looks something is tainting my kernel so it sens the abrt signal | |
| 09:54:16 | zigo | bauwser: Switch to Debian ! :) | |
| 09:54:25 | bauwser | *cough cough* | |
| 09:54:45 | bauwser | I was running Ubuntu before, but I feel somehow corporate | |
| 09:55:03 | bauwser | although I have nothing preventing me to install any other distro | |
| 09:55:28 | bauwser | and tbh, I'm pretty happy with F27, works like a charm | |
| 10:02:02 | Spaz-Work | I may go back to Fedora next time | |
| 10:02:10 | Spaz-Work | It's been a long journey and i'm ready to come home and brick after updates | |
| 10:02:17 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Replace raise with self.fail() in unit test https://review.openstack.org/550914 | |
| 10:02:40 | Spaz-Work | scheduler you make my head hurt.. but imma learn you. | |
| 10:03:30 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Raise a proper exception in unit test https://review.openstack.org/550914 | |
| 10:26:21 | zigo | finucannot: bauwser: All of this doc and talk doesn't answer the practical questions: how do I setup my all-in-one-server setup, how do I db-sync the cell0 and cell1 dbs. | |
| 10:26:38 | zigo | And how can I make multiple conductors access to different rabbitmq vhosts ... | |
| 10:26:50 | zigo | They'd be sharing the same nova.conf... | |
| 10:27:07 | lyaaaaaaaarwood | ratailor: re https://review.openstack.org/#/c/401009 one nit and a quick comment based on the initial .save() in that method. | |
| 10:27:28 | bauwser | zigo: devstack can help | |
| 10:27:44 | zigo | bauwser: How can it help me doing that setup with packages? | |
| 10:27:55 | zigo | Do you mean, i should read devstack's scripts? | |
| 10:27:59 | bauwser | by looking on how it's setup | |
| 10:28:00 | bauwser | yeah | |
| 10:28:03 | zigo | k | |
| 10:28:23 | finucannot | It's a horrible answer, but yeah. It's that or use some other tool to install this stuff :( | |
| 10:28:56 | finucannot | I will note that DevStack configures multiple nova.conf files. nova.conf, nova-cpu.conf, nova_cell1.conf, ... | |
| 10:28:56 | zigo | bauwser: finucannot: Should I attempt to just implement this on my single server setup? https://docs.openstack.org/nova/latest/_images/graphviz-609e99b08abf22927d47c939873591359f7e50e3.png | |
| 10:29:27 | zigo | (ie: single conductor, 1 db for each of cell0 and cell1) | |
| 10:30:10 | bauwser | zigo: finucannot: AFAIK, we pass the transport_url when we create the cell as a parameter | |
| 10:31:43 | bauwser | zigo: so, there are 2 points : one is the MQ, the other is the DB | |
| 10:32:09 | bauwser | Nova requires 3 logical DBs, but both can share the same DB engine | |
| 10:32:16 | zigo | bauwser: How do I get the cell0 db populated with tables? Is there some kind of cell-db-sync ? | |
| 10:32:43 | Roamer` | is there a way to let Nova know that the new host that just appeared is actually a renamed version of an old host that it knew about, but that it shows as down now? :) | |
| 10:32:51 | Roamer` | (short of mucking around in the MySQL database, I mean...) | |
| 10:33:20 | Roamer` | I guess I could just delete the database record for the old host, but I wonder what inconsistencies this will cause :) | |
| 10:33:45 | bauwser | zigo: sec, there are docs for that | |
| 10:33:55 | zigo | Thanks a lot. | |
| 10:34:27 | zigo | I'm sure all of what I'm asking for will end up being just a few lines in my shell script, but those aren't easy to find out. | |
| 10:34:29 | bauwser | zigo: https://docs.openstack.org/nova/latest/install/controller-install-ubuntu.html#install-and-configure-components | |
| 10:34:55 | bauwser | zigo: and https://docs.openstack.org/nova/latest/install/compute-install-ubuntu.html#add-the-compute-node-to-the-cell-database | |
| 10:35:57 | zigo | Allright, I'm back to the original issue. | |
| 10:36:26 | zigo | When I do a nova-manage db sync, I get this: | |
| 10:36:30 | zigo | http://paste.openstack.org/show/696077/ | |
| 10:37:03 | zigo | Indeed, my novadb_cell0 is completely empty, that's to be expected. | |
| 10:37:34 | zigo | How come "nova-manage db sync" doesn't get that? | |
| 10:37:37 | zigo | How to fix? | |
| 10:44:55 | cdent | zigo: there's yet another page about cells, with some additional steps https://docs.openstack.org/nova/latest/user/cells.html#setup-of-cells-v2 (or better https://docs.openstack.org/nova/latest/user/cells.html#step-by-step-for-common-use-cases ) | |
| 10:45:22 | zigo | cdent: Thanks! | |
| 10:45:45 | zigo | THAT's what I was looking for ! :) | |
| 10:45:58 | cdent | :) | |
| 10:46:14 | zigo | Hum... Still... | |
| 10:46:25 | zigo | cdent: cell0 should point to the "normal" novadb, or ... ? | |
| 10:46:42 | openstackgerrit | Stephen Finucane proposed openstack/nova master: Rename '_numa_get_constraints_XXX' functions https://review.openstack.org/385072 | |
| 10:46:43 | openstackgerrit | Stephen Finucane proposed openstack/nova master: Standardize '_get_XXX_constraints' functions https://review.openstack.org/385071 | |
| 10:47:29 | cdent | zigo: if you go up the page to "first time setup" there's some info on what database cell0 gets (its own) | |
| 10:47:53 | cdent | if don't use --database-connection a default will be chosen | |
| 10:48:23 | zigo | cdent: Then I'm back to square 1, with db-sync failing because there's no migrate_version table. | |
| 10:48:48 | zigo | ie: paste.openstack.org/show/696077/ | |
| 10:48:57 | zigo | pymysql.err.ProgrammingError: (1146, "Table 'novadb_cell0.migrate_version' doesn't exist") | |
| 10:49:00 | zigo | :/ | |
| 10:49:25 | cdent | zigo: I think you need create the db yourself | |
| 10:49:34 | zigo | cdent: I did. | |
| 10:49:41 | cdent | once it exists then those commands will have something to talk to | |
| 10:49:45 | zigo | The db is created with username and password. | |
| 10:50:05 | zigo | Hum... probably it's just a login/pass issue, and I need to fix it? | |
| 10:51:26 | cdent | zigo: probably can compare with what the steps here: https://github.com/openstack-dev/devstack/blob/master/lib/nova#L716 | |
| 10:52:08 | cdent | if you're using user and pass then they probably need to be the same across database, unless you are being explicit and using --database-connection | |
| 10:55:36 | zigo | I checked, my credentials are correct for cell0 | |
| 10:57:41 | zigo | And login / pass is the same for the novadb and novadb_cell0... | |
| 10:57:43 | zigo | So it's not that. | |