Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2022-07-28
12:19:49 sean-k-mooney sound promising
12:21:11 gibi I think it is worth to fix as there was really stupid mistakes caught by this. like mocking the same thing twice in two consecutive decorator and applying assert on the second mock :)
12:43:08 opendevreview ribaudr proposed openstack/python-novaclient master: Microversion 2.91: Support specifying destination host to unshelve https://review.opendev.org/c/openstack/python-novaclient/+/831651
12:48:43 bauzas ok, will rebase my novaclient change on top of ^
12:50:06 opendevreview ribaudr proposed openstack/python-novaclient master: Microversion 2.91: Support specifying destination host to unshelve https://review.opendev.org/c/openstack/python-novaclient/+/831651
12:52:01 Uggla bauzas, hoping you will not have too many conflicts.
12:52:17 bauzas Uggla: nope, just one I guess
12:52:23 bauzas for the microversion number
13:01:28 opendevreview Sylvain Bauza proposed openstack/python-novaclient master: Add support for 2.92 : keypair import mandatory https://review.opendev.org/c/openstack/python-novaclient/+/851231
13:01:41 bauzas Uggla: actually, no merge conflict I got \o/
13:01:45 bauzas just a simple rebase
13:01:56 Uggla bauzas, \o/
13:07:10 opendevreview Amit Uniyal proposed openstack/nova master: For evacuation, ignore if task_state is not None https://review.opendev.org/c/openstack/nova/+/848886
13:33:54 stephenfin gibi: nice work :)
14:12:47 gibi TIL: if there are both side_effect and return_value is defined on a mock then return_value is ignored
14:14:50 gibi bauzas: when you are around, could you hit https://review.opendev.org/c/openstack/os-traits/+/844336 ?
14:15:04 bauzas I am and I can
14:15:21 gibi thank you
14:15:32 gibi I will update the os-trait release patch once that lands
14:15:58 bauzas done
14:17:31 gibi awesome thanks
14:26:56 opendevreview Merged openstack/os-traits master: Add traits for vIOMMU https://review.opendev.org/c/openstack/os-traits/+/844336
14:30:31 gibi sean-k-mooney, bauzas: I've updated the os-traits release patch https://review.opendev.org/c/openstack/releases/+/851390
14:33:18 efried :*
14:33:30 gibi o/
14:34:27 gibi efried: how is life?
14:35:38 efried Living the dream. Working on https://github.com/openshift/hive/ mostly. Enjoy it a lot.
14:35:38 efried What's new with y'all?
14:38:12 gibi good to hear that
14:39:25 gibi I got involved writing k8s operators to deploy OpenStack so I will look at hive at some point to learn some tricks :)
14:40:05 gibi while also trying to make PCI tracking in Placement happen
14:44:18 efried Not gonna claim hive is the examplar of k8s operators, but it has some pretty cool stuff going on.
14:44:18 efried In theory hive will deploy OpenStack -- though I've never tried it myself.
14:47:19 whoami-rajat hi #openstack-nova , wanted to request reviews on my feature, it's been sitting for 3 months without reviews and just wanted to get it in in the Zed cycle https://review.opendev.org/c/openstack/nova/+/831219
14:47:33 whoami-rajat 1) https://review.opendev.org/c/openstack/nova/+/820368
14:47:36 whoami-rajat 2) https://review.opendev.org/c/openstack/nova/+/831219
14:47:42 whoami-rajat 3) https://review.opendev.org/c/openstack/nova/+/830883
14:49:16 dansmith whoami-rajat: sorry I lost track of that.. tempest tests are now working?
14:49:38 dansmith whoami-rajat: the top patch is in merge conflict
15:01:14 sean-k-mooney i started reviewing on eof thos yesterday but was tied up with downstream stuff
15:01:49 sean-k-mooney oh it was the tempest patch
15:03:12 sean-k-mooney dansmith: im pretty sure that passwords are optional for rebuild even if you orginaly set one but im not sure about what tempest needs
15:03:18 sean-k-mooney im expecting it to use the ssh key it has
15:03:35 sean-k-mooney and fall back to the default cirros one otherwise based on its config
15:04:15 dansmith sean-k-mooney: yeah, but that code was copied from somewhere else, and it looked to me like it might be required in case they're using passwords instead of keys
15:04:29 dansmith but regardless, it's just not part of the test so the comment was irrelevant I think
15:05:31 sean-k-mooney ack
15:05:58 sean-k-mooney i havent looked at the nova change yet
15:06:18 sean-k-mooney its on my radar
15:06:55 sean-k-mooney whoami-rajat:^
15:07:02 sean-k-mooney what was for you
15:43:47 whoami-rajat dansmith, yeah, tempest is working, i will rebase the nova patch, not sure if there is any change needed on the tempest patch based on the comments
15:44:44 whoami-rajat sean-k-mooney, some part of the test is taken from rebuilding an image backed instance, and thanks for looking
16:14:15 opendevreview Rajat Dhasmana proposed openstack/nova master: Add support for volume backed server rebuild https://review.opendev.org/c/openstack/nova/+/820368
16:14:15 opendevreview Rajat Dhasmana proposed openstack/nova master: Add conductor RPC interface for rebuild https://review.opendev.org/c/openstack/nova/+/831219
16:14:16 opendevreview Rajat Dhasmana proposed openstack/nova master: Add API support for rebuilding BFV instances https://review.opendev.org/c/openstack/nova/+/830883
16:14:35 whoami-rajat dansmith, ^ rebased
16:21:19 dansmith ack
17:21:32 opendevreview Oliver Walsh proposed openstack/nova master: Skip cell0 in all-cells min version check https://review.opendev.org/c/openstack/nova/+/851440
17:22:52 opendevreview Oliver Walsh proposed openstack/nova master: Skip cell0 in all-cells min version check https://review.opendev.org/c/openstack/nova/+/851440
17:26:50 dansmith owalsh_: so your templated urls don't have DB names in them, it seems
17:27:14 owalsh_ dansmith: correct, has --database_connection='{scheme}://{username}:{password}@{hostname}/?{query}'
17:27:14 dansmith they have hostnames per cell, but I'm not exactly sure why.. are those pointing to hostnames that are VIPs for separate DB instances or something/
17:27:50 owalsh_ so on the cell1 conductor in nova.conf it the db connection will point to it's mariadb, cell0 is on a different mariadb
17:28:40 owalsh_ both are VIPs
17:28:48 dansmith owalsh_: I need to see the full stack of where this is called from
17:29:24 dansmith owalsh_: I guess the problem is that this is being called from a cell conductor and not a superconductor and that's the problem.. and I guess that's because the cell conductor has api_database configured and thus thinks it is supposed to do the fanout?
17:29:35 dansmith but I want to know what the operation is that gets us here
17:30:06 dansmith actually,
17:30:31 dansmith the superconductor will be the same because templated urls are never going to work if you have DBs on different hostnames but try to use {hostname} since that will only ever resolve to one
17:30:51 dansmith so you probably need to just not use {hostname} in the URLs if the DBs are at specific locations
17:31:06 dansmith if you're using DNS names pointing to VIPs like cell1.foo and cell0.foo, then you don't need the template at all right?
17:31:59 owalsh_ full stack doesn't give any more info but I'll post it. It's called from utils.raise_if_old_compute() in service create() AFAICT
17:32:40 dansmith okay, but need to know what's being done to know who is calling that and for why
17:33:01 dansmith but if we want to exclude cell0 from that check (which *is* specific to computes) we should do it there and not in this generalized method
17:33:19 dansmith however, like I say, that's just punting the can down the road, because a superconductor arrangement like this with multiple cells will not work
17:33:46 owalsh_ ack, in tripleo superconductor will use the same DB server as cell0 so it happens to work
17:34:12 dansmith okay
17:34:31 dansmith but if you use per-db hostnames, then no need for the {hostname} in the template at all right?
17:34:53 dansmith (per-db hostnames being far superior, of course)
17:35:17 dansmith the template was really for the credentials and not the other things, I just added every part of the url into the templating engine for flexibility
17:41:03 sean-k-mooney[m] im having some issues with my work laptop so im going to set up my personal one quickly so ill be on matrix for a bit rather then irc
17:41:11 sean-k-mooney[m] i should see either ping
17:41:15 sean-k-mooney[m] but just an fyi
17:43:20 gibi_pto I will be off tomorrow, but back on Monday
17:53:23 opendevreview Balazs Gibizer proposed openstack/nova master: Remove double mocking https://review.opendev.org/c/openstack/nova/+/851445
17:54:56 gibi_pto stephenfin, sean-k-mooney[m]: ^^ it is just a partial fix but I run out of time today and I will be off tomorrow so I pushed it. feel free to take over tomorrow if you have time
17:59:25 sean-k-mooney[m] gibi_pto: ack
18:00:42 owalsh_ dansmith: nova-conductor logs - https://paste.openstack.org/show/b85LsPayPl2QYN6RrDi1/, just noticed that it times out waiting for cell0 after 1 minutes so not fatal
18:03:11 owalsh_ dansmith: even if we just switch to the VIP hostnames might have issues if there is a firewall between the cell conductor and the cell0 db
18:06:12 dansmith owalsh_: yeah, so making that compute-specific check might be good, just needs to be one level up
18:06:16 dansmith I'll look at the logs in just a sec
18:08:48 dansmith owalsh_: yeah it should handle the failure gracefully
18:09:35 dansmith owalsh_: that's not enough to tell what's going on because of how the stack frame gets chopped for a thread, unfortunately
18:09:44 dansmith can you correlate that req-id with the api logs or something?
18:18:19 dansmith owalsh_: nm, I got it.. on service startup
18:40:12 owalsh_ dansmith: hrmm, I guess it's not just the hostname that could be an issue. If cell0 username/password is different to the cell nova db it would also fail
18:41:04 dansmith owalsh_: yep for sure, that's one of the compromises
18:41:18 dansmith but the assertion during design was that credentials are more per-host and less per-database
18:43:15 owalsh_ ack, IIRC for tripleo it's really just the {query} part that mattered as it included the local bind address for each host

Earlier   Later