Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-01-09
15:48:50 bauzas cdent: dansmith: for all the drivers but Ironic, host == node
15:49:37 bauzas cdent: dansmith: should I add a new DB call for getting the host name just because we have Ironic ? meh
15:51:53 cdent bauzas: I'm not understanding what you're trying to do? Are you saying: you have a list of compute nodes from placement api, now you want to make a collection of host states?
15:52:03 bauzas cdent: exactly
15:52:11 bauzas cdent: but we're keying per a tuple
15:52:18 bauzas cdent: which (host, node)
15:52:24 bauzas cdent: so I need to know the host value
15:52:47 cdent I believe jay has made some adjustment on that, at least in the resource tracker, but I'm not sure if it impacts where you're at
15:52:51 bauzas cdent: for 90% of our deployments, host == hypervisor_name
15:53:39 bauzas but given Ironic has host != hypervisor_hostname, I probably need to add a helper method in the ComputeNode object like compute_node.get_by_resource_provider(uuid)
15:53:56 bauzas or I could just get_by_uuid() because fortunately, cn.uuid == rp.uuid
15:54:15 bauzas but that sounds silly to DB lookup just because of Ironic
15:54:21 bauzas graaaah
15:55:25 bauzas cdent: and yeah, I know jay's cleanups but AFAICT, we're not changing yet that
15:56:24 cdent presumably he'll show up at some point today and we can get his input too.
16:24:33 cdent mriedem: could you remind me where you left the ptg planning etherpad?
16:24:45 mriedem cdent: https://etherpad.openstack.org/p/nova-ptg-pike
16:24:53 cdent thanks
16:27:42 openstackgerrit Christopher Martin proposed openstack/nova: Change soft delete resources to prevent DOS https://review.openstack.org/386756
16:30:43 mriedem ildikov: i won't be at the nova/cinder meeting probably, i'll be around but have about an hour today to get some stuff done before i have to catch a flight
16:31:00 mriedem ildikov: i'd say priority this week should be figuring out what's up with the endpoint 404 for using cinder v3
16:31:38 ildikov mriedem: tnx for the notice, safe travels!
16:31:47 ildikov mriedem: I didn;t have much time to play more with that
16:32:36 ildikov mriedem: I fixed the issue locally with hard coding the endpoint_template, but for instance the new attach calls in Cinder still acted weird while testing
16:32:55 smcginnis ildikov: Acted weird?
16:32:56 ildikov mriedem: like on call worked from the CLI, but not from Nova
16:33:02 ildikov *one
16:33:03 smcginnis Ah
16:33:17 mriedem well, which cli?
16:33:29 mriedem remember that latest novaclient doesn't proxy through to cinder anymore
16:33:34 ildikov I called attachment-create and I looked into the logs, the exact same looking call was 202 from the CLI and got 404 when Nova called it...
16:33:36 mriedem so things like listing attachments in novaclient aren't using cinder
16:33:41 ildikov Cinder CLI, sorry
16:33:55 ildikov the old Nova attach call worked fine
16:34:11 ildikov but the new call failed for at least seemingly no good reason
16:34:12 smcginnis ildikov: Well that's good. At least the existing stuff didn't break with it.
16:34:15 openstackgerrit Sylvain Bauza proposed openstack/nova: WIP/DNM : Put all the magic in a PlacementClient https://review.openstack.org/417111
16:34:15 openstackgerrit Sylvain Bauza proposed openstack/nova: WIP: Scheduler calling the Placement API https://review.openstack.org/417961
16:34:23 openstackgerrit Dan Smith proposed openstack/nova: WTFWIP: Multicell support for instance listing https://review.openstack.org/396775
16:34:23 openstackgerrit Dan Smith proposed openstack/nova: WIP: Make metadata server know about cell mappings https://review.openstack.org/417962
16:35:03 ildikov mriedem: smcginnis: in Devstack with admin cred I got 'Volume not found' and with demo cred I got 'Malformed request' error
16:35:44 ildikov it can still be some misconfig on my side, not saying it's not, but still looks odd
16:36:06 ildikov smcginnis: no, it seemed to work, but I didn't do any thorough testing on it either
16:36:18 mriedem idk, i guess if you're hitting just issues with using cinder (no nova involved) with v3 then i'd start opening bugs
16:36:22 openstackgerrit sean mooney proposed openstack/os-vif: introduces MTU support for vhost-user https://review.openstack.org/410135
16:36:35 ildikov smcginnis: and with just setting catalog_info my env failed like the logs on the gate for the switch to v3 patch
16:37:03 ildikov mriedem: with Cinder CLI it works like a charm really
16:37:09 smcginnis ildikov: Did you see the call hit c-api? Is it hitting the right endpoint?
16:37:20 openstackgerrit Sergey Nikitin proposed openstack/nova: Change parent of diagnostics objects to NovaObject class https://review.openstack.org/394480
16:37:21 openstackgerrit Sergey Nikitin proposed openstack/nova: Standardization of VM diagnostics info API. https://review.openstack.org/355540
16:37:21 openstackgerrit Sergey Nikitin proposed openstack/nova: Completed implementation of instance diagnostics for Xen https://review.openstack.org/399613
16:37:33 ildikov mriedem: I think somehow the endpoint or _SESSION info is not OK in Nova, or it looks like it at least
16:37:47 ildikov smcginnis: I checked the Cinder API logs
16:38:02 ildikov smcginnis: same looking request one succeeds one fails
16:38:17 ildikov smcginnis: after setting endpoint_template
16:38:52 ildikov smcginnis: the other in Nova was that the Nova API could reach Cinder after setting only the catalog_info to v3 and Nova compute failed with endpoint not found
16:42:45 openstackgerrit Andrey Volkov proposed openstack/nova-specs: Add hierarchy quota driver specification https://review.openstack.org/394422
16:49:25 mriedem andreykurilin: have you seen the bug i reported about the novaclient functional tests failing?
16:49:42 mriedem https://bugs.launchpad.net/python-novaclient/+bug/1654806
16:49:42 openstack Launchpad bug 1654806 in devstack "Functional tests (with neutron) failing with "ERROR (Conflict): Multiple possible networks found, use a Network ID to be more specific. (HTTP 409)" since 1/6" [High,In progress] - Assigned to Matt Riedemann (mriedem)
16:50:12 mriedem i tried reverting a few devstack changes from 1/6 related to neutron which i thought were causing the problem and then ran those here https://review.openstack.org/#/c/417645/ but that doesn't seem to be the issue
16:50:55 mriedem sc68cal: would https://review.openstack.org/#/c/404931/ cause multiple tenant networks to start showing up? one for flat and one for vxlan?
16:51:57 sc68cal let me see
16:52:12 sc68cal mriedem: no, that's still a lib/neutron thing
16:52:29 sc68cal we're still running the lib/neutron-legacy code
16:53:10 andreykurilin mriedem: hi! unfortunately, no. my country has holidays after New Year (not like most of countries has Christmas holidays at the end of December) :)
16:53:28 mriedem andreykurilin: ah ok nvm then
16:53:35 sc68cal mriedem: let me start doing some git log on stuff and see if I can help you
16:53:41 mriedem sc68cal: hrmph, ok, i'm a bit stuck then
16:54:14 mriedem sc68cal: i had also reverted https://review.openstack.org/#/c/406186/ but that doesn't appear to be the cause either
16:54:19 mriedem whatever it is it started around 1/6
16:54:42 sc68cal mriedem: yeah anything that's just lib/neutron won't be relevant
16:56:03 sc68cal mriedem: sorry for sending you on a wild goose chase, I think I need to document the state of neutron vs neutron-legacy better
16:56:28 mriedem ildikov: if you're using single node devstack i don't think nova-api would work with cinder v3 but nova-compute wouldn't
16:56:34 mriedem they are using the same config and code
16:57:11 andreykurilin mriedem: I'll return to work tomorrow
17:09:06 openstackgerrit Pavel Gluschak proposed openstack/nova: libvirt: make live migration possible with Virtuozzo https://review.openstack.org/355805
17:17:33 openstackgerrit Sylvain Bauza proposed openstack/nova: WIP: Scheduler calling the Placement API https://review.openstack.org/417961
17:17:33 openstackgerrit Sylvain Bauza proposed openstack/nova: WIP/DNM : Put all the magic in a PlacementClient https://review.openstack.org/417111
17:17:34 openstackgerrit Sylvain Bauza proposed openstack/nova: Expose a REST API for a specific list of RPs https://review.openstack.org/392569
17:18:09 bauzas dansmith: do you agree with not adding a DB API method just for a object method ?
17:18:21 bauzas dansmith: ie. just calling the DB directly within the object method ?
17:18:36 bauzas dansmith: ie. I need to add a ComputeNode.get_by_uuid method
17:18:51 bauzas dansmith: I just want to write the DB call within that method like we do for the API DB
17:18:58 dansmith I'm buried with other stuff and didn't really see what you asked before
17:19:08 dansmith but if you want a ComputeNode.get_by_uuid call that's probably fine
17:19:28 bauzas dansmith: no worries, nevermind my previous pings, just asking for writing this method
17:19:53 bauzas it seems to me not necessary to modify the DB API for something like that
17:49:52 openstackgerrit Matt Riedemann proposed openstack/python-novaclient: DNM: Test devstack revert for neutronclient functional test failures https://review.openstack.org/417645
17:52:32 sc68cal mriedem: so.... this OS_NOVACLIENT_NETWORK variable
17:53:05 sc68cal neutron bails out if there are multiple networks to choose from, and some code spelunking has led me to that env var
17:54:04 mriedem sc68cal: that's in novaclient itself?
17:54:54 sc68cal yeah I'm looking at this function pick_network
17:57:20 sc68cal mriedem: https://github.com/openstack/python-novaclient/blob/master/novaclient/tests/functional/base.py#L232
17:59:57 sc68cal mriedem: basically, in the presence of multiple networks (which we've always had) you need to pick a network to spawn the instance on
18:00:29 sc68cal mriedem: the only time it's not the case, is if there is only one network available. So if you're a tenant that only has access to one network, it's all good
18:01:16 mriedem sc68cal: kevinbenton is also looking at this in -neutron
18:01:22 openstackgerrit Eric Brown proposed openstack/nova: Remove references to Python 3.4 https://review.openstack.org/417997
18:01:36 mriedem sc68cal: and yeah i know that's how it's always worked, but apparently something changed since 1/6 to make multiple networks available to the tenants in these tests
18:01:56 mriedem if it turns out that some bug was fixed and the novaclient tests were always just wrong and getting by, and we have to fix novaclient, then ok

Earlier   Later