Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-06-07
15:38:19 jaypipes gbarros: you know what would make reviewing that easier? more than 0 lines of documentation on the opencontrail vif plugin repo :)
15:38:57 gbarros jaypipes, tell me about it :) mhenkel keeps telling me it's fully documented... somewhere... :)
15:40:05 gbarros mhenkel, could you help us out by pointing out the docs? Maybe a link to it in the readme?
15:40:21 jaypipes mriedem, gbarros: I thought I remembered some conversation where we wanted to make "vhostuser_vrouter_plug" mention contrail specifically before we merged this patch?
15:40:52 jaypipes gbarros: or is vhostuser_vrouter_plug already in the Neutron vif details and we can't change it?
15:40:58 mriedem jaypipes: the latter
15:41:07 mriedem see my question about that in the 2nd to last patch set
15:41:16 mriedem their plugin now registers with contrail in the name,
15:41:21 gbarros jaypipes, I believe there is a TODO on that?
15:41:22 mriedem but the vif details thing is already baked into neutron
15:41:23 openstackgerrit Merged openstack/nova master: Revert "Remove Babel from requirements.txt" https://review.openstack.org/471498
15:41:26 mriedem yeah i added the todo
15:42:05 jaypipes gbarros: yeah, totes, there's a TODO, was just seeing if I remembered reading a discussion about it correctly :)
15:42:25 jaypipes gbarros: I'm getting old and can't remember all the things.
15:43:40 gbarros jaypipes, you're ahead of me there, I could never remember all the things.
15:45:38 jaypipes mhenkel, gbarros: +W'd.
15:45:45 mriedem i don't understand this https://github.com/openstack/nova/blob/7815108d4892525b0047c787cbd2fe2f26c204c2/nova/virt/libvirt/driver.py#L568
15:45:51 mriedem live_migration_tunneled is a boolean
15:45:53 mriedem so why would it be None?
15:46:04 mriedem if Falsey or True
15:46:06 mriedem god
15:46:30 jaypipes mriedem: the default before the option came into existence?
15:46:37 mriedem it might have been a string
15:46:44 mriedem that's my guess, so i'm checking history
15:47:00 jaypipes ya
15:47:47 mriedem nope it was always boolean https://github.com/openstack/nova/commit/621594fc41d0e07fd63dfe7c3c5cfee9edc380ad#diff-f4019782d93a196a0d026479e6aa61b1R171
15:48:07 mriedem oh but not default
15:48:47 gbarros ty jaypipes
15:49:33 jaypipes mriedem: right.
15:50:18 jaypipes mriedem: True, False, None. The unholy Triumvirate.
15:50:31 openstackgerrit Merged openstack/nova master: Fix cell0 naming when QS params on the connection https://review.openstack.org/471487
15:52:37 openstackgerrit Baodong (Robert) Li proposed openstack/nova-specs master: Expose vlan trunking in metadata/configdrive https://review.openstack.org/471815
15:53:41 gbarros jaypipes to it's a trool type instead of bool? (dad joke...)
15:54:50 openstackgerrit Matt Riedemann proposed openstack/nova master: Cleanup the plethora of libvirt live migration options https://review.openstack.org/471834
15:54:51 mriedem it's a flaming bag of pug manure
15:54:54 jaypipes gbarros: :) yes indeed.
15:55:16 jaypipes mriedem: hey now, that is a step too far sir!
15:55:42 mriedem i shouldn't sully pug manure by relating it to live_migration_tunneled?
15:56:07 gbarros ha
15:56:54 jaypipes mriedem: correct.
15:57:28 jaypipes mriedem: lest I would have to resort to a comment about human babies.
15:57:52 mriedem jaypipes: if you want to talk about baby feces i'm game
15:57:59 mriedem let's bring comstud into this as well
15:58:05 jaypipes mriedem: good god no.
15:58:20 dansmith what is the deal with this traits fail? http://logs.openstack.org/34/469634/10/check/gate-nova-tox-functional-ubuntu-xenial/a007f54/testr_results.html.gz
15:58:37 dansmith doesn't look related, but it seems like a thing that would be 100% fail
15:58:45 dansmith maybe something we hit yesterday and have since resolved?
15:59:08 cdent dansmith: rebase, it's been fixed:
15:59:29 cdent https://review.openstack.org/#/c/471508/
15:59:32 dansmith okay, so it needs a rebase to pass?
15:59:36 dansmith cdent: ^
16:00:21 mriedem it's not 100% fail so you could recheck
16:00:28 openstackgerrit Dan Smith proposed openstack/nova master: placement: use separate tables for projects/users https://review.openstack.org/470645
16:00:59 openstackgerrit Dan Smith proposed openstack/nova master: placement project_id, user_id in PUT /allocations https://review.openstack.org/469634
16:02:02 dansmith really, seems like it would be
16:02:04 dansmith either way, rebased
16:03:58 openstackgerrit Merged openstack/nova master: Send request_id on neutron calls https://review.openstack.org/468356
16:06:44 openstackgerrit Merged openstack/nova master: Send out notifications when instance tags changed https://review.openstack.org/459923
16:13:15 cfriesen_ mriedem: regarding your live migration uri stuff...don't we have another issue too? we pass the unmodified "migrate_uri" to guest.migrate(), and that gets written to the 'migrate_uri' parameter that gets passed to self._domain.migrateToURI3()
16:14:21 mriedem cfriesen_: no
16:15:11 mriedem cfriesen_: you're talking about this https://github.com/openstack/nova/blob/9dde4fe0eacc01c135401da3313c493caa99a30c/nova/virt/libvirt/driver.py#L6256
16:15:25 mriedem migrate_uri might be None https://github.com/openstack/nova/blob/9dde4fe0eacc01c135401da3313c493caa99a30c/nova/virt/libvirt/driver.py#L6210
16:15:35 mriedem if live_migration_inbound_addr is not set
16:15:40 mriedem or you're tunneling
16:16:10 mriedem or, jupiter is aligned with saturn
16:16:24 cfriesen_ mriedem: right...but if we need to add usernames and such to the first parameter, would we ever need to add it to the second?
16:16:38 mriedem it would get added either way
16:16:50 mriedem because they both go through the _live_migration_uri method
16:17:07 mriedem ooops sorry,
16:17:11 mriedem there is a _migrate_uri method
16:17:19 cfriesen_ mriedem: exactly
16:17:19 mriedem which != _live_migration_uri
16:17:20 mriedem nice
16:17:53 mriedem https://github.com/openstack/nova/blob/9dde4fe0eacc01c135401da3313c493caa99a30c/nova/virt/libvirt/driver.py#L681
16:17:59 mriedem "QEMU accept two schemes: tcp and rdma"
16:18:02 mriedem that's wrong, it also takes ssh
16:18:06 cfriesen_ I didn't see the patch adding CONF.libvirt.live_migration_scheme, but arguably it should have tweaked _migrate_uri as well I think
16:32:26 melwitt mriedem: on the config option move thing, it is part of the [compute] group but it's also in the [DEFAULT] group. it looks like all of the compute options are in both groups
16:35:22 mriedem melwitt: are you sure about that?
16:35:30 mriedem i think i sorted out that it wasn't in the [compute] group
16:35:37 mriedem the variables for the groups in nova/conf/compute.py were confusing me
16:35:58 mriedem if it were in both groups and we moved it out of one, w/o deprecating, that would be a bug
16:35:58 melwitt mriedem: I could be wrong, but I thought I saw they are added to both
16:36:30 melwitt hm, yeah
16:37:54 melwitt oh, 'compute_opts' vs 'compute_group_opts', I think I confused them together before
16:38:43 mriedem me too
16:40:43 melwitt phew
16:41:01 mriedem melwitt: while you're here https://review.openstack.org/#/c/446240/
16:41:25 melwitt oh, thanks
16:44:34 sfinucan gabor_antal: Looks like https://review.openstack.org/#/c/451789 is failing now?
16:45:45 sfinucan mriedem: I did some work around libvirt live migration options, because what was there wasn't enough
16:45:50 bauzas oh, weird bug https://bugs.launchpad.net/nova/+bug/1696417
16:45:50 openstack Launchpad bug 1696417 in OpenStack Compute (nova) "nova-manage db online_data_migrations can fail when upgrading to newton under certain conditions" [Undecided,New]
16:46:16 sfinucan mriedem, cfriesen_: tada https://review.openstack.org/#/c/456571/
16:46:27 sfinucan (ignore the spurious build failure)
16:47:00 gabor_antal sfinucan: :O it should not imho but it looks like
16:47:24 mriedem sfinucan: cool that looks like what we need in devstack
16:47:26 mriedem the username one
16:48:29 mriedem sfinucan: i'll test that out with my devstack change
16:48:42 sfinucan (y)

Earlier   Later