13:19:51 <lbeliveau> #startmeeting sriov
13:19:52 <openstack> Meeting started Tue Oct  4 13:19:51 2016 UTC and is due to finish in 60 minutes.  The chair is lbeliveau. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:19:54 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:19:56 <openstack> The meeting name has been set to 'sriov'
13:20:05 <wznoinsk> so, hello ;-)
13:20:27 <sfinucan> howdy
13:20:31 <lbeliveau> sorry about the confusion, moshele did tell me he would be on holidays this month, but didn't told me when :)\
13:20:58 <lbeliveau> #topic Improving SR-IOV/PCI Passthrough CI
13:21:23 <lbeliveau> so got one tempest for cold migration merged, revert case review is ongoing with a +2
13:21:55 <lbeliveau> this is the API tests, and we'll push the more complex test with networking connectivity test once the revert is in
13:22:06 <lbeliveau> wznoinsk: anything on your side ?
13:22:56 <wznoinsk> lbeliveau: I'm getting blocked with nodepool trying to spawn physical servers via ironic for my SRIOV CI sandboxing, I need to fix it or try my luck with SRIOV/PCI passthru in VMs, I'm not sure which is the lesser evil
13:23:34 <wznoinsk> unfortunately nothing to celebrate yet
13:23:43 <lbeliveau> did you try to ping folks on infra ?
13:23:53 <sfinucan> lbeliveau: For context, are these Tempest tests public?
13:23:57 <lbeliveau> I've never tried that myself ...
13:24:02 <wznoinsk> yes, nodepool spawning baremetal is not something they do and test
13:24:08 <sfinucan> (a la the intel-nfv-ci-tests)
13:24:50 <wznoinsk> lbeliveau: I guess when I fix it and infra will like it nodepool will be able to handle tests on baremetal nodes
13:25:29 <lbeliveau> sfinucan: yes, I have to follow up with moshele is it is now part of mellanox CI, for the gate testing with multi-node it's still experimental job
13:25:37 <wznoinsk> has anyone tried double pci passthru? host -> devstack VM -> tempest VM ?
13:26:00 <lbeliveau> not me :(
13:26:07 <sfinucan> wznoinsk: Not yet, but I might try it later this week
13:26:16 <sfinucan> If I do, I'll provide results
13:26:30 <sfinucan> danpb also provided a link for faking PCI devices yesterday, if it's of any use
13:26:33 <wznoinsk> how about PCI passthru of SRIOV nic to a devstack VM for it to pass VFs down to tempest VMs?
13:26:47 <lbeliveau> sfinucan: what is the link ?  I'd like to look at that
13:26:53 <wznoinsk> sfinucan: saw that, I want to test the real devs tho :)
13:27:03 <sfinucan> lbeliveau: see https://paste.fedoraproject.org/442587/54996401
13:27:07 <sfinucan> and eavesdrop logs from yesterday for context
13:27:26 <sfinucan> wznoinsk: Aye, that's fair enough
13:27:39 <sfinucan> could be a good interim measure though. Works for NUMA, right?
13:27:49 <sfinucan> but sorry, I digress
13:28:52 <lbeliveau> wznoinsk: when you try it, what kind of erros are you getting ?
13:30:03 <wznoinsk> nodepool: http://pasteboard.co/ay7KU4myZ.png
13:30:15 <wznoinsk> manual attemps at the same time: http://pasteboard.co/ay8d9Ryoq.png
13:31:05 <lbeliveau> vm is stuck in reboot loop ?
13:32:02 <wznoinsk> just so we're on the same page, this is when trying to get a baremetal server provisioned by ironic using nodepool asking
13:32:30 <wznoinsk> the problem is simple, nodepool tests ssh connection when sshd is still saying 'server is still booting', it needs a proper retry loop
13:32:45 <wznoinsk> I just need to setup the nodepool on my dev env
13:33:13 <lbeliveau> do you see anything unusual in /var/log/libvirt/libvirtd.log and /var/log/libvirt/qemu/instance-000000x.log ?
13:33:39 <lbeliveau> ok ... I taught the vm was maybe not coming up
13:33:52 <wznoinsk> it's provision the physical server, not a VM on it
13:34:29 <lbeliveau> ok we'll need help from lennyb and moshele I guess, I don't know much about CI env myself
13:34:36 <wznoinsk> anyways I hope to have better news next week, we aim to get the SRIOV/PCI CIs over by the end of the year (hopefully earlier)
13:34:51 <lbeliveau> cool !
13:34:57 <lbeliveau> keep us posted
13:35:04 <wznoinsk> lbeliveau: they've tried with containers before, didn't get that working unfortunately that's why I'm trying with baremetal for start
13:35:12 <wznoinsk> will sure do
13:35:14 <lbeliveau> #topic Patches for Subteam Review
13:35:27 <lbeliveau> anybody has patches for review that we're not aware of ?
13:36:05 <sfinucan> none on my end
13:36:37 <lbeliveau> me neither, just need cores to review those that are pending
13:36:41 <lbeliveau> #topic Specs for Review
13:36:54 <lbeliveau> so lots of activity there, a few specs under review
13:37:17 <lbeliveau> anybody wants to say something about specs ?
13:37:35 <lbeliveau> we should plan a discussion at the summit
13:37:47 <sfinucan> lbeliveau: You've seen the two NUMA-PCI specs, yes?
13:38:23 <lbeliveau> yes, I've been commenting on a few recently
13:38:46 <sfinucan> lbeliveau: Then other than that, no, I'm good
13:38:56 <sfinucan> Yes to meeting at the summit tho
13:39:16 <vladikr> lbeliveau, I've also submitted  https://review.openstack.org/381370 yesterday
13:40:02 <lbeliveau> vladikr: will have a look
13:40:38 <sfinucan> vladikr: ditto
13:40:46 <vladikr> sfinucan, thanks
13:41:24 <lbeliveau> not sure how this is usually working with cores, maybe we should push for a few of them, not sure we can commit on all of them
13:41:46 <wznoinsk> I have a feeling we're missing some people from the meeting today due to the confusion
13:41:48 <wznoinsk> sry, lost connection
13:42:04 <lbeliveau> yes, next week will be better :)
13:42:16 <lbeliveau> #topic Open discussion
13:42:30 <lbeliveau> anybody needs to bring something on the table ?
13:42:34 <sfinucan> So I just have this bug to point out https://bugs.launchpad.net/nova/+bug/1628301
13:42:35 <openstack> Launchpad bug 1628301 in OpenStack Compute (nova) "SR-IOV not working in Mitaka and Intel X series NIC" [Undecided,Incomplete]
13:42:56 <lbeliveau> ok was not aware of this one, let me check
13:43:03 <sfinucan> mriedem raised it last week. I tried setting up a virtual SR-IOV env but haven't succeeded yet
13:43:07 <sfinucan> lbeliveau: Cool
13:43:38 <lbeliveau> looks like config issues at first sight
13:44:16 <lbeliveau> ok I will comment on it a bit later
13:44:50 <lbeliveau> anything else ?
13:45:10 <sfinucan> not from me
13:45:29 <lbeliveau> thanks guy, we'll talk next week
13:45:34 <lbeliveau> #endmeeting