13:01:41 <baoli> #startmeeting PCI Passthrough
13:01:42 <openstack> Meeting started Tue Sep 16 13:01:41 2014 UTC and is due to finish in 60 minutes.  The chair is baoli. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:01:43 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:01:46 <openstack> The meeting name has been set to 'pci_passthrough'
13:01:55 <baoli> Hi, there
13:02:01 <irenab> Hi
13:03:04 <heyongli> hi
13:03:19 <heyongli> thank your hard work, baoli
13:03:32 <rpothier> hi
13:03:51 <baoli> So we made it by the end of last Friday
13:04:07 <heyongli> congratulation, every one
13:04:14 <irenab> baoli: Great news!
13:05:01 <baoli> yes indeed.
13:05:30 <baoli> Thanks everyone for helping out
13:06:17 <irenab> so now we should focus on testing and fixeng bugs
13:06:26 <baoli> Irenab, this morning, one of my coworkers run into problems with n1kv neutron plugin. Due to vnic_type not being set, he couldn't bring up VMs.
13:07:08 <irenab> baoli: is there any open bug for this, to look for details?
13:07:09 <baoli> I remember that we require that neutron plugins set the vnic_type for each port.
13:07:30 <baoli> irenab, not yet.
13:08:06 <irenab> if not set, its normal
13:08:08 <baoli> the question is we should open a bug for nova or for neutron, which depends on the question if every port should have an explicit vnic_type.
13:08:46 <irenab> is it for plugin or Ml2 Mechanism Driver?
13:09:02 <baoli> irenab, this is a monolithic plugin.
13:09:59 <irenab> For my understanding it should not be a problem, if vnic_type not set, it assumes normal. And for Monolitic Plugin there are no options for different vnic_types, right?
13:10:40 <irenab> baoli: we can try to troubleshoot it offiline
13:10:42 <BrianB_> baoli, monolithic is being depericated
13:11:08 <baoli> BrianB_, so n1kv monolithic plugin will be depreicated?
13:11:25 <BrianB_> baoli, yes
13:11:40 <baoli> Brian
13:11:51 <baoli> BrianB_, cool.
13:12:52 <heyongli> irenab, i recall you had a etherpad for next step, i cant find it, could you paste a link again?
13:13:25 <baoli> Irenab, itzik contacted me last week. Does he still have the problem?
13:13:47 <baoli> Irenab, please open a bug if it's still an issue.
13:14:19 <irenab> heyongli: https://etherpad.openstack.org/p/kilo_sriov_pci_passthrough
13:14:33 <heyongli> thanks
13:14:35 <irenab> baoli: seems that macvtap is not working for hw_veb
13:15:10 <irenab> TRACE nova.compute.manager mlibvirtError: unsupported configuration: an interface of type 'direct' is requesting a vlan tag, but that is not supported for this type of connection
13:15:33 <irenab> heyongli: did you test this?
13:16:30 <irenab> We have created wiki page: https://wiki.openstack.org/wiki/SR-IOV-Passthrough-For-Networking
13:16:45 <heyongli> not yet, i foucs other thing, i will check this
13:16:48 <irenab> for general overview, nova and neutron configs
13:17:01 <heyongli> great
13:17:10 <baoli> irenab, that's cool
13:17:18 <irenab> currently covers hw_veb MD, since cisco MD is not merged yet
13:17:43 <irenab> Please, review and update
13:17:53 <baoli> irenab, please open a bug to track that issue, then.
13:18:03 <irenab> baoli: sure, will do
13:18:24 <irenab> baoli: you mentioned resize operation. Should it work?
13:18:45 <heyongli> it's buggy now
13:19:01 <heyongli> i will open bug if there is not
13:19:07 <baoli> heyongli, did you try it with the officail juno ?
13:19:24 <irenab> heyongli: should it ever work if required rescheduling to some other node?
13:19:30 <baoli> heyongli, I opened a bug for it already.
13:19:38 <heyongli> cool
13:19:47 <heyongli> it's now work for a while
13:20:21 <irenab> heyongli: for both resize on same Server and another Server?
13:20:28 <baoli> heyongli, you mean it used to work for a while?
13:20:52 <heyongli> yeah, i remember , long time ago, it's work
13:21:15 <heyongli> resize to another server is lack of testing
13:22:22 <irenab> I think if there are some limitation for certain operations, it should be documented
13:22:37 <baoli> https://bugs.launchpad.net/nova/+bug/1368201
13:22:38 <uvirtbot> Launchpad bug 1368201 in nova "resize with PCI devices doesn't work" [Low,Confirmed]
13:22:50 <irenab> baoli: thanks
13:23:20 <heyongli> boali, got it
13:23:21 <baoli> it took me a couple of days to test it, and in the process my compute node was hosed,
13:24:08 <baoli> And it took me a while to figure out what's wrong and recover it. Still I need to make a little change in the nova code to make it work again.
13:24:09 <heyongli> sorry for that, i check our internal ci find it not work, according our meeting ar
13:24:49 <irenab> baoli: Did you test it on same Server or between two Servers?
13:25:02 <baoli> For some reason, libvirt doesn't like the memballoon config in the domain xml although it has the same libvirt version as other servers in our lab.
13:25:29 <baoli> irenab, I have one compute node only.
13:25:46 <heyongli> i suspect it broken for a while, during icehouse, pci is stuck
13:27:15 <irenab> I am not sure it will work between nodes by design, but aso do not hve setup for now...
13:28:17 <baoli> would 'resize' move VMs across nodes? It's basically a cold migration with different flavor.
13:28:24 <irenab> another question regarding documentation in openstack admin guides
13:28:41 <heyongli> irenab, between servers , i need check the design,
13:28:43 <irenab> is there a documentation bug for this?
13:28:54 <baoli> irenab, that's right. We should open one for it.
13:29:23 <baoli> irenab, maybe the wiki should suffice?
13:29:28 <irenab> baoli: can you please open one for nova and 'll proceed with one for neutron
13:29:59 <irenab> baoli: not sure, how is it for general PCI?
13:30:28 <heyongli> between servers? irenab
13:30:30 <baoli> heyongli, do we have admin doc for general PCI?
13:30:51 <heyongli> i and yunhong not write the doc
13:31:07 <heyongli> we just setup a wiki
13:31:18 <irenab> heyongli: queocumentation
13:31:47 <irenab> The question if we need to make it part of openstack admin guide
13:32:20 <heyongli> it should
13:33:05 <baoli> who usually write the text for the admin doc?
13:33:23 <heyongli> some people focus on documetation
13:33:25 <baoli> I mean it's the developer or do we have tec writer for it?
13:33:54 <heyongli> most doc might not wirte by dev, but just guess
13:34:07 <irenab> I think that the developer of the feature, but tech writer reviews
13:34:24 <irenab> at least it was so for vnic_type that I added in Icehouse
13:34:42 <baoli> irenab, did you add it to the neutron admin doc?
13:35:14 <irenab> it was added to API Reference doc
13:35:33 <baoli> but you did it yourself?
13:35:55 <irenab> The neutron doc guru (emagana) contacted me then to handle this
13:36:23 <baoli> so you came up with the text, and the doc guru added it.
13:36:50 <irenab> baoli: no, did the writing and patch for documentation
13:36:57 <irenab> by myself
13:37:20 <baoli> irenab, ok. I think the doc bugs may follow similar processes as bug fixes.
13:37:25 <irenab> once you add DocImpact in the patch commit, it opens documentation bug
13:37:47 <baoli> irenab, no one has asked about a DocImpact during the code review
13:38:51 <heyongli> should we open a doc bug now?
13:38:59 <irenab> yes, but probably need to open documentation bug  anyway
13:39:06 <baoli> So if we ever do this for nova, we need something for pci passthrough, and one section in it for sr-iov
13:39:37 <irenab> baoli: +1
13:39:56 <heyongli> agree
13:41:55 <baoli> shall we just open a bug for it now, and whoever has bandwidth can start working on it?
13:42:20 <irenab> baoli: Yes
13:43:16 <baoli> #note open a admin doc bug for PCI-passthrough and sr-iov
13:43:29 <irenab> Did you have a chance to look at the ehterpad for kilo features?
13:43:36 <baoli> ok, let's move on to kilo list
13:44:14 <baoli> I did. And I think it's a very good list.
13:44:27 <baoli> Let's add the 'resize' thing in the list too.
13:45:21 <itzikb> Hi, I have 2 questions regarding pci_passthrough_whitelist
13:45:27 <heyongli> i can not access the etherpad from home
13:45:33 <baoli> Another thing is that let's put them in two categories: one for nova, one for neutron, to make it clear. It seems that the majority work will be in Nova
13:46:08 <heyongli> and nova will enter tick toke mode, will it impact it?
13:46:16 <irenab> baoli: we can add 'project' for each item.
13:46:26 <baoli> ireanb, that sounds good
13:46:40 <irenab> baoli: I'll update
13:46:53 <baoli> heyongli, what is tick toke mode?
13:47:33 <irenab> Lets try for next week to focus on setting priority for each task
13:47:38 <heyongli> in one release nova might foucus on bug fix, limit the feature merge, will it be perform right now, i'm not sure
13:48:28 <baoli> heyongli, ok
13:48:57 <irenab> heyongli: I think we can prioritize and see if need Spec/Bug. Then depending on general nova development in Kilo, we caneither push it further or wait
13:49:26 <baoli> irenab, yes, let's do that next meeting.
13:49:42 <baoli> We can discuss existing bugs, and then go over the list
13:49:48 <irenab> baoli: agree
13:50:26 <irenab> Do you think we will need some design session at November Summit?
13:51:07 <baoli> irenab, I think so with the first several items on the list
13:51:55 <irenab> Then we need to hurry and discuss it asap to make some proposal before November
13:52:24 <irenab> Are you going to make it to Paris?
13:52:51 <baoli> irenab, I just need to get the visa, everything else is all set.
13:53:31 <irenab> baoli: great
13:54:07 <heyongli> i afraid can not go
13:54:20 <heyongli> some one else should be ther
13:54:34 <baoli> irenab, I guess you are going too, right?
13:55:06 <irenab> baoli: yes, still need to close travel logistics
13:56:14 <irenab> heyongli: Sorry you won't be there
13:56:37 <heyongli> me too
13:56:50 <baoli> heyongli, I guess it's not because of visa, right?
13:56:54 <heyongli> one of our guy should be there
13:56:57 <heyongli> no
13:57:00 <heyongli> baoli
13:57:21 <heyongli> usally we sent local guy go to meeting
13:57:57 <heyongli> attach port to vm seems kind of hotplug
13:58:22 <baoli> Hopefully, next time it's going to Beijing or Shanghai
13:58:41 <irenab> heyongli: exactly
13:58:51 <heyongli> that's great ,but china have hongkong already
13:59:15 <irenab> so let's come prepared for next week with items for kilo.
13:59:25 <heyongli> sure
13:59:32 <baoli> yep
14:00:04 <irenab> See you next week, have to switch to other meeting
14:00:08 <baoli> I'll open the admin doc bug, and check Irenab's wiki.
14:00:14 <heyongli> see you
14:00:16 <irenab> baoli: Thanks!
14:00:18 <baoli> #endmeeting