13:09:13 <pczesno> #startmeeting pci-passthrough 13:09:14 <openstack> Meeting started Tue Jun 23 13:09:13 2015 UTC and is due to finish in 60 minutes. The chair is pczesno. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:09:15 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:09:18 <openstack> The meeting name has been set to 'pci_passthrough' 13:17:31 <pczesno> :) 13:17:31 <pczesno> there is not much on the agenda 13:17:31 <pczesno> shoudl we go straight to open discussion? 13:17:31 <moshele> pczesno: sure 13:17:31 <pczesno> moshele: so on the agent required thing 13:17:31 <moshele> pczesno: do you plan to response on https://review.openstack.org/#/c/135331/ it is look like johnthetubaguy is waiting for response 13:17:32 <pczesno> moshele: yes we will 13:17:32 <pczesno> moshele: did you have any luck talking with Jay about his resource templates thingy? 13:17:32 <moshele> I didn't talk to jaypips they are busy with the spec freeze review 13:17:32 <pczesno> moshele: i don't see any specs relating to that 13:17:32 <moshele> me too 13:17:32 <pczesno> moshele: ok, it seems the resource templates are prerequisite for this work 13:18:17 <moshele> yes 13:18:41 <pczesno> moshele: so it will take time to update the niantic drivers to support setting state 13:19:31 <moshele> for now just announce deprecation for L we will remove it in M 13:19:54 <moshele> is that ok with you? 13:21:19 <pczesno> moshele: i wonder if we can modify the agent to not fail ont the setting state not supported msg from ip 13:22:09 <pczesno> moshele: this way we could use the agent with older nics 13:22:54 <moshele> ok, I am fine with that 13:23:22 <pczesno> moshele: if that's the case we could make it mandatory in Liberty 13:25:44 <moshele> I am not sure if that is the correct process, usually we need to declare deprecation and then remove it in the next release 13:26:40 <baoli> pczesno: moshele: sorry about losing the context. can you brief me a little bit on the subject? 13:26:55 <pczesno> moshele: yes, that's true, i don't mind either 13:27:25 <moshele> agent_required parameter in sriovnicswitch 13:27:31 <pczesno> baoli: the sriovnicswith mech driver can run with agent and without, moshele whant's to make it required 13:27:37 <pczesno> baoli: to support QOS 13:28:30 <baoli> pczesno, moshele: thanks. So it's only related to the specific mech driver. 13:28:45 <pczesno> baoli: yes, no impact to cisco :) 13:28:45 <moshele> yes 13:28:55 <baoli> cool 13:29:18 <moshele> by the way this is the Qos SR-IOV bp https://review.openstack.org/#/c/187895/ which is part of the Qos feature 13:29:29 <pczesno> moshele: so will the agent stay in neutron tree? 13:29:51 <moshele> currently yes 13:30:19 <pczesno> moshele: i saw that, you want to support both new and old ip api:? 13:30:31 <moshele> yes 13:31:20 <pczesno> moshele: it seems to me we will have to handle nic driver feature parity somehow 13:31:38 <baoli> pczesno: what's ip api? 13:31:45 <pczesno> moshele: like with the max_burst_kbps 13:32:10 <pczesno> baoli: the ip utility from iproute2 package 13:32:44 <pczesno> baoli: via ip utility we set the qos parameters on VF's 13:33:27 <moshele> I know I ask the driver team to add max_burst_kbps support I am not sure when it happen, any way the Qos plugin should handle unsupported attributes in Qos rules 13:34:05 <baoli> pczesno: I see. thanks. 13:35:12 <pczesno> moshele: what do you mean by "the Qos plugin shoudl handle unsupported attr in Qos rules"? 13:38:53 <baoli> moshele, thanks for letting me know multiple PF's are on mlnx's radar. 13:39:01 <moshele> I mean we are working on validation mechanism in Qos to validate what are the supported Qos rule and attribute per mechanism driver 13:39:07 <moshele> baoli: sure :) 13:40:28 <pczesno> moshele: but how can the MD know ? will it be a config option for the MD? 13:41:28 <moshele> pczesno: it still open issue hopefully we can close it in the Qos sprint next week 13:42:04 <moshele> I will update you on the next meeting I guess 13:42:06 <pczesno> moshele: ok, do you know if it will be possible to attend the sprint remotely? 13:42:13 <moshele> yes 13:42:34 <moshele> you need to install redhat app to connect 13:42:40 <moshele> you can to talk ajo 13:42:48 <pczesno> moshele: cool, thanks 13:42:57 <pczesno> moshele: you will be there in person, right? 13:43:09 <moshele> yes 13:44:07 <pczesno> anything else to discuss? 13:44:23 <moshele> baoli: I added a spec for the whitelist https://review.openstack.org/#/c/179577/ 13:45:28 <pczesno> gerrit is down again 13:45:57 <baoli> moshele: can't open it either 13:47:19 <moshele> baoli: I updated the syntax as danpb suggested to support regex 13:47:49 <baoli> moshele: I guess this is related to the patch you were working on ealier. 13:47:50 <moshele> baoli: just see that it make sene to you 13:47:51 <pczesno> moshele: it's already approved, right? 13:48:05 <moshele> it as 2 +2 13:48:15 <moshele> but not approved yet 13:48:50 <moshele> baoli: yes 13:49:33 <moshele> pczesno: are you going to talk to jaypipes regarding the resource template 13:49:52 <pczesno> moshele: yes, i will 13:50:24 <baoli> moshele: cool, will take a look, and hopefully the patch will land soon. 13:52:37 <moshele> ok I think we are finished, right? 13:52:43 <pczesno> yes 13:52:44 <baoli> moshele: one question. This is not to support the general regex, but a specific syntax of range, right? 13:53:46 <moshele> no danpb wanted regex but the old syntax will still exist 13:53:58 <pczesno> #endmeeting pci-passthrough