03:00:44 <Sundar> #startmeeting openstack-cyborg
03:00:45 <openstack> Meeting started Thu Apr  2 03:00:44 2020 UTC and is due to finish in 60 minutes.  The chair is Sundar. Information about MeetBot at http://wiki.debian.org/MeetBot.
03:00:46 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
03:00:48 <openstack> The meeting name has been set to 'openstack_cyborg'
03:00:51 <Sundar> Hi all
03:01:08 <Sundar> #topic Roll call
03:01:12 <Sundar> o/ Sundar
03:01:20 <s_shogo> Hi all
03:01:20 <brinzhang_> o/
03:01:48 <xinranwang> Hi all
03:01:55 <Sundar> Just the 3 of us? May be we'll wait a couple of min.
03:02:01 <chenke> Hi all
03:02:23 <chenke> o/
03:02:23 <Yumeng> hi chenke, hi all
03:02:36 <Sundar> Hi all
03:02:43 <Sundar> Good, let's get started
03:02:49 <Sundar> First the good news!
03:03:15 <Sundar> The Nova integration is essentially done! https://review.opendev.org/#/q/project:openstack/nova+bp/nova-cyborg-interaction+(status:merged%20OR%20status:open)
03:03:32 <s_shogo> Congrats !!!
03:03:45 <Sundar> I am just finsihing up on the release notes and a patch for a corner case.
03:03:47 <Yumeng> bravo!! such a great effort!
03:04:08 <chenke> Great!
03:04:31 <brinzhang_> nice job
03:04:42 <xinranwang> Congrats
03:04:49 <xinranwang> Thanks for your efforts
03:04:57 <Sundar> It has been a long journey -- we started in Rocky release, then we had the cyborg-nova pilot code, got the spec merged, and this patch series has been going for  about a year.
03:05:08 <Sundar> Thanks :)  I am relieved.
03:06:29 <Sundar> I am also happy that many companies plan to use Cyborg and, in fact, have already been using it.
03:06:41 <chenke> a big step.
03:06:54 <Sundar> Some telco companies in China also look forward to Cyborg with some networking support.
03:07:08 <Yumeng> Yes, it's not easy. Great news for us!
03:07:52 <Sundar> The best part is, all of you are developers and also users. That is the ideal situation: users guide the development of the project.
03:07:59 <brinzhang_> We are using now
03:09:42 <Yumeng> brinzhang_: great! so does ZTE.
03:10:20 <Sundar> The part that is not so good for me is that I have moved onto another role. So, I can spend only my spare time outside of work on Cyborg. I will continue to review patches and conduct the meetings (unless somebody else wants to take that over). But I may not have time to write  lots of new patches.
03:10:57 <Sundar> It is a bit tough for me to step aside from Cyborg, but that is the way forward.
03:11:44 <brinzhang_> Sundar: That's enough, I know that will take your free time
03:11:50 <brinzhang_> thanks
03:12:08 <Sundar> I will complete the 4 open patches here: https://review.opendev.org/#/q/status:open+project:openstack/cyborg+branch:master+owner:Sundar
03:12:37 <Sundar> Re. https://review.opendev.org/684456, I think xinranwang and I agree that it is not necessary now.
03:13:13 <Sundar> Is there any other important piece left for me, apart from completing the PTL duties for U?
03:15:28 <s_shogo> >https://review.opendev.org/684456  I approve that, to install OPAE in other way is ok, IMO.
03:15:47 <Sundar> s_shogo: Good, thanks.
03:16:04 <Sundar> If any of you think of something I have missed, you can LMK later too.
03:16:43 <Sundar> For the rest of this meeting, we should probably prioritize the patches that need to merge before Apr 9.
03:16:54 <Sundar> #topic Patches before Apr 9
03:16:56 <Yumeng> Thank you Sundar! You've contributed many to Cyborg. It's hard to say en end,but you've done a good begining and good end.
03:17:16 <Sundar> Thanks, Yumeng
03:17:34 <Sundar> https://review.opendev.org/#/q/status:open+project:openstack/cyborg+branch:master
03:18:30 <chenke> I will review your patches as soon as possble.
03:18:45 <Sundar> The most important patches are the ones I am working on. Umm, no, just kidding :)
03:18:49 <Yumeng> I will also review them.
03:18:56 <Sundar> It is the microversion patch: https://review.opendev.org/696860
03:19:05 <Yumeng> haha.
03:19:19 <xinranwang> Yes, please review it when you got time.
03:19:28 <Yumeng> will do.
03:19:56 <Sundar> What else would we consider highest priority?
03:20:54 <xinranwang> I also submit a patch completing v2 api doc
03:21:19 <xinranwang> https://review.opendev.org/715980  please review this too :)
03:21:57 <chenke> Hi sundar. I found a bug when deleing device profile before a vm(with fake-device) was deleted.
03:22:07 <Sundar> xinranwang: I agree it is important but we can also do it after Apr 9, if we need to.
03:22:19 <Sundar> chenke: Interesting, what's the bug?
03:22:48 <chenke> That's because the foriegn key constraint.
03:22:50 <Sundar> The original idea was to allow the device profile deletion when the VM is running, but we don't allow it today
03:23:03 <Sundar> chenke: Yes, the ARQs refer to the device profile entry
03:23:29 <Sundar> But they don't need to -- we copy the device profile groups into the ARQs, so we can remove that link
03:23:36 <chenke> So, is it reasonable to throw this error?
03:23:55 <chenke> No. Just copy the device profile groups is not enough.
03:24:24 <chenke> the error like this: `cyborg`.`extended_accelerator_requests`, CONSTRAINT `extended_accelerator_requests_ibfk_1` FOREIGN KEY (`device_profile_id`)
03:25:22 <Sundar> chenke: For now, the shortest improvement would be to block it at the API level rather than throw an error
03:26:03 <Sundar> For the future, we may need more changes to use the device profile groups copied into the ARQ
03:26:13 <Sundar> However, there is a fundamental problem.
03:26:47 <Sundar> On the Nova side, the operator may do a hard reboot or other ops for the VM anytime, and those ops will delete the old ARQs, and create new ones from the device profile name.
03:26:59 <Sundar> So, if the devic eprofile has been deleted, that is a problem.
03:29:30 <Sundar> Any thoughts?
03:29:46 <chenke> Ok.
03:30:04 <chenke> So we should keep the device profile.
03:30:17 <Sundar> So, let us block the deletion at the device profile API level rather than throw an error
03:30:26 <chenke> Agree.
03:31:00 <Sundar> Cool.
03:31:34 <Sundar> BTW, I will be proposing a patch shortly for CYborg's highlights for the U release. I'll share that link.
03:31:42 <Yumeng> yes ,agree now we just blocl the API level. but later maybe we can do like nova flavor. create a copy in nova for ops like reboot
03:32:54 <Yumeng> but that needs nova approval..   :-
03:33:13 <Sundar> Yumeng: The instance object has a copy of the flavor inside it, but does not have the ARQs. It has to call Cyborg with the device profile name to create the ARQ for some instance ops.
03:33:32 <Sundar> Yes, it needs more changes in Nova
03:33:57 <Sundar> We may need to make ARQs into an OVO, so it can be stored in the instance object.
03:34:10 <Sundar> Plus, db changes in Nova
03:34:26 <Yumeng> yes.
03:35:10 <chenke> actually that's hard.
03:36:04 <Sundar> Yup
03:36:19 <Sundar> That's why I didn;t do it :)
03:37:42 <chenke> Ok.  I have no other needed to be discussed.
03:38:14 <Sundar> Anything else, folks?
03:38:35 <brinzhang_> one patch
03:39:26 <brinzhang_> Sundar, chenke, s_shogo: I think this is why py37 timeout previous https://review.opendev.org/#/c/716796/1/tox.ini
03:39:59 <brinzhang_> the ignore_basepython_conflict flag introduced in 3.1.1, before we are in 2.0
03:40:32 <brinzhang_> that I upgrade this to 3.1.1 to adopt python3
03:41:18 <chenke> good catch.
03:41:29 <Sundar> Yea, makes sense to me.
03:41:46 <s_shogo> good
03:42:18 <Sundar> #topic AoB
03:42:39 <Sundar> Anything else?
03:43:20 <brinzhang_> Nothing else, many thing talked in wechat, and have done of mine ^^
03:43:56 <Sundar> Thanks a lot, folks. Have a good day, and stay safe away from the virus!
03:44:00 <Sundar> #endmeeting