03:01:33 <Sundar> #startmeeting openstack-cyborg
03:01:35 <openstack> Meeting started Thu Sep 26 03:01:33 2019 UTC and is due to finish in 60 minutes.  The chair is Sundar. Information about MeetBot at http://wiki.debian.org/MeetBot.
03:01:36 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
03:01:38 <openstack> The meeting name has been set to 'openstack_cyborg'
03:01:41 <Sundar> Hi all
03:01:50 <Sundar> #topic Who's here
03:01:51 <shaohe_feng> good evening
03:01:57 <Sundar> o/
03:02:01 <s_shogo> Hi all
03:02:04 <shaohe_feng> o/
03:02:18 <chenke> o/
03:02:22 <chenke> Hi all.
03:02:50 <Yumeng> #info Yumeng
03:02:53 <Yumeng> hi all
03:03:10 <xinranwang> Hi all
03:03:16 <xinranwang> #info xinranwang
03:03:25 <s_shogo> #info s_shogo
03:03:39 <Sundar> Let's start with some good news. In Chenke's patch https://review.opendev.org/#/c/682565/, Eric asked to try 'strict_proxies=False'
03:03:47 <Sundar> I tried and it works for me at least :)
03:03:53 <chenke> Wow
03:03:59 <chenke> Sound great.
03:04:20 <Sundar> If chenke can resubmit with that change, I'll trigger the tempest check
03:04:39 <chenke> I had done it.
03:04:57 <chenke> And triggered the tempest check.
03:05:36 <Sundar> https://review.opendev.org/#/c/682565/15/cyborg/common/utils.py@149 still has i, right?
03:05:39 <Sundar> *it
03:06:00 <Sundar> I set that to False
03:06:44 <chenke> The eric's advice is to see it to False, pls see Line 130 the default value.
03:07:02 <xinranwang> I discussed with colleague who has experience on CI, there is no other solution except add "depend on" in each patch.
03:07:30 <xinranwang> it is better to merge nova patches as soon as possible.
03:08:12 <Sundar> chenke: I see.
03:08:31 <Sundar> Ok. Hope that does the trick.
03:08:39 <chenke> Hope so.
03:08:45 <Yumeng> agree with Xinran
03:08:57 <Yumeng> we have this to test related patches :https://review.opendev.org/#/c/670999/
03:09:17 <Yumeng> to test nova-cyborg-integration related patches
03:09:44 <chenke> Agree with Xinran. Now Yumengs' patch is a better solution so far.
03:09:44 <Sundar> It is not up to us to get Nova patches merged. There are challenge sin getting reviewer attention, not only for Cyborg, but for other subjects too.
03:10:59 <Sundar> Are we saying we should put Depends-On for a Nova patch in every Cyborg patch?
03:11:47 <xinranwang> yes
03:12:56 <Sundar> All agree?
03:12:59 <chenke> but this will have too many dependencies.
03:13:56 <Sundar> Too many dependencies on Nova code?
03:14:14 <chenke> Yes.
03:15:02 <xinranwang> 1. need keep nova patch updated all the time to avoid merge conflict.
03:15:02 <xinranwang> 2. need add "depend on" in every patch to trigger correct tempest test.
03:15:02 <Sundar> Yea. May be we can do it judiciously based on whether the patch has Nova interaction.
03:15:10 <chenke> But it is also a good way to fully test our cyborg code.
03:15:13 <chenke> Agree
03:16:23 <Sundar> I am the only one maintaining those patches. Nova moves fast and breakages happen often. I cannot commit to keep doing that all the time, unless we are getting Nova reviewers' attention.
03:16:48 <shaohe_feng> something wrong with my net
03:17:04 <chenke> welcome back.
03:17:11 <xinranwang> It's better to get nova folks attention.
03:17:28 <xinranwang> Longer we wait, more efforts it takes.
03:17:29 <Sundar> That's what we tried to do for entire Train cycle
03:17:36 <Yumeng> yes. so better to have a successful tempest test at nova side and get merg asap, then we add more features in cyborg
03:19:23 <Yumeng> Sundar: I can help to maintain and fix patches, if you need help.
03:21:08 <Sundar> Let's be clear. Now we put Depends-On *from* Nova to the Cyborg patch that we want to test, i..e. chenke's patch. We can't do that for every patch.
03:21:52 <chenke> Agree.
03:22:29 <chenke> just some special patch should do like this.
03:22:29 <xinranwang> +1
03:22:37 <s_shogo> +1
03:22:47 <Yumeng> +1
03:24:16 <Sundar> Makes sense. Any patch that touches Nova or other services or any significant patch can be tested this way. If/when we get Nova reviews early in U, we can make it a common practice.
03:24:39 <Sundar> All good with that?
03:25:47 <Sundar> Ok, I'll take previous +1s as yes :)
03:26:04 <Sundar> #topic RC status
03:26:26 <Sundar> https://review.opendev.org/#/c/684056/1 has been filed
03:27:03 <Sundar> I am saying we have a few potential RC blockers.
03:27:21 <Sundar> •	Privsep: https://review.opendev.org/673957 : affects security
03:27:28 <Sundar> •	Privsep for FPGA prog: https://review.opendev.org/683733 : affects security
03:27:37 <Sundar> •	V1 API: https://review.opendev.org/684193 : affects backwards compatibility
03:27:56 <Sundar> Others are important for us, and we should merge as much of them as we can.
03:28:04 <shaohe_feng> https://review.opendev.org/#/c/681005/
03:28:32 <openstackgerrit> chenker proposed openstack/cyborg master: Fix the hardcoding of user role using sdk_adapter approach  https://review.opendev.org/682565
03:29:20 <Sundar> shaohe_feng, an RC blocker has  a high bar -- we have to explain why the entire release should be held up till we fix that
03:30:10 <Sundar> We should continue to review your patch and try to merge it. May be this RC may not be the last one.
03:30:19 <shaohe_feng> One questions about:  sdk_adapter patch. I remember we did user sdk_adapter in V1, why we change it to connection
03:30:31 <shaohe_feng> and change back to sdk_adapter again?
03:31:40 <chenke> nova recommand us to use sdk now.
03:31:45 <Sundar> We used ksa_adapter in v1. We were asked to move to sdk_adapter as a new method and then to COnnection as the latest one. There are some bugs apparently, so we have settled on what works.
03:32:00 <Sundar> chenke: +1
03:32:45 <shaohe_feng> what's the pron of sdk?
03:34:16 <shaohe_feng> O, I remember xinranwang use sdk_adapter ?
03:34:27 <shaohe_feng> why not give up?
03:34:39 <shaohe_feng> then sdk_adapter again?
03:35:42 <shaohe_feng> pros and cons?
03:36:16 <Sundar> You can ask Eric in that patch. Nova has some requirements and considerable interest in this area.
03:36:33 <shaohe_feng> OK.
03:37:16 <Sundar> Are there any other patches that are potential RC blockers, i.e, have security or critical implications, so we should hold up the release?
03:37:29 <shaohe_feng> maybe we can do more investigation on it. but not now
03:37:37 <shaohe_feng> now we have more things to do.
03:39:10 <Sundar> Sure. If nothing else, let us please focus on reviewing the ones above first, and as much of the rest as we can. It is possible that there may be future RCs, so they may still make it
03:40:00 <Sundar> Any other comments or thoughts?
03:40:01 <Yumeng> https://review.opendev.org/#/c/679406/ this one please
03:40:15 <Yumeng> conductor writes device_profile
03:40:33 <Sundar> Yumeng: sure. IMHO, it is mostly there. I'll try it out tonight and review it
03:40:46 <Yumeng> just updated.
03:40:55 <Yumeng> thanks!
03:41:33 <Sundar> Shaohe, have you tried your patch with the Nova patches?
03:41:49 <shaohe_feng> yes.
03:41:55 <Sundar> Great
03:41:59 <shaohe_feng> xinranwang also help to do it.
03:42:09 <Sundar> Thanks to both of you
03:42:28 <Sundar> Anything else, folks?
03:42:35 <shaohe_feng> actually, I did not change your code logical
03:42:51 <shaohe_feng> just put them in thread pool as async job. :)
03:43:10 <Sundar> Cool, np even if you changed it, as long as it works :)
03:43:23 <xinranwang> This patch is almost good for me
03:44:01 <Sundar> almost?
03:44:43 <Sundar> Anyway, we can continue the review in Gerrit.
03:44:55 <xinranwang> “Almost”means need other’s review as well
03:45:17 <chenke> About this patch:https://review.opendev.org/#/c/684193/
03:45:33 <chenke> I agree with xinran's idea.
03:45:53 <Sundar> Do you want to see 6 endpoints with devstack? :)
03:46:05 <Sundar> v1 is end-of-life.
03:46:06 <chenke> no
03:46:11 <chenke> We don't need to enable v1.
03:46:14 <Sundar> After Train, we can remove it early in U
03:47:37 <Sundar> Can you explain further, chenke? v1 is deprecated but supported in this release. So, we do the bare minimum to keep it alive, but don't bother develoeprs with legacy endpoints
03:48:20 <Sundar> We can remove v1 altogether early in U, and even these devstack comments will go away
03:49:16 <Sundar> Anybody disagree?
03:49:16 <chenke> Sorry, I misunderstood this meaning.
03:49:24 <chenke> just go on.
03:49:35 <Sundar> You are ok with that patch as is?
03:50:34 <Sundar> Anyway, let's continue the review in Gerrit
03:50:39 <chenke> +1
03:50:55 <Sundar> If nothing else, we can call it a day (or night or afternoon)
03:51:36 <Sundar> Have a good day, everyone! Bye for now. (I'll bother you on WeChat after this -- JK)
03:51:42 <Sundar> #endmeeting