Thursday, 2020-03-05

*** brinzhang_ has joined #openstack-cyborg00:07
*** brinzhang has quit IRC00:11
*** brinzhang has joined #openstack-cyborg00:12
*** brinzhang_ has quit IRC00:15
*** igordc has quit IRC00:33
*** igordc has joined #openstack-cyborg00:53
*** chenke has joined #openstack-cyborg01:13
*** openstackgerrit has joined #openstack-cyborg01:44
openstackgerritchenker proposed openstack/cyborg master: Remove V1 API and related codes
*** brinzhang_ has joined #openstack-cyborg01:49
*** brinzhang_ has quit IRC01:50
*** brinzhang_ has joined #openstack-cyborg01:51
*** brinzhang has quit IRC01:52
*** igordc has quit IRC02:07
*** xinranwang has joined #openstack-cyborg02:42
*** Sundar has joined #openstack-cyborg02:46
*** brinzhang has joined #openstack-cyborg02:50
brinzhangSundar: ask a question
Sundarbrinzhang: Ok, looking at it02:53
*** brinzhang_ has quit IRC02:53
brinzhangSundar: "if getattr(self.bind, "is_job", False) and need_job is not False:" if need_job=False, is it what would you want?02:53
SundarWhat is the question?02:53
brinzhangSundar: check_extra_job(context, deployable) this just only return  True or False? or there is other volue?02:54
brinzhangDou you accept current changes?02:55
brinzhang"need_job is not False" has the different value when need_job init None or False02:56
*** Yumeng has joined #openstack-cyborg02:57
SundarIMHO, need_job should be boolean (True/False). The check should be: "if need_job: ..."02:58
SundarHaving a possible value of None just complicates other code that has to check for None.02:59
SundarSo, I agree with this change.03:00
brinzhangso "if getattr(self.bind, "is_job", False) and need_job is not False:" should change to "if getattr(self.bind, "is_job", False) and need_job:"03:00
SundarYes, agree03:01
chenke need_job or need_job is not False both ok.03:01
brinzhangAnd I think we need add some UT for this, now the change dont run some UT03:01
Sundarchenke: Sure. But shorter is better.03:01
SundarLooks like we have quite a few people. So, let's get started.03:02
Sundar#startmeeting openstack-cyborg03:02
openstackMeeting started Thu Mar  5 03:02:31 2020 UTC and is due to finish in 60 minutes.  The chair is Sundar. Information about MeetBot at
openstackUseful Commands: #action #agreed #help #info #idea #link #topic #startvote.03:02
*** openstack changes topic to " (Meeting topic: openstack-cyborg)"03:02
openstackThe meeting name has been set to 'openstack_cyborg'03:02
Sundar#topic Agenda03:02
*** openstack changes topic to "Agenda (Meeting topic: openstack-cyborg)"03:02
chenke#info chenke03:02
Sundar#info Sundar03:03
Li_Liu#info Li_Liu03:03
brinzhang#info brinzhang03:03
Yumeng#info Yumeng03:03
xinranwang#info xinranwang03:04
SundarWhat do people have to discuss today?03:04
*** shaohe_feng has joined #openstack-cyborg03:04
brinzhangAbout remove  v1 API03:05
SundarBTW, the time for PTG reservation is up. I didn't hear confirmation that even 3 people plan to go to Vancouver, SO I haven't confirmed anything.03:05
shaohe_fengencourage developers to access the PTG03:07
shaohe_fengif possible.03:07
chenkeV1's API is now worthless, and v2 can also be seen. v2's deployable api, Shogo is also doing it, I think it can be removed. Even if the v1 api is retained, the call may not necessarily be as expected.
SundarTo have focused discussion on V release goals and plans, we could either do it in IRC and etherpads or in openstack-discuss ML. If we do a Zoom call.  or WeChat discussion, they should be posted on IRC or ML. Otherwise, other projects will not have know what we do.03:10
Sundarbrnzhang, chenke: Agree v1 API can be removed.03:10
Sundarbrinzhang: ^03:10
brinzhangSundar, but in v2 we dont have the deployable, right?03:11
brinzhangif we remove now, I think it's too early to do this.03:11
brinzhangIMO, we should done all things in v2, then to deal with the v103:12
Sundarbrinzhang: There is a deployable object in v2, but it means something else than the v1 deployable.03:13
SundarI think we plan to have APIs for patching deployables in v2.03:13
brinzhangSundar: sure03:14
brinzhangIf all agree to remove now, I am ok too.03:14
Sundarbrinzhang: I think it is better to remove api/controllers/v1/ and add api/controllers/v2/deployables.py03:15
chenkeshogo is doing this03:15
brinzhangSundar: I also agree, but I just think it's earlier03:16
*** s_shogo has joined #openstack-cyborg03:16
Sundarbrinzhang: I see. What should we wait for, before we remove the v1 API?03:16
brinzhangI think so03:16
chenkedo you use v1?03:17
brinzhangchenke, Sundar: Need a process here, right?03:19
Yumengbrinzhang: Brin, if I didn't remember it wrong, we've made an agreement Train version that we make an deprecation anounce and remove the code in U release.03:19
brinzhangYumeng, Ah, if so, I think I missed this message, ignore my comments03:21
YumengBut I forgot where the discussion was happen.03:21
SundarYes, we followed the process of deprecating in Stein itself and reinforced the deprecation in Train.03:21
Yumengyes, it's safe to remove now. Don't worry.03:21
SundarYumeng: We have discussed this in IRC before (I need to look up when). But our API docs clearly say v1 is deprecated, both in Stein and Train.03:22
Yumengok. good that we have announcement!03:23
brinzhangSundar: that should an ML or etherpad to recore this aggrement, otherwise will have different opinion03:24
brinzhangAnd we should add that links to the commit message, that reviewer can get more info03:24
chenkeOk. I will update message03:25
brinzhangSundar, chenker, Yumeng: ok, we can continue next topic03:26
Sundarbrinzhang: The above link is published already. I have said this to Nova developers in ML too, not sure when.03:26
SundarFor example, Sean mooney and others are testing Nova with Cyborg v2. No expectation of v1 support.03:27
Sundar#topic PTG03:27
*** openstack changes topic to "PTG (Meeting topic: openstack-cyborg)"03:27
SundarTo reinforce what I said: To have focused discussion on V release goals and plans, we could either do it in IRC and etherpads or in openstack-discuss ML. If we do a Zoom call.  or WeChat discussion, they should be posted on IRC or ML. Otherwise, other projects will not have know what we do.03:28
SundarDoes anybody disagree?03:29
xinranwangWe can create a ether pad03:29
xinranwangTo list the goals03:29
SundarOk, so everybody agrees to this ^, right?03:31
Yumengagree.  etherpad, zoom are always good.03:31
brinzhangagree, wWe can publish content on etherpad regularly03:31
brinzhangsend that info to ML03:31
Sundar#agree V release planning process above ^03:32
SundarThanks, all03:32
SundarAny other topic before I move to Nova and related stuff?03:32
YumengSundar, xinranwang, chenke: I will abandon this:
Yumengbut we still need to update the valid controlpath_id_type values here, do you agree?03:33
Yumengwe should remove "MDEV" in cureent valid values03:34
SundarYumeng: agree03:34
Sundarto both03:35
Yumengok. great! will do.03:35
Yumengok. thanks all!03:37
SundarOk, on the Nova series: it is a bit mixed news. One patch got merged, one more got +2. But we are still spinning on UT details. Sometimes seems like too much detail or nits.03:38
SundarI was hoping to wrap this up by Feb. I have offers for a next role, but yet to decide. But  I will be moving to a different role in March. So, I am looking to wrap it up one way or the other.03:40
Yumengwow, congrats!03:41
chenkeSundar congratulate!03:41
s_shogocongrats! >> the merged patch03:44
SundarThanks, mixed feelings for me, after putting so much effort into Cyborg.03:44
SundarCan one of you volunteer to be an additional PTL after I move on?03:45
brinzhangSundar, congrats! That is a new beginning and it will be brilliant.03:45
SundarThanks, brinzhang.03:45
shaohe_fengSo what‘s your next role?03:45
shaohe_fengstill about cloud related work?03:46
chenkeYes, you spend a lot of energy on cyborg, about 16 hours a day. Good luck cares for those who work hard.03:46
SundarNot decided on next role, yes still about cloud,03:47
Sundarchenke: I was doing CYborg plus some Kubernetes work till about a month ago03:48
SundarThat's why it took long hours03:48
chenkeGot it.03:49
xinranwang Thanks for your great efforts!03:49
SundarThanks, xinranwang03:50
SundarOK, maybe you all can discuss offline on additional PTL. We still have time.03:50
YumengSundar: I can help with the rest work after you move on. But not sure if we need more volunteers. we can discuss the tasks and seperate them.03:50
shaohe_fengkubernets is popular in cloud.03:51
SundarYumeng: Ok. We need to inform the OpenStack Foundation so that they can send all communication to two people, not just me alone.03:51
SundarThat's all, folks. Anything else before we wrap up?03:52
shaohe_fengcan anyone who attend PTG can tell something about Cyborg and Neutron integration?03:53
shaohe_fengon smartnic03:53
shaohe_fengI have look into it.03:54
Sundarshaohe_feng: I have rescinded (cancelled, taken back) my proposals for Cyborg Neutron integ. It would not be easy to drive a 3-way change across Nova, neutron and Cyborg.03:54
SundarThe next best approach would be to assume a static programming model, and use Nova to create the resource providers for smart NICs using the provider config (YAML-based) approach.03:55
shaohe_fengThe neutron Docs tells us, it support a new vif type smartnic03:55
*** links has joined #openstack-cyborg03:56
shaohe_fengthat's means cyborg should not manage these kind of accelerators?03:56
shaohe_fengI have  think it how to  support it.03:57
shaohe_fengany method, nova can not do many help03:57
shaohe_fengsome one want to know details about it.03:58
shaohe_fengor we can discuss offline03:58
Sundarshaohe_feng: First, to my knowledge, most folks do not plan to dynamically program smart NICs, i.e., change its image based on VM type, traffic type, workload, time of day, etc.03:58
SundarInstead they'll do it statically, maybe even offline when no VMs are present and the host is in maintenance mode03:58
SundarThe strength of CYborg is dynamic programming.03:59
shaohe_fengI know other company will support it. and maybe they are work in progress, not sure they will upstream.03:59
shaohe_fengthe cyborg is only for program?03:59
SundarOk, if you have concrete proposals, and plan to follow up with Nova/neutron community, along with other developers, that would be good.03:59
SundarYea, we can take this offline, or in the next IRC call.04:00
SundarI need to go now. And it is lunch time for you all :)04:00
shaohe_fengNo, it will do more not only program.04:00
SundarAnything else, folks?04:00
Yumengnothing from me.04:01
SundarHave a good day! Bye for now.04:02
*** openstack changes topic to "Pending patches (Meeting topic: openstack-cyborg)"04:02
openstackMeeting ended Thu Mar  5 04:02:03 2020 UTC.  Information about MeetBot at . (v 0.1.4)04:02
openstackMinutes (text):
*** s_shogo has quit IRC04:02
*** brinzhang has quit IRC04:36
*** brinzhang has joined #openstack-cyborg04:37
*** brinzhang has quit IRC04:37
*** brinzhang has joined #openstack-cyborg04:38
*** Sundar has quit IRC04:57
*** brinzhang has quit IRC05:52
openstackgerritMerged openstack/cyborg master: UT for job manager
openstackgerritMerged openstack/cyborg master: add testcase for check_bindings_result failed
*** efried has quit IRC06:07
*** efried1 has joined #openstack-cyborg06:07
openstackgerritMerged openstack/cyborg master: change default SimpleQueue to _PySimpleQueue for queue
openstackgerritMerged openstack/cyborg master: add testcases for async job bind
openstackgerritMerged openstack/cyborg master: Revert "Solve py37 timeout"
*** efried1 is now known as efried06:09
openstackgerritchenker proposed openstack/cyborg master: Fix minor error in extarq job
openstackgerritchenker proposed openstack/cyborg master: Remove V1 API and related codes
*** Yumeng has quit IRC06:36
openstackgerritYumengBao proposed openstack/cyborg master: Fix attach_handle_type of gpu device with improper value
*** efried has quit IRC07:22
*** efried has joined #openstack-cyborg07:22
*** efried has quit IRC07:23
*** efried has joined #openstack-cyborg07:23
*** xinranwang has quit IRC07:35
*** tetsuro has joined #openstack-cyborg07:44
*** brinzhang has joined #openstack-cyborg08:03
openstackgerritYumengBao proposed openstack/cyborg master: Fix GPU's improper cpid_type allowed value and attach_handle_type value
*** irclogbot_0 has quit IRC09:08
*** tetsuro has quit IRC09:29
*** irclogbot_0 has joined #openstack-cyborg09:43
*** openstackstatus has quit IRC09:45
openstackgerritYumengBao proposed openstack/cyborg master: Fix GPU's improper cpid_type allowed value and attach_handle_type value
*** shaohe_feng has quit IRC10:33
*** chenke has quit IRC11:02
*** tetsuro has joined #openstack-cyborg12:03
*** tetsuro has quit IRC12:06
*** brinzhang has quit IRC12:16
*** brinzhang has joined #openstack-cyborg13:31
*** efried is now known as efried_gone15:18
*** links has quit IRC15:27
*** brinzhang has quit IRC15:45
*** TxGirlGeek has joined #openstack-cyborg16:39
*** igordc has joined #openstack-cyborg17:22
*** TxGirlGeek has quit IRC22:35
*** brinzhang has joined #openstack-cyborg23:42
*** brinzhang_ has joined #openstack-cyborg23:57

Generated by 2.15.3 by Marius Gedminas - find it at!