Monday, 2018-08-13

*** efried1 has joined #openstack-cyborg02:26
*** efried has quit IRC02:28
*** efried1 is now known as efried02:28
*** openstackgerrit has quit IRC05:18
*** openstackgerrit has joined #openstack-cyborg08:27
openstackgerritMerged openstack/cyborg master: Update html_theme so cyborg doc page displays in standard page theme
openstackgerritMerged openstack/os-acc master: Update reno for stable/rocky
*** jaypipes has joined #openstack-cyborg13:26
openstackgerritMerged openstack/cyborg master: Follow the new PTI for document build
*** shaohe_feng has joined #openstack-cyborg13:59
*** Sundar has joined #openstack-cyborg14:00
shaohe_feng#startmeeting openstack-cyborg-driver14:01
openstackMeeting started Mon Aug 13 14:01:10 2018 UTC and is due to finish in 60 minutes.  The chair is shaohe_feng. Information about MeetBot at
shaohe_feng#topic Roll Call14:01
openstackUseful Commands: #action #agreed #help #info #idea #link #topic #startvote.14:01
openstackThe meeting name has been set to 'openstack_cyborg_driver'14:01
shaohe_feng#info shaohe_feng14:01
edleafe#info ed leafe14:02
Sundar#info Sundar14:05
*** annabelleB has joined #openstack-cyborg14:05
shaohe_fengSundar, edleafe morning14:05
*** dolpher has joined #openstack-cyborg14:06
edleafegood UGT morning to you!14:06
SundarGood day, Shaohe14:06
shaohe_fengedleafe, Sundar now only we three on line.14:07
*** Li_Liu has joined #openstack-cyborg14:08
Li_Liu#info Li_Liu14:08
Li_Liudo we still have the meeting today?14:08
Sundarshaohe: I had sent an email yesterday to openstack-dev. Shall we talk about that?14:08
shaohe_fengSundar: OK, go ahead.14:09
shaohe_fengLi_Liu: morning.14:09
*** xinran has joined #openstack-cyborg14:09
shaohe_fenggood evening,  xinran14:09
Li_Liushaohe_feng: good evening :P14:09
xinranHi all14:09
shaohe_fengSundar: You can talk about your email14:10
SundarIt was apparently decided in some meeting that, to record the discovered devices, Cyborg agent will call Cyborg REST API. Also, to allocate and deallocate accelerators. If so, that will make the public and ha many disadvantages.14:10
SundarAmong other things, it means it is not internal to Cyborg any more. Any user can call it. So, we should authenticate. Even if we open it only to operators, it is still error-prone. We can just keep it Cyborg-internal, right?14:11
xinranWhen Cyborg agent will call restful API?14:12
shaohe_fengCyborg agent should not  call Cyborg REST API.14:13
Li_LiuAgent should stay with rpc14:13
Sundarxinran: Rest API is meant for things that can be accessed by external users, operators or other services.14:13
*** wangzhh has joined #openstack-cyborg14:14
xinranSundar:  can you give us an example when agent call restful api14:14
Sundarshaohe, Li_Liu: Agreed. I amreferring to #link (Line 44)14:15
wangzhhHi all. Test connection.14:15
xinranIMHO agent should not call restful api14:16
Li_LiuSundar, that is exposed by cyborg-api, not meat to be used by Agent14:16
SundarLi_Liu: Was that agreed only for GET, not for PUT/POST?14:17
shaohe_fengwangzhh: good evening.14:18
wangzhhshaohe_feng: good evening. :)14:19
Li_LiuSundar, PUT/POST are open to admin I think, in case they want to tune some of the deployables14:19
SundarLi_Liu: the issue there is that we will have to maintain backwards compat for REST APIs. That is going to constrain our development and/or cause upgrade issues.14:20
Li_Liuwangzhh, do you have comments on this?14:21
wangzhhSundar,  what do you mean about "we will have to maintain backwards compat for REST APIs. That is going to constrain our development and/or cause upgrade issues."14:22
*** HongboZhao has joined #openstack-cyborg14:22
wangzhhCan u give me an example which current APIs can not handle?14:23
Sundarwangzhh: In #link (Line 44), if we allow PUT/POST in addition to get, we may have such considerations14:23
wangzhhWe have PATCH now. If we want to update deployable. we can use it.14:25
shaohe_fengsure we should allow PUT/POST,  that depends on14:25
Sundarwangzhh: Can you elaborate?14:26
wangzhhshaohe_feng: Do we need create deployable via API?14:26
shaohe_fengThat depends on.14:27
Li_Liuwangzhh, maybe we should allow that for now14:28
shaohe_fengat present, agent creates deployable.14:28
Li_Liuto let admin able to add stuff in manaully14:28
SundarLi_Liu, shaohe: can you explain why that is necessary? And how we will ensure backwards compat and avoid upgrade issues?14:29
shaohe_fengSundar: we should allow users to update some attribution of a deployable14:30
wangzhhI agree that users need to update something.14:30
SundarBy users, I hope you mean operators. End users (tenants) should not be touching this14:31
wangzhhBut can user create a new deployable? In which case? Could u give some example?14:31
Li_LiuThe users here should mean admin/operators14:31
shaohe_feng^ IMHO,  Li_Liu, the intention of your attribution design should allow user to update them, right?14:32
Li_Liuyes, indeed14:32
Sundar If operators need a manual way, can't we provide a script of some kind? Before opening up REST APIs, we need to understand their implications.14:32
Li_LiuI agree, right now Create api for deployables might not be very useful14:33
Li_LiuSundar, I assume the script you mena cyborg-pythonclient?14:34
shaohe_fengSundar: yes, at present it means admin/operations, Unless there are sufficient reasons that end users should update them14:34
wangzhhSundar, I think update is necessary. For example, I need change my deployable's name. Some attr which we can't collect from driver.14:35
SundarLi_Liu: isn't pythonclient a wrapper around the REST API?14:35
Sundarwangzhh: Sure, we can provide some script that not guarantees backwards compat till it all matures14:36
Sundar*does not guarantee14:36
SundarPerhaps @edleafe can comment but, my understand is that REST APIs need to be honored in future releases14:37
Li_LiuSo you suggestion is take out the deployable creation rest api?14:38
edleafeSundar: it has nothing to do with REST - any published API should not be removed or modified14:38
Sundaredleafe: What APIs do we publish to users apart from REST? RPC APIs are internal and can evolve across releases, right?14:39
edleafeA public API is a contract with your users14:39
edleafeI'm speaking more generally14:40
edleafeWhether your API is REST, RPC, SOAP, or anything else: public APIs need to be maintained14:40
SundarIf, say, Cyborg offers a RPC API to Nova or os-acc, we can evolve that in future releases based on mutual understanding, right? IOW, that doesn;t count as 'public' - exposed to users14:42
*** annabelleB has quit IRC14:42
*** annabelleB has joined #openstack-cyborg14:42
edleafeI'm not following. If another service is relying on an API, then it's public14:43
wangzhhSundar, if we use script. It's hard to intergrate with other project or software.14:43
wangzhhFor example, If we want to update attr by horizon?14:43
wangzhhCommand line and script is not friendly. :) IMHO.14:44
xinranIf other service interact with cyborg there may be the rest ful apis14:44
*** annabelleB has quit IRC14:45
wangzhhAgree with xinran.14:45
Sundarwangzhh: Before we make something public and incur all the costs for all future releases, let us keep it minimal. If it is not absolutely essential to do something with public APIs, we should look at alternatives first14:45
Sundaredleafe: Has nova/os-vif/neutron RPC APIs never changed? I thought those have evolved too14:46
edleafeSundar: I don't really know - I haven't worked with Neutron or os-vif14:47
Li_LiuSundar, I am thinking about it in the other direction tho. I think there is no harm to open couple rest apis and maintain them. As we couldn't confine what the users can/want to do in their scenarios.14:47
Li_LiuBy offering something more, can open up more opportunities for them.14:48
edleafeLi_Liu: you should always think very carefully before publishing an API14:48
SundarLi_Liu: That means keeping your deployables structure backwards compatible? Before we know all the use cases?14:49
shaohe_fengYes, API should be carefully, as edleafe said it is a contract :)14:50
Li_LiuFor sure, I am not saying we should do it without a thought :)14:51
wangzhhNova have API version  to handle these issue.14:51
Sundarshaohe: Not just the API. It will constrain how you can evolve the data structure that we call Deployable. How do we add new fields, drop old fields, or change the meaning/type/format of any field?14:51
*** annabelleB has joined #openstack-cyborg14:52
*** Coco_gao has joined #openstack-cyborg14:54
shaohe_fengSundar: yes, we need a API version for them.14:55
edleafeSundar: you're describing alpha development, where things can be added/removed/changed. Is that you would describe Cyborg's current state?14:55
wangzhhCoCo, Good evening. :)14:55
shaohe_fengCoco_gao: evening.14:55
Sundarwangzhh, shaohe, IIUC, having API versions does not mean that older versions can be dropped. They are still public14:55
Coco_gao#evening or morning, everyone.14:56
Sundaredleafe: Not quite, I am just advocating minimal public APIs till Cyborg has got some adoption and use cases are known and well-supported14:57
edleafeSundar: exactly. The point of API versioning is that the default never changes, but users can opt in to a newer version14:57
shaohe_fengSundar: yes. They are still public.  For the end users's app still use older versions, depends on older  instead of latest.14:57
wangzhhYep. LTS. But it can be dropped after depatched declaration.14:57
Li_LiuAs edleafe said, Cyborg is still at its very early dev stage14:58
SundarWhen you introduce API v2 with many changes, you still have to maintain v1 at least for some time. That means handling upgrades, conversions without breaking old functionality.14:58
edleafeI would think that given Cyborg's early development status, that any APIs you create are labeled as alpha, with a note that they may be changed in the future14:59
edleafeSundar: some would contend that you have to maintain v1 forever.14:59
edleafe(not me, though!)14:59
Sundaredleafe: iIf we can mark REST APIs as alpha, that will be great15:00
Li_LiuI think that's want we should be doing for now15:00
wangzhhSundar, all that mean that  we should do with thought. Instead of not to do.15:00
SundarThanks, edleafe15:02
SundarFolks, shall we agree to mark Deployables API as alpha (subject to change without backwards compat) for now?15:02
Li_LiuI vote yes15:03
wangzhhAgree with Sundar. :) :) :) :)15:04
Coco_gaoI agree, I think maybe the alpha is the version when cyborg finish interactions with nova.  Before that, cyborg is not workable by users.15:05
SundarGreat! Thanks, guys. :)15:06
shaohe_fengSundar: you should take more time/though on how to define the API. Post your ideas and let we discuss together.15:06
shaohe_fengSundar: have you consider the API that nova how to apply an accelerators from cyborg?15:07
Sundarshaohe: Yes, that should be part of the os-acc spec. We are still in the stage of getting agreement on the overall workflows and interactions. But I agree we should document that in detail15:08
shaohe_fengSundar: and also the parameters that API supports.15:09
SundarSure. Please see Line 525 for example15:10
shaohe_fengSundar: OK, for example, as a user I want to apply a accelerator,  accelerator type is FPGA instead of GPU, vendor is intel, model is A10, and it's function is Crypto.15:11
shaohe_fengSundar: maybe more parameters will be extended.15:12
Li_Liushaohe_feng, use the attribute in deployables, if the fields are not supported15:13
shaohe_fengSundar: these parameters, nova will extract for your defined traits/RC in the flavor15:13
SundarOperators will define flavors based on the attributes you mentioned, and users will pick one of the flavors. The accelerator requirement is conveyed though extra specs, and that is aprt of these function signatures15:13
shaohe_fengLi_Liu: Yes, I already use attribute  :)15:13
Li_Liuright now, when creating Deployables any parameters that are not native to deployable, will be added to the attribute_list15:14
shaohe_fengyes. I did use attribute as this way.15:14
shaohe_fengSundar: also can we support to apply batch accelerators in that API?15:15
Sundarshaohe: if you mean that a sigle VM may want more than one accelerator, sure.15:16
shaohe_fengSundar: for example, I want 2 accelerator,   both accelerator type is FPGA, vendor is intel, model is A10, and it's function is Crypto.15:16
Coco_gaoHave we make an agreement on users' input,  parameters or sections we may support by now?15:16
SundarCurrently, the API gets a list of device RPs selected by Nova, and extra specs. But that is not ideal. Still trying to think of a better way15:17
shaohe_fengSundar: or I want 2 different accelerators, one is GPU another is FPGA?15:17
Sundarshaohe: All that should be possible. The currently proposed APIs can technically handle them, but we can probably improve upon them15:17
shaohe_fengCoco_gao: Sundar is working on them.15:18
shaohe_fengSundar when will we see your API define?15:18
shaohe_fengSundar: one api to apply 2 accelerators instead of two api call?15:19
Sundarshaohe: I am updating the os-acc spec. It should be out by this week, if not in next couple of days.15:19
shaohe_fengSundar: nice.15:19
Sundarshaohe: The current API looks something like: prepareVANs(device_rp[], extra_specs, instance_info)15:19
Coco_gaoGreat jobs.15:20
SundarThe extra specs will contain the details of the user request, like: CUSTOM_ACCELERATOR_FPGA=2 and the traits15:20
SundarThe device_rp[] will contain the device resource providers selected by Nova for each of those accelerators15:20
shaohe_fengSundar: anyway, the api should well match the user the expect accelerators15:21
Li_Liudevice_rp is a list, and extra_specs will be applicable to all the RPs in the list?15:21
SundarThe problem is, how do we correlate each device_rp to a specific accelerator in extra specs. If we can massociate each user request with its own device RP, that will simplify our livs15:21
Li_Liucan we make extra_specs a list as well?15:23
SundarLi_Liu: yes. The extra specs may contain groups based on granular request syntax #link
SundarThe extra specs is a pre-defined thing coming from Nova.15:24
Li_Liuah, ok15:24
SundarNot sure if we can modify it. Still looking into that15:24
openstackgerritOpenStack Release Bot proposed openstack/cyborg master: Update reno for stable/rocky
shaohe_fengI have already use resource groups to support batch accelerators :)15:26
shaohe_fengSundar: remind,  have you notify rodrego that you and Coco_gao are refactoring the new returning  data of  the driver?15:26
SundarOh yes.15:27
SundarWe are currently looking at making the current driver patch work with Zuul15:27
shaohe_fengCoco_gao: how is the refactor going on.15:27
Coco_gaoyeah, I am working on it right now. Still need to test the code by now.15:28
SundarCoco_gao: we probably need to sync up. Are you using OVOs?15:29
shaohe_fengCoco_gao: good. Guess other developers are depending on your change.15:29
shaohe_fengSundar: have resolve the OPAE package install in Zuul?15:29
wangzhhCoco:Good job.  Plz don't forget test with my GPU driver.15:29
Coco_gaoOur Datacenter is going to move to other location,  so the test work need to be done several days later.15:30
Sundarshaohe: Rodrigo and I discussed the pkg install, and he is making changes15:31
Coco_gaoSundar, you mean VersionedObject?15:31
SundarCoco_gao: yes15:31
SundarWe should update the spec before submitting the patch?15:32
Coco_gaoyes,  I am using VersionedObject15:32
shaohe_fengSundar: good.15:32
SundarI am still backlogged on os-acc. I'll update the driver/agent spec. Coco_gao, I cna send you an early version for review15:33
shaohe_fengSundar: you should sync up with Coco_gao well.15:33
shaohe_fengLi_Liu: do we have deadline for this refactor?15:33
Coco_gaoI don't know15:34
Coco_gaomaybe we can discuss the deadline.15:34
Li_Liushaohe_feng, if you wanna catch the Rocky release, then Aug 20 - Aug 2415:35
Li_Liubut if not, there's not hard deadline on this15:35
Coco_gaoSundar, any detailed questions or informations , plz send me an email(
shaohe_fengyes, Your working is very important. for other drivers base on your work.15:36
SundarCoco_gao: sure15:36
Li_Liuwell, yes, we need to save time for reviews as well(if you wanna catch the R release)15:36
shaohe_fengLi_Liu:  I need to change the agent code, base on your new pf/vf model.15:37
Coco_gaoshaohe_feng, I know, maybe this week I will give an patch on the object first.15:38
SundarR release is past code freeze, right?15:38
Li_Liuok, let me know if you need any help/discussion15:38
Li_Liuwe can have ZOOM meeting anytime for discussion until Rocky releases15:38
Coco_gaothen shaohe can update the agent code, others can update the driver code.15:40
shaohe_fengSundar,  Li_Liu, will we not touch the API code until R release any more?15:41
shaohe_fengeven Sundar have define the new API, right?15:41
Li_LiuI think so15:42
Li_Liuno time for that15:42
Coco_gaoHi, shaohe, I think maybe zhenghao can help update agent code, since we need to test gpu driver together.15:43
wangzhhYep. shaohe_feng: Can I take it?15:43
Coco_gaoIf the agent code don't finished, we can't test our code seperately.15:44
shaohe_fengthat's great15:45
shaohe_feng3ks,  wangzhh15:45
wangzhhYou're welcome :)15:45
Coco_gaoEveryone, if I have any questions I will contact you directly and thank you for you support and patience~15:46
Li_LiuThank you Coco :)15:47
Li_LiuThanks a lot everyone for the hard work15:47
shaohe_fengcool, Coco.15:47
wangzhhNo problem. ;)15:48
Coco_gaoTime to bed if we don't have other questions.15:48
shaohe_fengyes, enough sleep keep girl beauty.15:48
wangzhhヾ(・ω・`。)  Good night.15:49
shaohe_fengAny thing wants to discuss?15:49
SundarGood night or good day, everybody15:49
shaohe_fengOK, let's end the meeting.15:49
Li_LiuHave a good nite guys15:49
wangzhhヾ( ̄▽ ̄)Bye~Bye~15:49
shaohe_fengThank you Li_Liu15:49
openstackMeeting ended Mon Aug 13 15:49:59 2018 UTC.  Information about MeetBot at . (v 0.1.4)15:50
openstackMinutes (text):
Coco_gaoenough sleep keep man strong...aha15:50
*** HongboZhao has quit IRC15:50
*** Li_Liu has quit IRC15:55
*** Sundar has quit IRC16:01
*** dolpher has quit IRC16:50
*** annabelleB has quit IRC16:54
*** annabelleB has joined #openstack-cyborg17:02
*** annabelleB has quit IRC17:07
*** annabelleB has joined #openstack-cyborg17:09
*** openstackgerrit has quit IRC17:19
*** wangzhh has quit IRC17:51
*** Coco_gao has quit IRC17:53
*** xinran has quit IRC18:19
*** ttk2[m] has quit IRC19:25
*** annabelleB has quit IRC19:25
*** efried has quit IRC19:58
*** efried has joined #openstack-cyborg20:13
*** annabelleB has joined #openstack-cyborg20:15
*** shaohe_feng has quit IRC20:42
*** annabelleB has quit IRC21:59

Generated by 2.15.3 by Marius Gedminas - find it at!