Wednesday, 2018-01-03

*** crushil has joined #openstack-cyborg03:15
*** crushil_ has joined #openstack-cyborg03:15
*** crushil_ has quit IRC03:15
*** mikeH has joined #openstack-cyborg03:17
*** mikeH has quit IRC03:23
*** masber has quit IRC03:32
*** masber has joined #openstack-cyborg04:21
*** crushil_ has joined #openstack-cyborg04:46
*** crushil_ has quit IRC04:51
*** crushil_ has joined #openstack-cyborg04:51
*** crushil_ has quit IRC05:25
*** crushil_ has joined #openstack-cyborg05:31
*** crushil_ has quit IRC05:33
*** crushil_ has joined #openstack-cyborg05:34
*** masber has quit IRC06:13
*** crushil_ has quit IRC06:32
*** jkilpatr has quit IRC11:00
*** jkilpatr has joined #openstack-cyborg11:49
*** jkilpatr has quit IRC12:28
*** crushil has quit IRC12:53
*** crushil has joined #openstack-cyborg13:02
*** crushil has quit IRC13:07
*** crushil has joined #openstack-cyborg13:08
*** jkilpatr has joined #openstack-cyborg13:42
*** zhipeng has joined #openstack-cyborg14:23
jkilpatrhowdy everyone14:47
*** vipparthy has joined #openstack-cyborg14:50
vipparthygood morning folks14:51
zhipenghappy new year folks:)14:52
*** mpaolino has joined #openstack-cyborg14:53
*** Li_Liu has joined #openstack-cyborg14:59
zhipengho ho ho15:00
zhipeng#startmeeting openstack-cyborg15:00
openstackMeeting started Wed Jan  3 15:00:24 2018 UTC and is due to finish in 60 minutes.  The chair is zhipeng. Information about MeetBot at
openstackUseful Commands: #action #agreed #help #info #idea #link #topic #startvote.15:00
*** openstack changes topic to " (Meeting topic: openstack-cyborg)"15:00
openstackThe meeting name has been set to 'openstack_cyborg'15:00
*** Yumeng__ has joined #openstack-cyborg15:00
zhipengmeetbot still on vacation ?15:00
Li_LiuHappy New Year guys!15:01
jkilpatrhappy new year15:01
zhipeng#startmeeting openstack-cyborg15:02
openstackzhipeng: Error: Can't start another meeting, one is in progress.  Use #endmeeting first.15:02
openstackzhipeng: Error: Can't start another meeting, one is in progress.  Use #endmeeting first.15:03
jkilpatrmeetbot seemed to have worked the first time?15:03
zhipengah looks like my network's problem15:03
zhipengwho do we have here?15:04
zhipeng#topic Roll Call15:04
*** openstack changes topic to "Roll Call (Meeting topic: openstack-cyborg)"15:04
zhipeng#info Howard15:04
jkilpatr#info Justin15:04
mpaolino#info Michele15:05
zhipengand also #info liliu and yumeng15:06
zhipeng#topic Queens Release Prep15:07
*** openstack changes topic to "Queens Release Prep (Meeting topic: openstack-cyborg)"15:07
zhipengfolks entering Jan we are officially in the mode for release prep15:07
vipparthyhere is Vipparthy15:07
zhipengso we need to accelerate our progress15:07
zhipenglet's run through the open patches, see what is still missing and what could be closed15:08
zhipeng#info code freeze is Feb 515:08
zhipeng#info 1. Internal API spec15:08
zhipengjkilpatr any update on this one ?15:09
jkilpatrzhipeng, waiting on feedback, I'm pretty happy with it.15:09
zhipengi think we should keep it open as we flash out the api/conductor/agent extension update ?15:10
jkilpatrsounds good to me, talk about what you want in the review.15:10
zhipengno problem, I think we should go back to the spec after we land zhuli's API patch15:11
zhipengok next up15:11
zhipeng#info 2. CRUD api extension15:11
Li_Liu  #info Li15:13
zhipengTL; DR for this patch is that15:13
zhipengfor Queens we will need a working API/DB that implements the resource provider data model15:13
zhipengthis patch is the attempt to achieve that15:14
jkilpatrI'll review today15:14
crushilzhipeng, The generic driver depends on this patch. Once this patch merges, I can add the implementation for the CRUD to the generic driver too15:14
crushilAnd it doesn't have unit tests yet15:14
zhipengso I think as I mentioned before, the implementation should align with jkilpatr's internal API spec15:14
zhipengcrushil ya I saw your comment15:15
zhipengI will work with zhuli and let's land this patch this week15:15
crushilI have already added create functionality to the generic driver15:15
jkilpatrI need to look at my own commit again, it's been a while.15:16
zhipengi know :P15:16
*** Amol has joined #openstack-cyborg15:16
zhipengokey without further a due15:16
zhipeng#info 3. generic driver15:16
zhipengcrushil take away15:16
crushilSo, as I already mentioned this patch depends on and once that patch merges, I can add functionality for the other use cases for Cyborg.15:17
zhipengwe also need unit test for this one right ? I mean later15:19
*** shaohe_feng_ has joined #openstack-cyborg15:19
zhipenghow generic should we expect for the driver when we release it ? Can we have very simple demo on NVIDIA GPU ?15:20
crushilThat's the goal15:21
crushilWe can do a simple POC at the OS summit15:21
zhipengok now let's move on two the relative two new specs15:22
zhipeng#info 4. FPGA data modeling spec15:22
vipparthycan you please plan for Demo online, so we can join and view functionality , Thank you15:22
zhipengthis is actually the spec that zhuli is implementing in his patch15:22
zhipengvipparthy we could discuss that at Dublin PTG for sure15:23
zhipengLi Liu you there ?15:23
vipparthyThank you Zhipeng15:23
zhipengcould you briefly introduce the spec ?15:24
zhipengso that everyone gets up to speed15:24
Li_LiuBasically, the spec introduces a way to describe hierarchy accelerator structure15:25
Li_LiuOne of the most typical usage is FPGA15:26
Li_LiuWe plan to add a new table called Deployables15:26
Li_LiuIt represent any accelerators that can be discovered/deployed15:27
Li_LiuThe key is it has 2 pointers, parent_id and root_ir15:27
Li_Liuparent_id points to the parent Deployable row15:28
Li_Liuroot_id points to the root Deployable row15:28
Li_LiuThis way, we can form nested tree structures15:29
Li_LiuFor instance, in the case of FPGA15:29
Li_LiuEach FPGA can have multiple Physical Functions15:30
Li_LiuEach Physical Functions can have multiple Virtual Functions15:30
Li_LiuAlso, we have a Attribute table which stores k-v pairs referencing back to the Deployables table15:32
zhipengis this aligned with the nested resource provider spec ?15:32
Li_Liuthis way, we can track arbitrary type of attributes for generic types15:32
zhipeng(which expected to be done in Queens)15:32
Li_LiuIt does, actually15:33
zhipengthx for the intro, everyone plz help review the spec15:34
Li_Liuthe Deployable is counterpart to resource providers in NOVA15:34
zhipengzhuli will also reflect the changes in the spec on the fly15:34
Li_LiuThanks a lot guys15:34
zhipengmoving on to the next15:35
zhipeng#info 5. Intel's FPGA device driver15:35
zhipengshaohe_feng_ could you help introduce the spec ?15:36
shaohe_feng_zhipeng: yes15:38
shaohe_feng_zhipeng: this spec is going to provide the fpga information to the Li_Liu's spec15:39
shaohe_feng_zhipeng: and support a program interface for FPGA15:40
shaohe_feng_ When cyborg agent starts or does resource check periodically, the cyborg FPGA driver should enumerate the list of the FPGA devices, and report the details of all available FPGA accelerator on the host. Such as BDF, PID, VID, IMAGE_ID, PF/VF type.15:41
zhipengthis has any impact on your work @crushil ?15:41
shaohe_feng_When user uses empty FPGA regions as its accelerator, cyborg agent should call driver's program() interface. Cyborg agent should provide BDF of VF, and local image path to the driver. More details ref15:41
crushilI'm not sure15:44
crushilI need to look at the spec15:44
shaohe_feng_zhipeng: I'm not familiar with other vendor's FPGA. Not sure we need a virtual FPGA driver interface.15:44
zhipengcrushil okey15:45
zhipengshaohe_feng_ I think Li_Liu could help you with the insight of other vendor implementation15:46
shaohe_feng_o, sorry. the driver should can program any FPGA regions not only empty region15:46
shaohe_feng_zhipeng: that's greate15:46
shaohe_feng_will update the spec.15:47
zhipengokey then we have the last non-priority patch15:47
zhipeng#info 6. SPDK driver code15:47
shaohe_feng_the agent decide which region need to be re-program.15:48
zhipengthe SPDK code is developed by people in my team, so if we could squeeze it in Queens then it would be great, if not it is also not an issue15:49
zhipengwill ping you guys when it is ready15:49
zhipeng#topic Queens Priorities15:49
*** openstack changes topic to "Queens Priorities (Meeting topic: openstack-cyborg)"15:49
jkilpatrproof of concept?15:49
zhipengI would suggest the following priorities for this month15:49
zhipengjkilpatr ya like that15:49
zhipengwe have patches need to be merged in spdk in the first place to make it work15:50
zhipengbut it is non-priority so no hurry on that one15:50
zhipengi will handle that quietly on my side :P15:50
zhipengokey back to priority suggestion15:51
zhipeng#info suggestion 1: land LiLiu and zhuli's patch this week15:51
*** Amol has quit IRC15:51
zhipengi know there are a lot of reviews to be done, but let's try to crunch it this week15:52
zhipengso that crushil and justin's work could move ahead15:52
zhipeng#info suggestion 2: we land justin's and shaohe's spec by the end of next week15:52
zhipengmeaning that once we agreed upon the API layer implementation this week, we could crunch out the conductor/agent/driver interface pretty much next Wed15:53
zhipengand we should be okey to have a full picture for the internal API spec, as well as shalhe's driver spec to align with it15:54
Li_Liusounds great15:54
zhipeng#info suggestion 3: at the moment we keep the vendor driver code in-tree since there are not a lot15:54
jkilpatrI'll give it a go.15:54
jkilpatrno reason to make things more complicated until it becomes an issue.15:55
zhipenganyone else comment on my three suggestions ?15:55
zhipengcrushil ?16:02
crushilWe should keep it in-tree16:03
crushilWe are not in that position yet to move it out of tree16:03
vipparthyzhipeng, who can help me to identify driver specs to be written for different FPGA's in our team ?16:04
zhipengcrushil exactly16:04
crushilAnd apologies for delays. I'm in a conference right now16:04
zhipengno problem :)16:04
zhipengvipparthy you could go over the log, there are info points that could help you identify it16:05
zhipengso if there are no other issues, I will close the meeting on time-ish :P16:06
zhipenglet's go to work and try to deliver an awesome first release16:06
*** jkilpatr has quit IRC16:06
*** dims has quit IRC16:09
*** openstack changes topic to "#openstack-cyborg"16:09
openstackMeeting ended Wed Jan  3 16:09:23 2018 UTC.  Information about MeetBot at . (v 0.1.4)16:09
openstackMinutes (text):
*** dims has joined #openstack-cyborg16:10
*** shaohe_feng_ has quit IRC16:11
*** zhipeng has quit IRC16:13
*** jkilpatr has joined #openstack-cyborg16:20
*** jkilpatr has quit IRC16:22
*** jkilpatr has joined #openstack-cyborg16:23
vipparthythank you folks16:23
*** vipparthy has quit IRC16:23
*** Yumeng__ has left #openstack-cyborg16:26
*** A_ has joined #openstack-cyborg16:50
*** A_ has quit IRC16:50
*** mpaolino has quit IRC17:01
*** jkilpatr has quit IRC18:03
*** jkilpatr has joined #openstack-cyborg18:04
*** jkilpatr has quit IRC18:40
*** mikeH_ has joined #openstack-cyborg18:41
*** mikeH_ has quit IRC18:43
*** mikeH_ has joined #openstack-cyborg19:07
*** mikeH_ has quit IRC19:08
*** mikeH_ has joined #openstack-cyborg19:09
*** mikeH_ has quit IRC19:10
*** jkilpatr has joined #openstack-cyborg20:21
*** am_ has joined #openstack-cyborg21:43
*** am_ has quit IRC21:44
*** crushil has quit IRC22:06
*** crushil has joined #openstack-cyborg22:33
*** vipparthy has joined #openstack-cyborg22:42
*** vipparthy has quit IRC22:42
*** crushil has quit IRC22:47
*** crushil has joined #openstack-cyborg22:48
*** crushil has quit IRC22:51
*** crushil has joined #openstack-cyborg22:55
*** crushil has quit IRC23:07
*** masber has joined #openstack-cyborg23:16

Generated by 2.15.3 by Marius Gedminas - find it at!