03:01:45 <Sundar> #startmeeting openstack-cyborg
03:01:46 <openstack> Meeting started Thu Oct 10 03:01:45 2019 UTC and is due to finish in 60 minutes.  The chair is Sundar. Information about MeetBot at http://wiki.debian.org/MeetBot.
03:01:47 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
03:01:49 <openstack> The meeting name has been set to 'openstack_cyborg'
03:01:57 <Sundar> #topic Who's here
03:02:05 <chenke> \0
03:02:09 <Yumeng> #info Yumeng
03:02:13 <chenke> #info chenke
03:02:17 <s_shogo> #info s_shogo
03:02:19 <xinranwang> #info xinranwang
03:02:47 <Sundar> Hi all, hope you had a good holiday last week.
03:03:23 <Sundar> #topic Status
03:03:49 <Sundar> Nothing much changed substantially last week. We got some doc patches into good shape.
03:04:34 <Sundar> Please see: https://review.opendev.org/686423, https://review.opendev.org/686518
03:04:55 <Sundar> The async bind patch got merged. Whew, finally
03:06:06 <Sundar> 'Install OPAE tools': https://review.opendev.org/684456 <-- this is specific to one driver but please do review it. Maybe you can look at the structure and style
03:07:03 <Sundar> 'Remove useless glance api_servers define': https://review.opendev.org/685542 Unfortunately, I always see failures with this patch. I provided the logs and my cyborg.conf in that patch.
03:07:52 <Sundar> Plus some generic patches like https://review.opendev.org/670973.
03:07:59 <chenke> yes. This patch in analysis.
03:08:05 <Sundar> Do any of you have any comments or questions?
03:08:35 <chenke> I guess this error is because the previous issues.
03:08:55 <chenke> endpoint = utils.get_endpoint(ksa_adap)
03:09:40 <chenke> The actually error is pointing here:  File "/opt/stack/cyborg/cyborg/image/glance.py", line 117, in get_api_servers\n    endpoint = utils.get_endpoint(ksa_adap)\n', u'  File "/opt/stack/cyborg/cyborg/common/utils.py", line 208, in get_endpoint\n    "interfaces: %s" % interfaces)\n
03:10:16 <Sundar> chenke: Could you tell us what the previous issues are?
03:10:35 <chenke> This code has not been executed
03:11:15 <chenke> https://github.com/openstack/cyborg/blob/master/cyborg/image/glance.py#L116
03:11:28 <chenke> Sundar pls see this if condition
03:11:54 <Sundar> Yes. So, the api_servers config seems necessary, right?
03:11:59 <chenke> no
03:12:14 <chenke> The api_servers hide the else condition.
03:13:06 <chenke> now when I delete this api_servers, the else condition raise error because "/opt/stack/cyborg/cyborg/common/utils.py", line 208, in get_endpoint"
03:13:36 <chenke> What I want to say is that this problem was hidden before.
03:14:32 <Sundar> Got it. So, the api_servers is mor elike a hack or workaround to hide the error in the 'else' clause
03:14:49 <chenke> Right.
03:15:36 <chenke> After our meeting, I will find why the get_endpoint raise error.
03:15:37 <Sundar> Then the questions are: A. Do we have time to find the issue in the get_endpoint()? B. Is it worth fixing -- may be we should go straight to openstacksdk in U?
03:16:03 <chenke> maybe we could use openstacksdk.
03:16:33 <chenke> need to have a try. I am not sure now. But eric suggest glance to use openstacksdk before.
03:17:22 <Sundar> Yes, I believe we tried that approach and gave up because of a bug there. The opestacksdk patch which was suggested as a fix has since been abandoned.
03:18:06 <Sundar> Anyway, I agree this is a U-release conversation. Perhaps we should get started soon so that we can shoot for a resolution in November, early in the U cycle
03:18:55 <chenke> agree.
03:20:34 <Sundar> Cool, thanks Chenke
03:21:09 <Sundar> All, please review the documentation patches. We need to close them for the next RC.
03:21:56 <Sundar> Also, like I said, please review: https://review.opendev.org/684456
03:22:15 <Sundar> #topic PTG prep
03:23:06 <Sundar> I will create an etherpad and share with you. Please indicate who among you plan to attend.
03:23:13 <xinranwang> next RC is this week or 18 Oct
03:24:25 <Sundar> xinranwang: The final RC is due this week. But I haven't seen a RC patch proposed yet.
03:24:32 <Sundar> 18 Oct is the release itself
03:25:30 <xinranwang> Sundar:  Ok, I will review it by this week.
03:25:35 <Sundar> Yumeng: we need to work together for the project update.
03:26:07 <Yumeng> sundar: yes, I was about to ask.
03:26:19 <Yumeng> do you have any plans about the structure?
03:26:20 <Sundar> Just curious, who among you plan to attend the PTG (after the Summit)?
03:27:06 <Sundar> Yumeng: kind of. Can you access Google slides?
03:27:29 <xinranwang> I will attend Summit and PTG, and I am preparing a demo at Intel booth during the Summit which show qat use case in cyborg and performance improvement.
03:27:45 <shaohe_feng> hi all.
03:27:51 <shaohe_feng> sorry for late
03:27:51 <Sundar> Cool
03:28:07 <shaohe_feng> something wrong with my pc.
03:28:21 <Sundar> shaohe_feng: welcome, np
03:28:53 <chenke> hi shaohe.
03:29:07 <shaohe_feng> morning chenke
03:29:21 <Yumeng> Sundar: google doc not always available. It will be appreciate if u can send me the file directly!
03:29:33 <Sundar> DO any of you have specific ideas for U release? Such as new drivers, or improvements to Cyborg? (Apart from nova integ, which IMO is most important.)
03:29:43 <s_shogo> I'll attend the Summit and PTG (all day).
03:29:46 <Sundar> Yumeng: got it, will do
03:29:51 <Yumeng> I will attend Summit and PTG too
03:30:11 <Sundar> Great, s_shogo and Yumeng
03:30:31 <Sundar> Hope we'll get to meet chenke and zhurong too ;)
03:30:58 <xinranwang> Sundar:  from Intel side, there are  HDDL and QAT driver. We can do it in U release.
03:31:01 <chenke> I will attend Summit and PTG too.
03:31:18 <Sundar> Great
03:31:25 <shaohe_feng> Yes, HDDL and QAT
03:31:30 <Sundar> #topic AoB
03:31:38 <Sundar> Anything else to bring up today?
03:31:49 <shaohe_feng> for FPGA, should we support alias for function_id?
03:32:03 <shaohe_feng> such as in DP:
03:32:12 <zhurong> Sundar hope to see you too
03:32:38 <Sundar> shaohe: Yes, that would be a good addition. After nova integ.
03:32:45 <xinranwang> Yes, as shaohe said. it remains many TODOs and improvement in Cyborg, we can discuss during PTG.
03:32:47 <Sundar> zhurong: Yes, I plan to be there
03:33:03 <Sundar> My first trip to China :)
03:33:29 <chenke> Sundar welcome to china.
03:33:33 <shaohe_feng> "accel:function_id": "123_456_789_333", this is not friendly to user.
03:33:52 <shaohe_feng> Sundar welcome to china
03:34:06 <Sundar> Thanks
03:34:14 <xinranwang> welcome to China :)
03:34:23 <shaohe_feng> maybe "accel:function_alias": "SmartNic_V1"
03:34:30 <shaohe_feng> or other describe it.
03:35:10 <Sundar> Yes, let's brainstorm on the function_id stuff. Perhaps I should create an etherpad for that.
03:35:18 <xinranwang> Sundar:  can you create a etherpad that we can add the idea on it.
03:35:25 <Sundar> Sure
03:35:28 <shaohe_feng> yes.
03:35:37 <shaohe_feng> a etherpad is good.
03:35:45 <Sundar> Cool. Anything else, folks?
03:35:51 <shaohe_feng> another things, now the async job is in API.
03:35:54 <xinranwang> And we can discuss them during the meeting
03:36:19 <shaohe_feng> we have issue, should be support async job in conductor or agent
03:36:35 <shaohe_feng> pros and cons?
03:36:47 <Sundar> IMHO it should be in the conductor
03:37:31 <shaohe_feng> OK, we can discuss and  improve it.  also put in etherpad
03:37:39 <Sundar> ok
03:38:32 <Sundar> From the next meeting, we can start focusing more on the PTG.
03:38:44 <shaohe_feng> also another thing, a mini improve for factory.
03:38:52 <shaohe_feng> https://stackoverflow.com/questions/3464061/cast-base-class-to-derived-class-python-or-more-pythonic-way-of-extending-class
03:39:00 <shaohe_feng> ^ #info link
03:39:26 <Sundar> Details like this are probably better discussed as a patch proposal, rather than in etherpad, I think
03:39:46 <shaohe_feng> this is no in etherpad.
03:39:53 <shaohe_feng> just discuss it here
03:40:10 <shaohe_feng> mini performance improvement
03:40:21 <shaohe_feng> now when I found it is a FPGA ARQ, I create a FPGA ARQ
03:40:36 <shaohe_feng> that means I access a DB again.
03:41:17 <shaohe_feng> we can cast base ARQ to FPGA ARQ to avoid the DB access
03:41:46 <Sundar> shaohe_feng: Ok. Please feel free to propose a patch with your ideas.
03:42:13 <shaohe_feng> do you think is it worth to do it?
03:42:17 <shaohe_feng> ^ Sundar
03:42:57 <Sundar> shaohe_feng: Yes. Not terribly sure about the casting across parent and child classes, though. That can be tricky.
03:43:34 <Sundar> Trying to avoid an extra db access makes sense.
03:43:43 <shaohe_feng> just a mini performance improvement.  and mini patch.
03:43:56 <shaohe_feng> you can see this example in the link
03:44:01 <shaohe_feng> https://stackoverflow.com/questions/3464061/cast-base-class-to-derived-class-python-or-more-pythonic-way-of-extending-class
03:44:13 <shaohe_feng> from math import pi
03:44:17 <shaohe_feng> print repr(c)
03:44:29 <shaohe_feng> only one line change:
03:44:31 <shaohe_feng> c.__class__ = CirclePlus
03:44:44 <Sundar> Perhaps it is better to focus on more basic stuff, heading into U. We really need to focus on nova integ. One consistent feedback i hear is that CYborg code and patches tend to lack UT
03:45:32 <Sundar> We completely lack functional testing with gabbi. These are not jazzy items but will help convince other developers and end users (operators) that we have good quality
03:45:35 <shaohe_feng> In cyborg: we just change one line the,  arqobj.__class__ = FPAGExaARQ
03:46:01 <shaohe_feng> also, we should list a task for testcase in the etherpad
03:46:20 <shaohe_feng> too many testcase should be added
03:46:29 <shaohe_feng> it is a huge task
03:47:02 <shaohe_feng> Hopeful everyone can take some testcase task.
03:48:11 <Sundar> shaohe_feng: I like that. We should certainly spread the UT.
03:48:33 <Sundar> What do others think? Yumeng, chenke, zhurong, xinranwang
03:50:09 <Sundar> We can split it into API, objects and db layer. By specific APIs (like device profile create), specific objects (e.g. ARQs, device profiles) and specific features (cross-service connections, etc.)
03:51:57 <Yumeng> yes agree,  it is urgent to improve UT and tempest case.
03:52:29 <Sundar> Cool. We can follow up in the etherpad. Anything else today? Will wait for 30 seconds ...
03:52:32 <chenke> nova integ and UT improve are the important things in U.
03:53:53 <Sundar> Agreed, along with the /v2/device,deployables API for listing inventory and programming
03:54:39 <Sundar> Thanks everybody. Have a good week! Bye.
03:54:43 <Sundar> #endmeeting