09:00:43 <priteau> #startmeeting blazar
09:00:45 <openstack> Meeting started Tue Jul  7 09:00:43 2020 UTC and is due to finish in 60 minutes.  The chair is priteau. Information about MeetBot at http://wiki.debian.org/MeetBot.
09:00:46 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
09:00:48 <openstack> The meeting name has been set to 'blazar'
09:01:03 <priteau> #topic Roll call
09:01:10 <tetsuro> o/
09:01:28 <priteau> Hi tetsuro
09:04:48 <priteau> No one else today, let's start
09:04:56 <priteau> #topic Code review priorities
09:05:29 <priteau> I've submitted the AZ fix to stable branches
09:05:39 <priteau> https://review.opendev.org/#/q/If6f6cf21170e95c58dd8a0a1b41b9b5f4797a627
09:05:41 <priteau> Please review
09:07:04 <priteau> I stopped at Stein as it is the oldest one still in Maintained status
09:08:09 <priteau> Also I don't think we have users using Rocky or older
09:08:23 <priteau> Chameleon is on Train, Nectar is using Stein
09:09:49 <priteau> The other review priority is https://review.opendev.org/#/c/731586/
09:16:27 <priteau> tetsuro: I looked at your vGPU reservation draft spec. I am not sure we are on the same page: I thought we were going to support general-purpose PCI passthrough and that would apply to GPUs but also FPGAs, fast interconnects, etc.
09:17:36 <priteau> Your spec works with vGPU support in Nova, which AFAIK is limited to specific nvidia and intel cards
09:18:13 <priteau> It is useful too, but it may be difficult to get access to the hardware to test it
09:18:54 <priteau> PCI-passthrough is more generic
09:19:27 <priteau> What are your thoughts? We can make two separate specs if needed
09:19:52 <tetsuro> Yeah, two separate specs would be nice
09:20:59 <priteau> For your own research projeect, are you interested more about Nova virtual GPU support than PCI passthrough?
09:21:55 <tetsuro> I am interested in PCI passthrough reservation as well as vGPU support, but I am not sure what you mean by general-purpose PCI passthrough reservation.
09:22:15 <tetsuro> Is it something to do with cyborg?
09:22:47 <priteau> By general-purpose I meant that PCI passthrough will work with any PCI device, it is not restricted to GPUs
09:23:16 <priteau> So by implementing support for PCI passthrough we would support GPUs, FPGAs, network cards, etc.
09:23:30 <tetsuro> Cool
09:25:36 <priteau> Can you create a new draft spec for PCI passthrough based on the one you already created?
09:26:09 <tetsuro> Well, that is cool, but to realize it, will blazar discover those devices, or it is another project to discover and manage the devices and blazar only manages the schedule slot of those devices?
09:27:25 <priteau> I don't know much about the role Cyborg plays with GPUs, FPGAs. I've used PCI passthrough only with Nova itself.
09:27:55 <priteau> I think we would need to teach Blazar about available devices on compute nodes, unless we can discover them automatically via the API
09:31:34 <priteau> So actions are:
09:31:45 <priteau> 1) create a new draft spec for PCI passthrough
09:32:04 <priteau> 2) look into what role Cyborg plays with accessing PCI devices
09:33:08 <priteau> 3) Adapt spec for PCI passthrough requirements
09:34:10 <tetsuro> Right
09:35:15 <priteau> #topic AOB
09:35:21 <priteau> Anything else to share today?
09:35:45 <priteau> A reminder that I am off next Tuesday so I am suggesting we cancel the next meeting
09:37:11 <tetsuro> I got it
09:38:28 <priteau> That's all from me this week
09:39:46 <priteau> Thanks for joining tetsuro
09:39:54 <priteau> #endmeeting