| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2021-04-22 | |||
| 11:53:15 | Dmitrii-Sh | sean-k-mooney: I see. Do you plan to replace InstancePCIRequests with placement resource requests? | |
| 11:53:15 | Dmitrii-Sh | sean-k-mooney: I see. Do you plan to replace InstancePCIRequests with placement resource requests? | |
| 11:53:36 | artom | Dmitrii-Sh, I think a sort of "smartnic for dummies" paragraph (or section) might be useful, just to give the rest of use context on what this thing even is | |
| 11:53:36 | artom | Dmitrii-Sh, I think a sort of "smartnic for dummies" paragraph (or section) might be useful, just to give the rest of use context on what this thing even is | |
| 11:55:07 | Dmitrii-Sh | artom: ty. I tried to make it largely agnostic of a specific vendor and target a class of devices: off-path SmartNICs with a separate (mostly ARM) CPU. There are several vendors out there that follow this pattern. | |
| 11:55:07 | Dmitrii-Sh | artom: ty. I tried to make it largely agnostic of a specific vendor and target a class of devices: off-path SmartNICs with a separate (mostly ARM) CPU. There are several vendors out there that follow this pattern. | |
| 11:55:34 | Dmitrii-Sh | but I can add a paragraph discussing it, or even better, a reference to an overview about those | |
| 11:55:34 | Dmitrii-Sh | but I can add a paragraph discussing it, or even better, a reference to an overview about those | |
| 11:55:52 | artom | I don't want to speak for everyone here, but you have a whole bunch of vocabulary there that isn't accessible to most of us ("switchdev-capable", "off-path") | |
| 11:55:52 | artom | I don't want to speak for everyone here, but you have a whole bunch of vocabulary there that isn't accessible to most of us ("switchdev-capable", "off-path") | |
| 11:56:12 | Dmitrii-Sh | I guess I should link this then https://netdevconf.info/0x14/pub/slides/39/Netdev%200x14%20--%20Taking%20Control%20of%20your%20SmartNIC%20v1.pdf | |
| 11:56:12 | Dmitrii-Sh | I guess I should link this then https://netdevconf.info/0x14/pub/slides/39/Netdev%200x14%20--%20Taking%20Control%20of%20your%20SmartNIC%20v1.pdf | |
| 11:56:25 | sean-k-mooney | Dmitrii-Sh: not replace them no but we would also claime resouces in placment for them | |
| 11:56:25 | sean-k-mooney | Dmitrii-Sh: not replace them no but we would also claime resouces in placment for them | |
| 11:56:44 | artom | Dmitrii-Sh, yeah, any kind of introduction or reference to what this hardware is and how it's different from the PCI devices that we're used to would be very useful. For me, at least. | |
| 11:56:44 | artom | Dmitrii-Sh, yeah, any kind of introduction or reference to what this hardware is and how it's different from the PCI devices that we're used to would be very useful. For me, at least. | |
| 11:56:54 | artom | Hopefully for other non-sean-mooney reviewers too | |
| 11:56:54 | artom | Hopefully for other non-sean-mooney reviewers too | |
| 11:57:25 | sean-k-mooney | Dmitrii-Sh: im not really trielled with nova calling neutron as part fo vif plugging as a clinet | |
| 11:57:25 | sean-k-mooney | Dmitrii-Sh: im not really trielled with nova calling neutron as part fo vif plugging as a clinet | |
| 11:57:31 | sean-k-mooney | even if that was doen via os-vif | |
| 11:57:31 | Dmitrii-Sh | artom: ack, will do. I understand that it's a lot of info to process | |
| 11:57:31 | sean-k-mooney | even if that was doen via os-vif | |
| 11:57:31 | Dmitrii-Sh | artom: ack, will do. I understand that it's a lot of info to process | |
| 11:58:07 | sean-k-mooney | Dmitrii-Sh: have you spoken to the cyborg team about this by the way | |
| 11:58:07 | sean-k-mooney | Dmitrii-Sh: have you spoken to the cyborg team about this by the way | |
| 11:58:34 | sean-k-mooney | they were tryign to enable reprogrambel smart nics via a different approch | |
| 11:58:34 | sean-k-mooney | they were tryign to enable reprogrambel smart nics via a different approch | |
| 11:58:50 | Dmitrii-Sh | sean-k-mooney: I've seen their specs but they seem to be targeting FPGA-based on-path NICs (Intel IIRC) | |
| 11:58:50 | Dmitrii-Sh | sean-k-mooney: I've seen their specs but they seem to be targeting FPGA-based on-path NICs (Intel IIRC) | |
| 11:59:10 | Dmitrii-Sh | and the approach there is to program NICs from the hypervisor host itself | |
| 11:59:10 | Dmitrii-Sh | and the approach there is to program NICs from the hypervisor host itself | |
| 11:59:11 | sean-k-mooney | Dmitrii-Sh: primarally | |
| 11:59:11 | sean-k-mooney | Dmitrii-Sh: primarally | |
| 11:59:26 | sean-k-mooney | i just wanted to make sure you were aware of it | |
| 11:59:26 | sean-k-mooney | i just wanted to make sure you were aware of it | |
| 11:59:27 | Dmitrii-Sh | where as here there is a security boundary between the hypervisor host and the SmartNIC host | |
| 11:59:27 | Dmitrii-Sh | where as here there is a security boundary between the hypervisor host and the SmartNIC host | |
| 12:00:01 | sean-k-mooney | so there was an effort to allow the neutron agent to manage remote host in the past | |
| 12:00:01 | Dmitrii-Sh | sean-k-mooney: ack, I am. I wish it addressed the things needed for off-path NICs to avoid fragmentation | |
| 12:00:01 | sean-k-mooney | so there was an effort to allow the neutron agent to manage remote host in the past | |
| 12:00:01 | Dmitrii-Sh | sean-k-mooney: ack, I am. I wish it addressed the things needed for off-path NICs to avoid fragmentation | |
| 12:00:08 | sean-k-mooney | what you are proposing is closer to that | |
| 12:00:08 | sean-k-mooney | what you are proposing is closer to that | |
| 12:00:37 | sean-k-mooney | i think we would still want to keep the binding host the same at the ml2 level | |
| 12:00:37 | sean-k-mooney | i think we would still want to keep the binding host the same at the ml2 level | |
| 12:00:44 | sean-k-mooney | the host on which the vm resides | |
| 12:00:44 | sean-k-mooney | the host on which the vm resides | |
| 12:00:53 | Dmitrii-Sh | There was some work in Ironic in the past https://specs.openstack.org/openstack/ironic-specs/specs/approved/support-smart-nic.html | |
| 12:00:53 | Dmitrii-Sh | There was some work in Ironic in the past https://specs.openstack.org/openstack/ironic-specs/specs/approved/support-smart-nic.html | |
| 12:01:03 | sean-k-mooney | but we may need to have other information that woudl be use to corralate that to the chassis in ovn | |
| 12:01:03 | sean-k-mooney | but we may need to have other information that woudl be use to corralate that to the chassis in ovn | |
| 12:01:31 | Dmitrii-Sh | as a result neutron-openvswitch-agent started using os-vif to plug representors (but that's not usable outside ironic at the moment) | |
| 12:01:31 | Dmitrii-Sh | as a result neutron-openvswitch-agent started using os-vif to plug representors (but that's not usable outside ironic at the moment) | |
| 12:02:16 | Dmitrii-Sh | sean-k-mooney: yes, I think there has to be a mapping table somewhere but I haven't quite decided where yet | |
| 12:02:16 | Dmitrii-Sh | sean-k-mooney: yes, I think there has to be a mapping table somewhere but I haven't quite decided where yet | |
| 12:03:09 | sean-k-mooney | Dmitrii-Sh: in neutron | |
| 12:03:09 | sean-k-mooney | Dmitrii-Sh: in neutron | |
| 12:03:29 | sean-k-mooney | i dont think nova should need to know what network backend is being used in this code | |
| 12:03:29 | sean-k-mooney | i dont think nova should need to know what network backend is being used in this code | |
| 12:03:35 | Dmitrii-Sh | sean-k-mooney: this makes sense. Otherwise Nova becomes too aware of a particular ML2 mechanism driver | |
| 12:03:35 | Dmitrii-Sh | sean-k-mooney: this makes sense. Otherwise Nova becomes too aware of a particular ML2 mechanism driver | |
| 12:03:47 | sean-k-mooney | yep | |
| 12:04:05 | sean-k-mooney | i have only skimmed the spec so i really need to set some time aside to digest it | |
| 12:04:05 | sean-k-mooney | i have only skimmed the spec so i really need to set some time aside to digest it | |
| 12:04:23 | Dmitrii-Sh | sean-k-mooney: ok, np, thanks for the early and quick feedback - much appreciated | |
| 12:04:23 | Dmitrii-Sh | sean-k-mooney: ok, np, thanks for the early and quick feedback - much appreciated | |
| 12:04:39 | Dmitrii-Sh | sean-k-mooney: I am happy to make adjustments to make it work with the ongoing efforts in Nova | |
| 12:04:39 | Dmitrii-Sh | sean-k-mooney: I am happy to make adjustments to make it work with the ongoing efforts in Nova | |
| 12:05:25 | sean-k-mooney | ill try and do that relitively soon. if we dont have time in the ptg we coudl set up a call afterward next week or the week after and see who is interested | |
| 12:05:25 | sean-k-mooney | ill try and do that relitively soon. if we dont have time in the ptg we coudl set up a call afterward next week or the week after and see who is interested | |
| 12:05:36 | sean-k-mooney | Dmitrii-Sh: have you discussed this with the neutron folks yet? | |
| 12:05:36 | sean-k-mooney | Dmitrii-Sh: have you discussed this with the neutron folks yet? | |
| 12:06:24 | Dmitrii-Sh | sean-k-mooney: ty. I think I'm going to hop in during the PCI device tracking in the Placement service discussion (added a note to ping me to the etherpad just in case). | |
| 12:06:24 | Dmitrii-Sh | sean-k-mooney: ty. I think I'm going to hop in during the PCI device tracking in the Placement service discussion (added a note to ping me to the etherpad just in case). | |
| 12:06:35 | Dmitrii-Sh | sean-k-mooney: I am working on a Neutron spec at the moment | |
| 12:06:35 | Dmitrii-Sh | sean-k-mooney: I am working on a Neutron spec at the moment | |
| 12:06:41 | sean-k-mooney | cool | |
| 12:06:41 | sean-k-mooney | cool | |
| 12:06:51 | Dmitrii-Sh | sean-k-mooney: There is some cross-project work to do in Neutron OVN mechanism driver and OVN itself | |
| 12:06:51 | Dmitrii-Sh | sean-k-mooney: There is some cross-project work to do in Neutron OVN mechanism driver and OVN itself | |
| 12:07:38 | sean-k-mooney | by the way one hack for the differnt hostname thing you could look at is changhing the [DEFAULT]/host value in the neutronl.conf for the ovs agent of the hostname value in the ovsdb for ovn | |
| 12:07:38 | sean-k-mooney | by the way one hack for the differnt hostname thing you could look at is changhing the [DEFAULT]/host value in the neutronl.conf for the ovs agent of the hostname value in the ovsdb for ovn | |
| 12:08:36 | sean-k-mooney | basically you could change those values so that you are declaring that you are manaing the network on behalf of host XYZ even if its not running on it | |
| 12:08:36 | sean-k-mooney | basically you could change those values so that you are declaring that you are manaing the network on behalf of host XYZ even if its not running on it | |
| 12:09:06 | Dmitrii-Sh | sean-k-mooney: there's a possibility that there will be 2 SmartNICs in one physical host | |
| 12:09:06 | Dmitrii-Sh | sean-k-mooney: there's a possibility that there will be 2 SmartNICs in one physical host | |
| 12:09:18 | sean-k-mooney | so assuming the agent/ovn would be runnign on the smart nic you could use the servers hostname for those values | |
| 12:09:19 | sean-k-mooney | so assuming the agent/ovn would be runnign on the smart nic you could use the servers hostname for those values | |
| 12:09:20 | Dmitrii-Sh | which creates an issue with this - I considered it as an initial hack | |
| 12:09:20 | Dmitrii-Sh | which creates an issue with this - I considered it as an initial hack | |
| 12:09:34 | sean-k-mooney | ah right | |
| 12:09:42 | sean-k-mooney | ya that does make things harder | |
| 12:09:42 | sean-k-mooney | ya that does make things harder | |
| 12:09:52 | Dmitrii-Sh | sean-k-mooney: hence the dances with PCI VPD serial numbers | |
| 12:09:52 | Dmitrii-Sh | sean-k-mooney: hence the dances with PCI VPD serial numbers | |
| 12:10:08 | Dmitrii-Sh | as we don't have another way of tracking which functions belong where on both sides | |
| 12:10:08 | Dmitrii-Sh | as we don't have another way of tracking which functions belong where on both sides | |
| 12:10:14 | sean-k-mooney | still i think we coudl encode some info in ovn or the agent config such that the ml2 driver could maintain the mapping | |
| 12:10:14 | sean-k-mooney | still i think we coudl encode some info in ovn or the agent config such that the ml2 driver could maintain the mapping | |