sriov-network-device-plugin
sriov-network-device-plugin copied to clipboard
k8snetworkplumbingwg
Add auxiliary network devices support
Auxiliary devices pool is created when auxNetDevice type is specified in plugin config file. Auxiliary devices discovered for each discovered network PCI device, meaning only auxiliary network devices with parent PCI device are supported. New AuxTypes selector allows to filter auxiliary device by the type. For ex., given following auxiliary devices list..
foo.bar.0
fancy.bar.0
foo.bar.1
foo.baz.0
.. and selector value ["bar"], first three devices will be added to the pool.
Auxiliary network devices accept selectors: vendors, devices, drivers, pfNames, rootDevices, linkTypes, isRdma.
NOTE patchset built on top of #441
Pull Request Test Coverage Report for Build 2910194077
Warning: This coverage report may be inaccurate.
This pull request's base commit is no longer the HEAD commit of its target branch. This means it includes changes from outside the original pull request, including, potentially, unrelated coverage changes.
- For more information on this, see Tracking coverage changes with pull request builds.
- To avoid this issue with future PRs, see these Recommended CI Configurations.
- For a quick fix, rebase this PR at GitHub. Your next report should be accurate.
Details
- 214 of 288 (74.31%) changed or added relevant lines in 15 files are covered.
- 106 unchanged lines in 9 files lost coverage.
- Overall coverage increased (+2.7%) to 79.949%
| Changes Missing Coverage | Covered Lines | Changed/Added Lines | % |
|---|---|---|---|
| pkg/infoprovider/genericInfoProvider.go | 2 | 3 | 66.67% |
| pkg/netdevice/netDeviceProvider.go | 2 | 3 | 66.67% |
| pkg/utils/rdma_provider.go | 0 | 3 | 0.0% |
| pkg/utils/testing.go | 1 | 4 | 25.0% |
| pkg/devices/rdma.go | 5 | 9 | 55.56% |
| pkg/factory/factory.go | 24 | 28 | 85.71% |
| pkg/auxnetdevice/auxNetDevice.go | 32 | 38 | 84.21% |
| pkg/utils/netlink_provider.go | 0 | 6 | 0.0% |
| pkg/utils/sriovnet_provider.go | 0 | 12 | 0.0% |
| pkg/utils/utils.go | 15 | 29 | 51.72% |
| <!-- | Total: | 214 | 288 |
| Files with Coverage Reduction | New Missed Lines | % |
|---|---|---|
| pkg/utils/sriovnet_provider.go | 1 | 28.57% |
| pkg/accelerator/accelDevice.go | 2 | 84.62% |
| pkg/factory/factory.go | 8 | 89.93% |
| pkg/netdevice/netResourcePool.go | 8 | 87.14% |
| pkg/accelerator/accelDeviceProvider.go | 9 | 87.5% |
| pkg/netdevice/netDeviceProvider.go | 14 | 87.18% |
| pkg/resources/deviceSelectors.go | 15 | 88.13% |
| pkg/netdevice/pciNetDevice.go | 16 | 69.84% |
| pkg/utils/utils.go | 33 | 82.91% |
| <!-- | Total: | 106 |
| Totals | |
|---|---|
| Change from base Build 2733786348: | 2.7% |
| Covered Lines: | 1870 |
| Relevant Lines: | 2339 |
💛 - Coveralls
@DmytroLinkin
i tested out your PR and added some comments.
i was able to successfully spin up a pod with SF + RDMA resources
one issue i found is that the injected env variable (PCIDEVICE_NVIDIA_COM_MLNX_SF) is showing the parent PF instead of the allocated SF.
/ # env
KUBERNETES_SERVICE_PORT=443
KUBERNETES_PORT=tcp://10.96.0.1:443
HOSTNAME=testpod1
SHLVL=1
HOME=/root
PCIDEVICE_NVIDIA_COM_MLNX_SF=0000:03:00.0
TERM=xterm
KUBERNETES_PORT_443_TCP_ADDR=10.96.0.1
PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
KUBERNETES_PORT_443_TCP_PORT=443
KUBERNETES_PORT_443_TCP_PROTO=tcp
KUBERNETES_SERVICE_PORT_HTTPS=443
KUBERNETES_PORT_443_TCP=tcp://10.96.0.1:443
KUBERNETES_SERVICE_HOST=10.96.0.1
PWD=/
@DmytroLinkin
i tested out your PR and added some comments.
i was able to successfully spin up a pod with SF + RDMA resources
one issue i found is that the injected env variable (PCIDEVICE_NVIDIA_COM_MLNX_SF) is showing the parent PF instead of the allocated SF.
/ # env KUBERNETES_SERVICE_PORT=443 KUBERNETES_PORT=tcp://10.96.0.1:443 HOSTNAME=testpod1 SHLVL=1 HOME=/root PCIDEVICE_NVIDIA_COM_MLNX_SF=0000:03:00.0 TERM=xterm KUBERNETES_PORT_443_TCP_ADDR=10.96.0.1 PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin KUBERNETES_PORT_443_TCP_PORT=443 KUBERNETES_PORT_443_TCP_PROTO=tcp KUBERNETES_SERVICE_PORT_HTTPS=443 KUBERNETES_PORT_443_TCP=tcp://10.96.0.1:443 KUBERNETES_SERVICE_HOST=10.96.0.1 PWD=/
It is genericInfoProvider which returns pciAddress of the device. I changed it to return pciAddr/auxiliary
minor nit on doc update.
otherwise LGTM. did some basic testing for the latest iteration on my dev setup for both SF and SR-IOV.
I have re-triggered Mellanox CI.
@SchSeba @Eoghan1232 PTAL, id like to try and get this merged soon.
i believe the PR Is in a good state and LGTM'd it.