changing "resource-list" config doesn't take effect

Bug #1913492 reported by Andrey Grebennikov
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
SR-IOV Network Device Plugin Charm
Triaged
Medium
Unassigned

Bug Description

Deployed Charmed Kubernetes 1.20 on baremetal.

Workers are running on physical machines
Activated SRIOV on the interface ens1f1
Deployed SRIOV-CNI and SRIOV-network-device-plugin.

Based on the logs from the pod of sriov-network-device-plugin daemonset it recognizes the virtual functions but doesn't add them to the pool of VFs:

I0127 17:10:37.391927 25 manager.go:98] ResourceList: [{ResourcePrefix: ResourceName:intel_sriov_netdevice IsRdma:false Selectors:{Vendors:[8086] Devices:[154c] Drivers:[i40evf] PfNames:[] LinkTypes:[] DDPProfiles:[]}}]
I0127 17:10:37.391984 25 manager.go:174] validating resource name "intel.com/intel_sriov_netdevice"
I0127 17:10:37.391990 25 main.go:60] Discovering host network devices
I0127 17:10:37.391998 25 manager.go:190] discovering host network devices
I0127 17:10:37.445129 25 manager.go:220] discoverDevices(): device found: 0000:12:00.0 02 Intel Corporation Ethernet Controller XXV710 for 25GbE ...
I0127 17:10:37.445644 25 manager.go:290] ens1f0 added to linkWatchList
I0127 17:10:37.445681 25 manager.go:220] discoverDevices(): device found: 0000:12:00.1 02 Intel Corporation Ethernet Controller XXV710 for 25GbE ...
I0127 17:10:37.445944 25 manager.go:290] ens1f1 added to linkWatchList
I0127 17:10:37.445972 25 manager.go:220] discoverDevices(): device found: 0000:12:02.0 02 Intel Corporation Ethernet Virtual Function 700 Series
I0127 17:10:37.446284 25 manager.go:220] discoverDevices(): device found: 0000:12:02.1 02 Intel Corporation Ethernet Virtual Function 700 Series
I0127 17:10:37.446720 25 manager.go:220] discoverDevices(): device found: 0000:12:02.2 02 Intel Corporation Ethernet Virtual Function 700 Series
I0127 17:10:37.447117 25 manager.go:220] discoverDevices(): device found: 0000:12:02.3 02 Intel Corporation Ethernet Virtual Function 700 Series
I0127 17:10:37.447552 25 manager.go:220] discoverDevices(): device found: 0000:12:0a.0 02 Intel Corporation Ethernet Virtual Function 700 Series
I0127 17:10:37.447891 25 manager.go:220] discoverDevices(): device found: 0000:12:0a.1 02 Intel Corporation Ethernet Virtual Function 700 Series
I0127 17:10:37.448266 25 manager.go:220] discoverDevices(): device found: 0000:12:0a.2 02 Intel Corporation Ethernet Virtual Function 700 Series
I0127 17:10:37.448662 25 manager.go:220] discoverDevices(): device found: 0000:12:0a.3 02 Intel Corporation Ethernet Virtual Function 700 Series
I0127 17:10:37.449112 25 manager.go:220] discoverDevices(): device found: 0000:37:00.0 02 Intel Corporation Ethernet Controller XXV710 for 25GbE ...

Trying to change the config of the charm to change the selector to only use PfName as the filter:
juju config sriov-network-device-plugin resource-list='[{"resourceName": "intel_sriov_netdevice","selectors": {"PfNames":["ens1f1"]}}]'

I see that the configmap got updated, however daemonset didn't reflect the change and still shows the original selector values.

Revision history for this message
George Kraft (cynerva) wrote :

Thanks for the report. I can reproduce this.

You can work around this by manually restarting the DaemonSet:

kubectl rollout restart -n $MODEL daemonset/sriov-network-device-plugin

Changed in charm-sriov-network-device-plugin:
importance: Undecided → Medium
status: New → Triaged
Revision history for this message
Andrey Grebennikov (agrebennikov) wrote :

yeah I could just kill respective containers and they will be recreated picking up updated configmap.
I guess this is a problem rather with the device-plugin itself, though maybe we should trigger the restart of the daemonset via the charm then.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.