bond-cni pod reboot

[ognjen011]

Could anyone help me understand if this is a normal behaviour. I setup deployment two pods as a test. I have two network interfaces ens8f1np1 and ens8f0np0 with VF's enabled. I setup a bond-cni and i see two interfaces are created in the pods including bond0 interface. Now when i shut down one physical link everything is working but after about 45s and than pods redeploy/restart. I dont want the pods to redeploy since the other VF should be enabled in via the bond interface if i understand this.

the other option i would like to try is active-actve mode

{
"cniVersion": "0.3.1",
"name": "sriov-dpdk-bond",
"type": "bond",
"name": "bond1",
"mode": "802.3ad",
"failOverMac": 0,
"linksInContainer": true,
"miimon": "100",
"mtu": 1500,
"links": [
{"name": "net1"},
{"name": "net2"}
],

I also see pods redploy when i use 802.3ad

[SchSeba]

The bond-cni for sure is not the one that reset the pod.
you need to check if the nic you move to down is not the k8s API communication nic and your node goes down.

is this the only pod that gets redeploy? or others in the same node?

[ognjen011]

Yes i did bring it down to test a failover I thought when it fails over to second nic it can be done without redeployment of a pod. Thanks for the help. I decided a better option is to use ovs and achieve VF failover that way without redeployment.

[SchSeba]

ok great!

@ognjen011 can we close this issue?

[ognjen011]

Yes we can close it. Thanks again.