aws/eks-anywhere-build-tooling

[BUG]netplan apply command delete the kernel local rule table in ubuntu 22.04 build

Opened this issue · 0 comments

What happened:
When testing with fresh build ubuntu22.04 image with v1.28 k8s release built using the image-builder tool, it is observed that if any changes made with the network configuration of the deployed baremetal host needs to be applied and the netplan apply command is executed, the apply action is successful but the connection to the baremetal node gets lost, the node drops out of network post which rebooting the baremetab brings back the node back into the network.

What you expected to happen:

  • netplane apply should not disrupt connectivity on the Baremetal host.

How to reproduce it (as minimally and precisely as possible):

  • Build an image using image-build tool with 22.04 and v1.28 k8s release version.
  • Deploy an EKSA ctlplane (air-gapped/internet facing doesn't matter).
  • Perform netplan apply command.
  • ping test from the node (via iLo or DRac console) to other local node or the gateway.
  • Try pinging this node from other hosts in the same L2 network.

Anything else we need to know?:

  • When the netplan apply is executed it for some reason removes the local rule table that the kernel will primarily refer to for reaching out local network and gateways: netplan apply command step
  • At this point if the rule is added back, the connection to and from the host resume: ip rule add preference 0 lookup local
  • This problem is not noticed in vanilla ubuntu22.04 deployed on the same baremetal hardware.

Environment:

  • EKS Anywhere Release: 0.18.2
  • EKS Distro Release: ubuntu 22.04 (baremetal)