when bootstrapping criod fails to start on the bootstrap node

[RobVerduijn]

when bootstrapping okd-scos 4.15.0-0.okd-scos-2024-01-18-223523 the criod service fails to start on the bootstrap machine.

according to the logs this is because /opt/cni already exists and is not a dir

• booting the bootstrap node with rhcos 4.16.3
• the bootstrap downloads and installs centos stream coreos 416.94.202406251923-0
• reboots
  then the criod is not started with the error:

/opt/cni already exists and is not a dir

/opt/cni is a symlink to a nonexisting location

ls -als /opt/
total 4
0 drwxr-xr-x.  2 root root   17 Aug 30 16:17 .
4 drwxr-xr-x. 24 root root 4096 Aug 30 16:20 ..
0 lrwxrwxrwx.  1 root root   21 Aug 30 16:17 cni -> ../../usr/lib/opt/cni

remove-ing the symlink solves the issue, criod starts and the bootstrapping of the cluster finishes propperly.

[RobVerduijn]

Hello,

I switched to using the fedora core os image that is specified by the openshift-install binary.

The instructions for this are present in the docs, although they are buried pretty deep in there on this location

There is the generic statement that will give you the urls of the iso images

openshift-install coreos print-stream-json | grep '\.iso[^.]'

I used the one for x86_64 which worked flawless

openshift-install coreos print-stream-json | jq -r .architectures.x86_64.artifacts.metal.formats.iso.disk.location