/run/cluster-api/bootstrap-success.complete is always created, even in case of failures

[anmazzotti]

What happened:

The bootstrap provider is expected to create the /run/cluster-api/bootstrap-success.complete sentinel file upon successful bootstrap.

Normally the runcmd command list will look like:

runcmd:
  - 'curl -sfL https://get.rke2.io | INSTALL_RKE2_VERSION=v1.28.4+rke2r1 sh -s - server'
  - 'systemctl enable rke2-server.service'
  - 'systemctl start rke2-server.service'
  - 'mkdir /run/cluster-api' 
  - 'echo success > /run/cluster-api/bootstrap-success.complete'

Since cloud-init will run all commands in the list (and collect errors later), the last command will always be executed.

What did you expect to happen:
The /run/cluster-api/bootstrap-success.complete file should not be created if any of the runcmd fails.

How to reproduce it:
Bootstrap any rke2 cluster using this provider

Anything else you would like to add:
To add complexity to this issue, the preRKE2Commands field can also be used to propagate arbitrary commands.
For example:

runcmd:
  - "mkdir -p /var/lib/rancher/rke2/server/manifests/ && ctr images pull ghcr.io/kube-vip/kube-vip:v0.6.4 && ctr run --rm --net-host ghcr.io/kube-vip/kube-vip:v0.6.4 vip /kube-vip manifest daemonset --arp --interface $(ip -4 -j route list default | jq -r .[0].dev) --address 192.168.122.50 --controlplane --leaderElection --taint --services --inCluster | tee /var/lib/rancher/rke2/server/manifests/kube-vip.yaml"
  - 'curl -sfL https://get.rke2.io | INSTALL_RKE2_VERSION=v1.28.4+rke2r1 sh -s - server'
  - 'systemctl enable rke2-server.service'
  - 'systemctl start rke2-server.service'
  - 'mkdir /run/cluster-api' 
  - 'echo success > /run/cluster-api/bootstrap-success.complete'

Environment:

• rke provider version: v0.2.2