Baremetal

Unblock a stuck oc delete bmh command

Sometimes, when deleting a BMH object with oc delete bmh -n openshift-machine-api <node name> the delete command is stuck forever. This happens because ironic is trying to decommision and delete stuff from the node itself, and does not always succeed with that.

To unblock the delete command, simply remove the object finalizer: oc patch -n openshift-machine-api <node name> -p '{"metadata":{"finalizers":null}}' --type=merge

Reprovisioning a node

Manual way (editing object YAML files)

1. for conveniency oc project openshift-machine-api
2. locate the correct secret, it’ll have the same name as the bmh with a ‘-bmc-secret’ postfix.
3. save the secret - oc get secret <bmh-name-bmc-secret> -o yaml > secret.yaml
4. save the bmh - oc get bmh <bmh-name> -o yaml > bmh.yaml
5. only then delete the bmh - oc delete bmh <bmh-name>
6. edit the secert.yaml file so it includes only the date, type, metadata.name and meteadata.namespace fields
7. edit the bmh.yaml so it includes only the oc spec, metadata.name and meteadata.namespace fields

Automatic way

• NOTE: Make sure to install oc neat if you don’t have it (https://github.com/itaysk/kubectl-neat)

1. for conveniency oc project openshift-machine-api
2. locate the correct secret, it’ll have the same name as the bmh with a ‘-bmc-secret’ postfix.
3. save the secret - oc get secret <bmh-name-bmc-secret> -o yaml | oc neat > secret.yaml
4. save the bmh - oc get bmh <bmh-name> -o yaml | oc neat > bmh.yaml
5. only then delete the bmh - oc delete bmh <bmh-name>
6. apply - oc apply -f secert.yaml and then oc apply -f bmh.yaml

The node should start reprovisioning and be ready after a while.

Rename a node

Evacuate the node:

oc adm drain NODE --ignore-daemonsets

Delete the node

oc delete node NODE

Make the DNS / hostname change if hostnames are not DNS names, you can use the following command on the node itself:

hostnamectl set-hostname NEW-NAME

Delete old certificates (which are valid only for the old name) on the node:

sudo rm /var/lib/kubelet/pki/*

Reboot the server

sudo reboot

Approve csr either use the procedure here, or look for the pending bootstrapper csr:

$ oc get csr
NAME        AGE     REQUESTOR                                                                   CONDITION
csr-6f9w7   33m     system:serviceaccount:openshift-machine-config-operator:node-bootstrapper   Approved,Issued
csr-8b9nm   40m     system:node:master-1.nnchange.lab.example.com                   Approved,Issued
csr-c6w6n   40m     system:serviceaccount:openshift-machine-config-operator:node-bootstrapper   Approved,Issued
csr-g5fpm   31m     system:node:worker-1.nnchange.lab.example.com                   Approved,Issued
csr-hsmlj   33m     system:serviceaccount:openshift-machine-config-operator:node-bootstrapper   Approved,Issued
csr-j4pct   31m     system:node:worker-0.nnchange.lab.example.com                   Approved,Issued
csr-jkkh7   39m     system:node:master-2.nnchange.lab.example.com                   Approved,Issued
csr-jnc5l   39m     system:node:master-0.nnchange.lab.example.com                   Approved,Issued
csr-nlpmv   2m27s   system:serviceaccount:openshift-machine-config-operator:node-bootstrapper   Pending
csr-pfmcl   40m     system:serviceaccount:openshift-machine-config-operator:node-bootstrapper   Approved,Issued
csr-r2d62   40m     system:serviceaccount:openshift-machine-config-operator:node-bootstrapper   Approved,Issued

$ oc adm certificate approve csr-XXX

Then accept the CSR for the node service account:

$ oc get csr
NAME        AGE     REQUESTOR                                                                   CONDITION
csr-6f9w7   35m     system:serviceaccount:openshift-machine-config-operator:node-bootstrapper   Approved,Issued
csr-7sw7b   14s     system:node:worker-a.nnchange.lab.example.com                   Pending
csr-8b9nm   41m     system:node:master-1.nnchange.lab.example.com                   Approved,Issued
csr-c6w6n   41m     system:serviceaccount:openshift-machine-config-operator:node-bootstrapper   Approved,Issued
csr-g5fpm   33m     system:node:worker-1.nnchange.lab.example.com                   Approved,Issued
csr-hsmlj   34m     system:serviceaccount:openshift-machine-config-operator:node-bootstrapper   Approved,Issued
csr-j4pct   33m     system:node:worker-0.nnchange.lab.example.com                   Approved,Issued
csr-jkkh7   41m     system:node:master-2.nnchange.lab.example.com                   Approved,Issued
csr-jnc5l   41m     system:node:master-0.nnchange.lab.example.com                   Approved,Issued
csr-nlpmv   3m51s   system:serviceaccount:openshift-machine-config-operator:node-bootstrapper   Approved,Issued
csr-pfmcl   41m     system:serviceaccount:openshift-machine-config-operator:node-bootstrapper   Approved,Issued
csr-r2d62   41m     system:serviceaccount:openshift-machine-config-operator:node-bootstrapper   Approved,Issued

$ oc adm certificate approve csr-XXX

And you should now be able to see the node with the new name:

oc get nodes