Set machine healthcheck per machine set (#261)

Signed-off-by: Jakub Stejskal <[email protected]>
skodjob · Oct 30, 2023 · e281a00 · e281a00
1 parent 2fa129b
commit e281a00
Show file tree

Hide file tree

Showing 3 changed files with 13 additions and 3 deletions.
diff --git a/install/roles/tealc/tasks/infra-setup/configure_worker_cluster.yaml b/install/roles/tealc/tasks/infra-setup/configure_worker_cluster.yaml
@@ -66,7 +66,15 @@
     kubeconfig: "{{ kubeconfig_path }}/{{ worker.name }}"
     state: present
     verify_ssl: no
-    template: "templates/worker-nodes-health-check.yaml"
+    template: "templates/worker-nodes-health-check.yaml.j2"
+  loop: "{{ worker.machine_pools }}"
+  loop_control:
+    loop_var: machine_pool
+  vars:
+    node_type: "{{ machine_pool.node_type }}"
+    max_unhealthy: "{{ machine_pool.max_unhealthy }}"
+    clusterName: "{{ worker.name }}"
+    clusterId: "{{ get_cluster_id_out.stdout }}"
 
 - name: "Set anonymous pull rights on {{ worker.name }}"
   shell: "oc patch --type=merge --patch='{ \"spec\": { \"registrySources\": { \"insecureRegistries\": [ \"default-route-openshift-image-registry.apps.{{ worker.name }}.{{ cluster_domain }}\" ] } } }' image.config.openshift.io/cluster"

diff --git a/.../templates/worker-nodes-health-check.yaml → ...mplates/worker-nodes-health-check.yaml.j2 b/.../templates/worker-nodes-health-check.yaml → ...mplates/worker-nodes-health-check.yaml.j2
@@ -2,18 +2,20 @@
 apiVersion: machine.openshift.io/v1beta1
 kind: MachineHealthCheck
 metadata:
-  name: worker-machines-health-check
+  name: worker-{{ node_type }}-machines-health-check
   namespace: openshift-machine-api
 spec:
   selector:
     matchLabels:
       machine.openshift.io/cluster-api-machine-role: worker
       machine.openshift.io/cluster-api-machine-type: worker
+      machine.openshift.io/cluster-api-machineset: {{ clusterId }}-{{ node_type }}-worker-0
   unhealthyConditions:
     - type: Ready
       status: Unknown
       timeout: 600s
     - type: Ready
       status: 'False'
       timeout: 600s
-  maxUnhealthy: 80%
+  nodeStartupTimeout: 20m
+  maxUnhealthy: {{ max_unhealthy }}
diff --git a/install/secrets/clusters.yaml b/install/secrets/clusters.yaml