From bb7d903e1ac7e28d43636246d236eb70982e5601 Mon Sep 17 00:00:00 2001
From: Matt Pryor <matt@stackhpc.com>
Date: Wed, 9 Aug 2023 16:42:16 +0100
Subject: [PATCH] Add support for nodeVolumeDetachTimeout and
 nodeDeletionTimeout (#57)

---
 .../control-plane/kubeadm-control-plane.yaml        |  6 +++---
 .../templates/node-group/machine-deployment.yaml    |  2 ++
 charts/openstack-cluster/values.yaml                | 13 ++++++++++++-
 3 files changed, 17 insertions(+), 4 deletions(-)

diff --git a/charts/openstack-cluster/templates/control-plane/kubeadm-control-plane.yaml b/charts/openstack-cluster/templates/control-plane/kubeadm-control-plane.yaml
index 57ee9a2..31fe40a 100644
--- a/charts/openstack-cluster/templates/control-plane/kubeadm-control-plane.yaml
+++ b/charts/openstack-cluster/templates/control-plane/kubeadm-control-plane.yaml
@@ -87,9 +87,9 @@ spec:
       apiVersion: infrastructure.cluster.x-k8s.io/v1alpha6
       name: {{ include "openstack-cluster.controlplane.mt.name" . }}
       namespace: {{ .Release.Namespace }}
-    {{- with .Values.controlPlane.nodeDrainTimeout }}
-    nodeDrainTimeout: {{ . }}
-    {{- end }}
+    nodeDrainTimeout: {{ .Values.controlPlane.nodeDrainTimeout }}
+    nodeVolumeDetachTimeout: {{ .Values.controlPlane.nodeVolumeDetachTimeout }}
+    nodeDeletionTimeout: {{ .Values.controlPlane.nodeDeletionTimeout }}
   kubeadmConfigSpec: {{
     include
       "openstack-cluster.controlplane.kubeadmConfigSpec"
diff --git a/charts/openstack-cluster/templates/node-group/machine-deployment.yaml b/charts/openstack-cluster/templates/node-group/machine-deployment.yaml
index 7885b34..1a70a62 100644
--- a/charts/openstack-cluster/templates/node-group/machine-deployment.yaml
+++ b/charts/openstack-cluster/templates/node-group/machine-deployment.yaml
@@ -61,4 +61,6 @@ spec:
         kind: OpenStackMachineTemplate
         name: {{ include "openstack-cluster.nodegroup.mt.name" (list $ $nodeGroup) }}
       nodeDrainTimeout: {{ $nodeGroup.nodeDrainTimeout }}
+      nodeVolumeDetachTimeout: {{ $nodeGroup.nodeVolumeDetachTimeout }}
+      nodeDeletionTimeout: {{ $nodeGroup.nodeDeletionTimeout }}
 {{- end }}
diff --git a/charts/openstack-cluster/values.yaml b/charts/openstack-cluster/values.yaml
index 4b4342e..cf39836 100644
--- a/charts/openstack-cluster/values.yaml
+++ b/charts/openstack-cluster/values.yaml
@@ -152,8 +152,12 @@ controlPlane:
     # availabilityZone:
   # The time to wait for a node to finish draining before it can be removed
   nodeDrainTimeout: 5m0s
+  # The time to wait for a node to detach all volumes before it can be removed
+  nodeVolumeDetachTimeout: 5m0s
+  # The time to wait for the node resource to be deleted in Kubernetes when a
+  # machine is marked for deletion
+  nodeDeletionTimeout: 5m0s
   # The rollout strategy to use for the control plane nodes
-
   # By default, the strategy allows the control plane to begin provisioning new nodes
   # without first tearing down old ones
   rolloutStrategy:
@@ -203,6 +207,8 @@ controlPlane:
     spec:
       # By default, unhealthy control plane nodes are always remediated
       maxUnhealthy: 100%
+      # If a node takes longer than 10 mins to startup, remediate it
+      nodeStartupTimeout: 10m0s
       # By default, consider a control plane node that has not been Ready
       # for more than 5 mins unhealthy
       unhealthyConditions:
@@ -241,6 +247,11 @@ nodeGroupDefaults:
     # availabilityZone:
   # The time to wait for a node to finish draining before it can be removed
   nodeDrainTimeout: 5m0s
+  # The time to wait for a node to detach all volumes before it can be removed
+  nodeVolumeDetachTimeout: 5m0s
+  # The time to wait for the node resource to be deleted in Kubernetes when a
+  # machine is marked for deletion
+  nodeDeletionTimeout: 5m0s
   # The rollout strategy to use for the node group
   # By default, this is set to do a rolling update within the existing resource envelope
   # of the node group, even if that means the node group temporarily has zero nodes