From 4cfa74a33caefaa5510f51ed64b6306e61f1b8f7 Mon Sep 17 00:00:00 2001
From: sethp-nr <30441101+sethp-nr@users.noreply.github.com>
Date: Wed, 26 Feb 2020 08:48:05 -0800
Subject: [PATCH] Report all healthcheck errors

Before:

```
E0225 23:47:06.883670       8 kubeadm_control_plane_controller.go:577] controllers/KubeadmControlPlane "msg"="waiting for control plane to pass etcd health check before adding an additional control plane machine" "error"="there are 3 control plane nodes, but 0 etcd members" "cluster"="test" "kubeadmControlPlane"="test" "namespace"="test"
```

After:

```
E0226 01:48:29.726200     166 kubeadm_control_plane_controller.go:577] controllers/KubeadmControlPlane "msg"="waiting for control plane to pass etcd health check before adding an additional control plane machine" "error"="[could not check etcd member health, node \"ip-10-0-0-3.ec2.internal\": failed to create etcd client: unable to create etcd client: context deadline exceeded, node \"ip-10-0-0-197.ec2.internal\": failed to create etcd client: unable to create etcd client: context deadline exceeded, node \"ip-10-0-0-59.ec2.internal\": failed to create etcd client: unable to create etcd client: context deadline exceeded]" "cluster"="test" "kubeadmControlPlane"="test" "namespace"="test"
```
---
 controlplane/kubeadm/internal/cluster.go | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/controlplane/kubeadm/internal/cluster.go b/controlplane/kubeadm/internal/cluster.go
index 18571b31cabc..5d365798d3cc 100644
--- a/controlplane/kubeadm/internal/cluster.go
+++ b/controlplane/kubeadm/internal/cluster.go
@@ -121,11 +121,11 @@ type healthCheck func(context.Context) (healthCheckResult, error)
 // healthCheck will run a generic health check function and report any errors discovered.
 // It does some additional validation to make sure there is a 1;1 match between nodes and machines.
 func (m *ManagementCluster) healthCheck(ctx context.Context, check healthCheck, clusterKey types.NamespacedName, controlPlaneName string) error {
+	var errorList []error
 	nodeChecks, err := check(ctx)
 	if err != nil {
-		return err
+		errorList = append(errorList, err)
 	}
-	errorList := []error{}
 	for nodeName, err := range nodeChecks {
 		if err != nil {
 			errorList = append(errorList, fmt.Errorf("node %q: %v", nodeName, err))
@@ -388,6 +388,10 @@ func (c *cluster) etcdIsHealthy(ctx context.Context) (healthCheckResult, error)
 		}
 	}
 
+	if len(response) > 0 {
+		return response, errors.New("could not check etcd member health")
+	}
+
 	// Check that there is exactly one etcd member for every control plane machine.
 	// There should be no etcd members added "out of band.""
 	if len(controlPlaneNodes.Items) != len(knownMemberIDSet) {