Do not drain before provision status check

Signed-off-by: Markus Blaschke <[email protected]>

Author: mblaschke

autopilot/lib.azure.go

@@ -80,7 +80,7 @@ func (r *AzureK8sAutopilot) azureVmRepair(contextLogger *log.Entry, nodeInfo k8s
 	return err
 }
 
-func (r *AzureK8sAutopilot) azureVmssInstanceUpdate(contextLogger *log.Entry, nodeInfo k8s.NodeInfo, doReimage bool) error {
+func (r *AzureK8sAutopilot) azureVmssInstanceUpdate(contextLogger *log.Entry, node *k8s.Node, nodeInfo k8s.NodeInfo, doReimage bool) error {
 	var err error
 
 	vmssClient := compute.NewVirtualMachineScaleSetsClient(nodeInfo.Subscription)
@@ -102,6 +102,11 @@ func (r *AzureK8sAutopilot) azureVmssInstanceUpdate(contextLogger *log.Entry, no
 
 	r.sendNotificationf("trigger automatic update of K8s node %v", nodeInfo.NodeName)
 
+	// drain node
+	if err := r.k8sDrainNode(contextLogger, node); err != nil {
+		return fmt.Errorf("node %s failed to drain: %v", node.Name, err)
+	}
+
 	// trigger update call
 	contextLogger.Info("scheduling Azure VMSS instance update")
 	vmssInstanceUpdateOpts := compute.VirtualMachineScaleSetVMInstanceRequiredIDs{

autopilot/task.update.go

@@ -85,17 +85,10 @@ vmssLoop:
 						return
 					}
 
-					// drain node
-					if err := r.k8sDrainNode(contextLogger, node); err != nil {
-						vmssInstanceContextLogger.Errorf("node %s failed to drain: %v", node.Name, err)
-						r.updateNodeLock(vmssInstanceContextLogger, node, r.Config.Update.LockDurationError)
-						continue vmssInstanceLoop
-					}
-
 					// trigger Azure VMSS instance update
 					r.prometheus.update.count.WithLabelValues().Inc()
 					doReimage := r.Config.Update.AzureVmssAction == "update+reimage"
-					err = r.azureVmssInstanceUpdate(vmssInstanceContextLogger, *nodeInfo, doReimage)
+					err = r.azureVmssInstanceUpdate(vmssInstanceContextLogger, node, *nodeInfo, doReimage)
 
 					if err != nil {
 						r.prometheus.general.errors.WithLabelValues("azure").Inc()