WIP: annealing

Author: justinsb

cmd/kops/anneal_cluster.go

@@ -0,0 +1,72 @@
+package main
+
+import (
+	"context"
+	"fmt"
+	"io"
+
+	"k8s.io/kops/cmd/kops/util"
+	"k8s.io/kops/pkg/apis/kops"
+)
+
+func AnnealCluster(ctx context.Context, f *util.Factory, out io.Writer, c *UpdateClusterOptions) error {
+	fmt.Fprintf(out, "Updating control plane configuration\n")
+	{
+		opt := *c
+		opt.Anneal = false // Prevent infinite loop
+		opt.InstanceGroupRoles = []string{
+			string(kops.InstanceGroupRoleAPIServer),
+			string(kops.InstanceGroupRoleControlPlane),
+		}
+		if _, err := RunUpdateCluster(ctx, f, out, &opt); err != nil {
+			return err
+		}
+	}
+
+	fmt.Fprintf(out, "Doing rolling-update for control plane\n")
+	{
+		opt := &RollingUpdateOptions{}
+		opt.InitDefaults()
+		opt.ClusterName = c.ClusterName
+		opt.InstanceGroupRoles = []string{
+			string(kops.InstanceGroupRoleAPIServer),
+			string(kops.InstanceGroupRoleControlPlane),
+		}
+		opt.Yes = c.Yes
+		if err := RunRollingUpdateCluster(ctx, f, out, opt); err != nil {
+			return err
+		}
+	}
+
+	fmt.Fprintf(out, "Updating node configuration\n")
+	{
+		opt := *c
+		opt.Anneal = false // Prevent infinite loop
+		// TODO: or just do all roles?
+		opt.InstanceGroupRoles = []string{
+			string(kops.InstanceGroupRoleBastion),
+			string(kops.InstanceGroupRoleNode),
+		}
+		if _, err := RunUpdateCluster(ctx, f, out, &opt); err != nil {
+			return err
+		}
+	}
+
+	fmt.Fprintf(out, "Doing rolling-update for nodes\n")
+	{
+		opt := &RollingUpdateOptions{}
+		opt.InitDefaults()
+		opt.ClusterName = c.ClusterName
+		// TODO: or just do all roles?
+		opt.InstanceGroupRoles = []string{
+			string(kops.InstanceGroupRoleBastion),
+			string(kops.InstanceGroupRoleNode),
+		}
+		opt.Yes = c.Yes
+		if err := RunRollingUpdateCluster(ctx, f, out, opt); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}

cmd/kops/update_cluster.go

@@ -96,6 +96,9 @@ type UpdateClusterOptions struct {
 	// The goal is that the cluster can keep running even during more disruptive
 	// infrastructure changes.
 	Prune bool
+
+	// TODO
+	Anneal bool
 }
 
 func (o *UpdateClusterOptions) InitDefaults() {
@@ -108,6 +111,7 @@ func (o *UpdateClusterOptions) InitDefaults() {
 	o.CreateKubecfg = true
 
 	o.Prune = false
+	o.Anneal = false
 
 	o.RunTasksOptions.InitDefaults()
 }
@@ -163,7 +167,7 @@ func NewCmdUpdateCluster(f *util.Factory, out io.Writer) *cobra.Command {
 	cmd.RegisterFlagCompletionFunc("lifecycle-overrides", completeLifecycleOverrides)
 
 	cmd.Flags().BoolVar(&options.Prune, "prune", options.Prune, "Delete old revisions of cloud resources that were needed during an upgrade")
-
+	cmd.Flags().BoolVar(&options.Anneal, "anneal", options.Anneal, "Anneal the cluster by rolling the control plane and nodes sequentially")
 	return cmd
 }
 
@@ -183,6 +187,13 @@ type UpdateClusterResults struct {
 }
 
 func RunUpdateCluster(ctx context.Context, f *util.Factory, out io.Writer, c *UpdateClusterOptions) (*UpdateClusterResults, error) {
+	if c.Anneal {
+		if !c.Yes {
+			return nil, fmt.Errorf("--anneal is only supported with --yes")
+		}
+		return nil, AnnealCluster(ctx, f, out, c)
+	}
+
 	results := &UpdateClusterResults{}
 
 	isDryrun := false