Random pool node suffixes (#9)

Also only place controlplanes in a placement group.
airfocusio · May 8, 2023 · 1a9364c · 1a9364c
1 parent 993d8fe
commit 1a9364c
Show file tree

Hide file tree

Showing 10 changed files with 23 additions and 52 deletions.
diff --git a/README.md b/README.md
@@ -5,7 +5,7 @@
 This CLI tool provides an easy way to manage [Talos](https://talos.dev/) powered [Kubernetes](https://kubernetes.io/) clusters on the [Hetzner Cloud](https://www.hetzner.com/cloud). Bootstrapping a new cluster performs the following steps:
 
 * Create private network `10.0.0.0/16` for inter-node communication
-* Create placement group to ensure nodes to not run on the same physical machine
+* Create placement group to ensure controlplane nodes to not run on the same physical machine
 * Create load balancer to access the controlplane nodes Kubernetes API server (port `6443`) or Talos API server (port `50000`)
 * Create firewall rules to block access to nodes from outside of the private network
 * Create first controlplane node
@@ -21,9 +21,9 @@ cd my-cluster
 
 export HCLOUD_TOKEN=...
 # bootstrap cluster
-hcloud-talos -v bootstrap-cluster --talos-version=1.2.9 --kubernetes-version=1.25.5 my-cluster controlplane-1
+hcloud-talos -v bootstrap-cluster --talos-version=1.2.9 --kubernetes-version=1.25.5 my-cluster controlplane-%id%
 
 # add more nodes
-hcloud-talos -v add-node --talos-version=1.2.9 controlplane-2 --controlplane
-hcloud-talos -v add-node --talos-version=1.2.9 worker-1
+hcloud-talos -v add-node --talos-version=1.2.9 controlplane-%id% --controlplane
+hcloud-talos -v add-node --talos-version=1.2.9 worker-%id%
 ```
diff --git a/cmd/reconcile_pool.go b/cmd/reconcile_pool.go
@@ -11,7 +11,6 @@ var (
 	reconcilePoolCmdServerType     string
 	reconcilePoolCmdNodeNamePrefix string
 	reconcilePoolCmdNodeCount      int
-	reconcilePoolCmdForce          bool
 	reconcilePoolCmdTalosVersion   string
 	reconcilePoolCmd               = &cobra.Command{
 		Use:   "reconcile-pool [pool-name]",
@@ -24,7 +23,6 @@ var (
 				NodeNamePrefix: reconcilePoolCmdNodeNamePrefix,
 				NodeCount:      reconcilePoolCmdNodeCount,
 				ServerType:     reconcilePoolCmdServerType,
-				Force:          reconcilePoolCmdForce,
 				PoolName:       args[0],
 				TalosVersion:   reconcilePoolCmdTalosVersion,
 			})
@@ -35,7 +33,6 @@ var (
 
 func init() {
 	reconcilePoolCmd.Flags().StringVarP(&reconcilePoolCmdConfigFile, "config", "c", defaultConfigFile, "")
-	reconcilePoolCmd.Flags().BoolVar(&reconcilePoolCmdForce, "force", false, "")
 	reconcilePoolCmd.Flags().StringVar(&reconcilePoolCmdServerType, "server-type", "cx21", "")
 	reconcilePoolCmd.Flags().StringVar(&reconcilePoolCmdNodeNamePrefix, "node-name-prefix", "worker", "")
 	reconcilePoolCmd.Flags().IntVar(&reconcilePoolCmdNodeCount, "node-count", 1, "")

diff --git a/e2etests/01_bootstrap_cluster_test.go b/e2etests/01_bootstrap_cluster_test.go
@@ -12,7 +12,7 @@ func TestBootstrapCluster(t *testing.T) {
 		ConfigFile:        configFile,
 		ClusterName:       clusterName,
 		ServerType:        "cx21",
-		NodeName:          "controlplane-1",
+		NodeName:          "controlplane-%id%",
 		Location:          "nbg1",
 		NetworkZone:       "eu-central",
 		Token:             hcloudToken,

diff --git a/e2etests/02_add_node_test.go b/e2etests/02_add_node_test.go
@@ -11,7 +11,7 @@ func TestAddNode(t *testing.T) {
 	_, err := internal.AddNode(&logger, clusterDir, internal.AddNodeOpts{
 		ConfigFile:   configFile,
 		ServerType:   "cx21",
-		NodeName:     "worker-1",
+		NodeName:     "worker",
 		TalosVersion: talosVersion,
 	})
 	assert.NoError(t, err)

diff --git a/e2etests/04_delete_node_test.go b/e2etests/04_delete_node_test.go
@@ -10,7 +10,7 @@ import (
 func TestDeleteNode(t *testing.T) {
 	err := internal.DeleteNode(&logger, clusterDir, internal.DeleteNodeOpts{
 		ConfigFile: configFile,
-		NodeName:   "worker-1",
+		NodeName:   "worker",
 		Force:      true,
 	})
 	assert.NoError(t, err)

diff --git a/internal/add_node.go b/internal/add_node.go
@@ -40,9 +40,13 @@ func AddNode(logger *utils.Logger, dir string, opts AddNodeOpts) (*hcloud.Server
 		return nil, err
 	}
 
-	placementGroup, err := clients.HcloudEnsurePlacementGroup(cl, nodePlacementGroupTemplate(cl), false)
-	if err != nil {
-		return nil, err
+	var placementGroup *hcloud.PlacementGroup
+	if opts.Controlplane {
+		controlplanePlacementGroup, err := clients.HcloudEnsurePlacementGroup(cl, controlplanePlacementGroupTemplate(cl), false)
+		if err != nil {
+			return nil, err
+		}
+		placementGroup = controlplanePlacementGroup
 	}
 
 	var nodeTemplate clients.HcloudServerCreateFromImageOpts

diff --git a/internal/bootstrap_cluster.go b/internal/bootstrap_cluster.go
@@ -65,7 +65,7 @@ func BootstrapCluster(logger *utils.Logger, dir string, opts BootstrapClusterOpt
 		return err
 	}
 
-	placementGroup, err := clients.HcloudEnsurePlacementGroup(cl, nodePlacementGroupTemplate(cl), true)
+	controlplanePlacementGroup, err := clients.HcloudEnsurePlacementGroup(cl, controlplanePlacementGroupTemplate(cl), true)
 	if err != nil {
 		return err
 	}
@@ -91,7 +91,7 @@ func BootstrapCluster(logger *utils.Logger, dir string, opts BootstrapClusterOpt
 	if err != nil {
 		return err
 	}
-	controlplaneServer, err := clients.HcloudCreateServerFromImage(cl, network, placementGroup, controlplaneNodeTemplate)
+	controlplaneServer, err := clients.HcloudCreateServerFromImage(cl, network, controlplanePlacementGroup, controlplaneNodeTemplate)
 	if err != nil {
 		return err
 	}

diff --git a/internal/reconcile_pool.go b/internal/reconcile_pool.go
@@ -2,7 +2,6 @@ package internal
 
 import (
 	"fmt"
-	"strings"
 
 	"github.com/airfocusio/hcloud-talos/internal/cluster"
 	"github.com/airfocusio/hcloud-talos/internal/utils"
@@ -15,7 +14,6 @@ type ReconcilePoolOpts struct {
 	NodeNamePrefix string
 	NodeCount      int
 	ServerType     string
-	Force          bool
 	TalosVersion   string
 }
 
@@ -51,10 +49,7 @@ func ReconcilePool(logger *utils.Logger, dir string, opts ReconcilePoolOpts) err
 
 		nodeCountDiff := len(poolServers) - opts.NodeCount
 		if nodeCountDiff < 0 {
-			nodeName, err := findNextNodeName(cl, opts.NodeNamePrefix, poolServers)
-			if err != nil {
-				return err
-			}
+			nodeName := opts.NodeNamePrefix + "-%id%"
 			_, err = AddNode(logger, cl.Dir, AddNodeOpts{
 				ConfigFile:   opts.ConfigFile,
 				ServerType:   opts.ServerType,
@@ -65,16 +60,6 @@ func ReconcilePool(logger *utils.Logger, dir string, opts ReconcilePoolOpts) err
 			if err != nil {
 				return err
 			}
-		} else if nodeCountDiff > 0 {
-			server := poolServers[len(poolServers)-1]
-			err := DeleteNode(logger, cl.Dir, DeleteNodeOpts{
-				ConfigFile: opts.ConfigFile,
-				NodeName:   strings.TrimPrefix(server.Name, cl.Config.ClusterName+"-"),
-				Force:      opts.Force,
-			})
-			if err != nil {
-				return err
-			}
 		} else {
 			logger.Debug.Printf("Pool %s/%s already has %d of %d nodes\n", cl.Config.ClusterName, opts.PoolName, len(poolServers), opts.NodeCount)
 			break
@@ -83,20 +68,3 @@ func ReconcilePool(logger *utils.Logger, dir string, opts ReconcilePoolOpts) err
 
 	return nil
 }
-
-func findNextNodeName(cl *cluster.Cluster, nodeNamePrefix string, poolServers []*hcloud.Server) (string, error) {
-	for i := 1; i <= 1000; i++ {
-		exists := false
-		proposedNodeName := fmt.Sprintf("%s-%d", nodeNamePrefix, i)
-		for _, server := range poolServers {
-			if server.Name == nodeName(cl, proposedNodeName) {
-				exists = true
-				break
-			}
-		}
-		if !exists {
-			return proposedNodeName, nil
-		}
-	}
-	return "", fmt.Errorf("unable to find next node name")
-}
diff --git a/internal/templates.go b/internal/templates.go
@@ -4,9 +4,11 @@ import (
 	"io/ioutil"
 	"net"
 	"path"
+	"strings"
 
 	"github.com/airfocusio/hcloud-talos/internal/clients"
 	"github.com/airfocusio/hcloud-talos/internal/cluster"
+	"github.com/airfocusio/hcloud-talos/internal/utils"
 	"github.com/hetznercloud/hcloud-go/hcloud"
 )
 
@@ -34,9 +36,9 @@ func nodeNetworkTemplate(cl *cluster.Cluster) hcloud.NetworkCreateOpts {
 	}
 }
 
-func nodePlacementGroupTemplate(cl *cluster.Cluster) hcloud.PlacementGroupCreateOpts {
+func controlplanePlacementGroupTemplate(cl *cluster.Cluster) hcloud.PlacementGroupCreateOpts {
 	return hcloud.PlacementGroupCreateOpts{
-		Name:   cl.Config.ClusterName + "-nodes",
+		Name:   cl.Config.ClusterName + "-controlplanes",
 		Type:   hcloud.PlacementGroupTypeSpread,
 		Labels: map[string]string{clusterLabel: cl.Config.ClusterName},
 	}
@@ -124,7 +126,7 @@ func nodeFirewallTemplate(cl *cluster.Cluster, network *hcloud.Network) hcloud.F
 }
 
 func nodeName(cl *cluster.Cluster, name string) string {
-	return cl.Config.ClusterName + "-" + name
+	return cl.Config.ClusterName + "-" + strings.Replace(name, "%id%", utils.RandString(6), 1)
 }
 
 func controlplaneNodeTemplate(cl *cluster.Cluster, serverType string, name string, talosVersion string) (clients.HcloudServerCreateFromImageOpts, error) {

diff --git a/internal/utils/rand.go b/internal/utils/rand.go
@@ -6,7 +6,7 @@ import (
 )
 
 var randStringInitialized = false
-var randStringRunes = []rune("0123456789abcdef")
+var randStringRunes = []rune("0123456789abcdefghijklmnopqrstuvwxyz")
 
 func RandString(n int) string {
 	if !randStringInitialized {