Merge pull request #8272 from johngmyers/need-update-annotation

Support the kops.k8s.io/needs-update annotation on nodes

cmd/kops/rollingupdatecluster.go

@@ -47,7 +47,8 @@ var (
 	This command updates a kubernetes cluster to match the cloud and kops specifications.
 
 	To perform a rolling update, you need to update the cloud resources first with the command
-	` + pretty.Bash("kops update cluster") + `.
+	` + pretty.Bash("kops update cluster") + `. Nodes may be additionally marked for update by placing a
+	` + pretty.Bash("kops.k8s.io/needs-update") + ` annotation on them.
 
 	If rolling-update does not report that the cluster needs to be rolled, you can force the cluster to be
 	rolled with the force flag.  Rolling update drains and validates the cluster by default.  A cluster is
@@ -318,6 +319,30 @@ func RunRollingUpdateCluster(f *util.Factory, out io.Writer, options *RollingUpd
 		return err
 	}
 
+	d := &instancegroups.RollingUpdateCluster{
+		MasterInterval:    options.MasterInterval,
+		NodeInterval:      options.NodeInterval,
+		BastionInterval:   options.BastionInterval,
+		Interactive:       options.Interactive,
+		Force:             options.Force,
+		Cloud:             cloud,
+		K8sClient:         k8sClient,
+		FailOnDrainError:  options.FailOnDrainError,
+		FailOnValidate:    options.FailOnValidate,
+		CloudOnly:         options.CloudOnly,
+		ClusterName:       options.ClusterName,
+		PostDrainDelay:    options.PostDrainDelay,
+		ValidationTimeout: options.ValidationTimeout,
+		// TODO should we expose this to the UI?
+		ValidateTickDuration:    30 * time.Second,
+		ValidateSuccessDuration: 10 * time.Second,
+	}
+
+	err = d.AdjustNeedUpdate(groups, cluster, list)
+	if err != nil {
+		return err
+	}
+
 	{
 		t := &tables.Table{}
 		t.AddColumn("NAME", func(r *cloudinstances.CloudInstanceGroup) string {
@@ -391,24 +416,7 @@ func RunRollingUpdateCluster(f *util.Factory, out io.Writer, options *RollingUpd
 			return fmt.Errorf("cannot create cluster validator: %v", err)
 		}
 	}
-	d := &instancegroups.RollingUpdateCluster{
-		MasterInterval:    options.MasterInterval,
-		NodeInterval:      options.NodeInterval,
-		BastionInterval:   options.BastionInterval,
-		Interactive:       options.Interactive,
-		Force:             options.Force,
-		Cloud:             cloud,
-		K8sClient:         k8sClient,
-		ClusterValidator:  clusterValidator,
-		FailOnDrainError:  options.FailOnDrainError,
-		FailOnValidate:    options.FailOnValidate,
-		CloudOnly:         options.CloudOnly,
-		ClusterName:       options.ClusterName,
-		PostDrainDelay:    options.PostDrainDelay,
-		ValidationTimeout: options.ValidationTimeout,
-		// TODO should we expose this to the UI?
-		ValidateTickDuration:    30 * time.Second,
-		ValidateSuccessDuration: 10 * time.Second,
-	}
+	d.ClusterValidator = clusterValidator
+
 	return d.RollingUpdate(groups, cluster, list)
 }

docs/cli/kops_rolling-update.md

@@ -10,7 +10,8 @@ Rolling update a cluster.
 This command updates a kubernetes cluster to match the cloud and kops specifications.
 
 To perform a rolling update, you need to update the cloud resources first with the command
-`kops update cluster`.
+`kops update cluster`. Nodes may be additionally marked for update by placing a
+`kops.k8s.io/needs-update` annotation on them.
 
 If rolling-update does not report that the cluster needs to be rolled, you can force the cluster to be
 rolled with the force flag.  Rolling update drains and validates the cluster by default.  A cluster is

docs/cli/kops_rolling-update_cluster.md

@@ -10,7 +10,8 @@ Rolling update a cluster.
 This command updates a kubernetes cluster to match the cloud and kops specifications.
 
 To perform a rolling update, you need to update the cloud resources first with the command
-`kops update cluster`.
+`kops update cluster`. Nodes may be additionally marked for update by placing a
+`kops.k8s.io/needs-update` annotation on them.
 
 If rolling-update does not report that the cluster needs to be rolled, you can force the cluster to be
 rolled with the force flag.  Rolling update drains and validates the cluster by default.  A cluster is

docs/operations/rolling-update.md

@@ -14,6 +14,7 @@ Cloud instances are chosen to be updated (replaced) if at least one of the follo
 * The instance was created with a specification that is older than that generated by the last
 `kops update cluster`.
 * The instance was detached for surging by a previous (failed or interrupted) rolling update.
+* The node has a `kops.k8s.io/needs-update` annotation.
 * The `--force` flag was given to the `kops rolling-update cluster` command.
 
 ## Order of instance groups

pkg/instancegroups/rollingupdate.go

@@ -69,6 +69,31 @@ type RollingUpdateCluster struct {
 	ValidateSuccessDuration time.Duration
 }
 
+// AdjustNeedUpdate adjusts the set of instances that need updating, using factors outside those known by the cloud implementation
+func (c *RollingUpdateCluster) AdjustNeedUpdate(groups map[string]*cloudinstances.CloudInstanceGroup, cluster *api.Cluster, instanceGroups *api.InstanceGroupList) error {
+	for _, group := range groups {
+		if group.Ready != nil {
+			var newReady []*cloudinstances.CloudInstanceGroupMember
+			for _, member := range group.Ready {
+				makeNotReady := false
+				if member.Node != nil && member.Node.Annotations != nil {
+					if _, ok := member.Node.Annotations["kops.k8s.io/needs-update"]; ok {
+						makeNotReady = true
+					}
+				}
+
+				if makeNotReady {
+					group.NeedUpdate = append(group.NeedUpdate, member)
+				} else {
+					newReady = append(newReady, member)
+				}
+			}
+			group.Ready = newReady
+		}
+	}
+	return nil
+}
+
 // RollingUpdate performs a rolling update on a K8s Cluster.
 func (c *RollingUpdateCluster) RollingUpdate(groups map[string]*cloudinstances.CloudInstanceGroup, cluster *api.Cluster, instanceGroups *api.InstanceGroupList) error {
 	if len(groups) == 0 {

pkg/instancegroups/rollingupdate_test.go

@@ -572,6 +572,103 @@ func TestRollingUpdateValidatesAfterBastion(t *testing.T) {
 	assertGroupInstanceCount(t, cloud, "bastion-1", 0)
 }
 
+func addNeedsUpdateAnnotation(group *cloudinstances.CloudInstanceGroup, node string) {
+	for _, igm := range group.Ready {
+		if igm.ID == node {
+			if igm.Node.Annotations == nil {
+				igm.Node.Annotations = map[string]string{}
+			}
+			igm.Node.Annotations["kops.k8s.io/needs-update"] = "somevalue"
+			return
+		}
+	}
+	for _, igm := range group.NeedUpdate {
+		if igm.ID == node {
+			if igm.Node.Annotations == nil {
+				igm.Node.Annotations = map[string]string{}
+			}
+			igm.Node.Annotations["kops.k8s.io/needs-update"] = "somevalue"
+			return
+		}
+	}
+	panic("did not find node " + node)
+}
+
+func TestAddAnnotatedNodesToNeedsUpdate(t *testing.T) {
+	c, cloud, cluster := getTestSetup()
+
+	groups := make(map[string]*cloudinstances.CloudInstanceGroup)
+	makeGroup(groups, c.K8sClient, cloud, "master-1", kopsapi.InstanceGroupRoleMaster, 2, 1)
+	makeGroup(groups, c.K8sClient, cloud, "node-1", kopsapi.InstanceGroupRoleNode, 2, 1)
+	makeGroup(groups, c.K8sClient, cloud, "node-2", kopsapi.InstanceGroupRoleNode, 2, 1)
+
+	addNeedsUpdateAnnotation(groups["node-1"], "node-1b")
+	addNeedsUpdateAnnotation(groups["node-2"], "node-2a")
+	addNeedsUpdateAnnotation(groups["master-1"], "master-1b")
+
+	err := c.AdjustNeedUpdate(groups, cluster, &kopsapi.InstanceGroupList{})
+	assert.NoError(t, err, "AddAnnotatedNodesToGroups")
+
+	assertGroupNeedUpdate(t, groups, "node-1", "node-1a", "node-1b")
+	assertGroupNeedUpdate(t, groups, "node-2", "node-2a")
+	assertGroupNeedUpdate(t, groups, "master-1", "master-1a", "master-1b")
+}
+
+func TestAddAnnotatedNodesToNeedsUpdateCloudonly(t *testing.T) {
+	c, cloud, cluster := getTestSetup()
+
+	groups := make(map[string]*cloudinstances.CloudInstanceGroup)
+	makeGroup(groups, c.K8sClient, cloud, "master-1", kopsapi.InstanceGroupRoleMaster, 2, 1)
+	makeGroup(groups, c.K8sClient, cloud, "node-1", kopsapi.InstanceGroupRoleNode, 2, 1)
+	makeGroup(groups, c.K8sClient, cloud, "node-2", kopsapi.InstanceGroupRoleNode, 2, 1)
+
+	addNeedsUpdateAnnotation(groups["node-1"], "node-1b")
+	addNeedsUpdateAnnotation(groups["node-2"], "node-2a")
+	addNeedsUpdateAnnotation(groups["master-1"], "master-1b")
+
+	c.CloudOnly = true
+	c.ClusterValidator = &assertNotCalledClusterValidator{T: t}
+
+	err := c.AdjustNeedUpdate(groups, cluster, &kopsapi.InstanceGroupList{})
+	assert.NoError(t, err, "AddAnnotatedNodesToGroups")
+
+	assertGroupNeedUpdate(t, groups, "node-1", "node-1a", "node-1b")
+	assertGroupNeedUpdate(t, groups, "node-2", "node-2a")
+	assertGroupNeedUpdate(t, groups, "master-1", "master-1a", "master-1b")
+}
+
+func TestAddAnnotatedNodesToNeedsUpdateNodesMissing(t *testing.T) {
+	c, cloud, cluster := getTestSetup()
+
+	groups := make(map[string]*cloudinstances.CloudInstanceGroup)
+	makeGroup(groups, c.K8sClient, cloud, "node-1", kopsapi.InstanceGroupRoleNode, 2, 1)
+
+	groups["node-1"].Ready[0].Node = nil
+	groups["node-1"].NeedUpdate[0].Node = nil
+
+	err := c.AdjustNeedUpdate(groups, cluster, &kopsapi.InstanceGroupList{})
+	assert.NoError(t, err, "AddAnnotatedNodesToGroups")
+}
+
+func assertGroupNeedUpdate(t *testing.T, groups map[string]*cloudinstances.CloudInstanceGroup, groupName string, nodes ...string) {
+	notFound := map[string]bool{}
+	for _, node := range nodes {
+		notFound[node] = true
+	}
+	for _, node := range groups[groupName].NeedUpdate {
+		if notFound[node.ID] {
+			notFound[node.ID] = false
+		} else {
+			t.Errorf("node %s of group %s is unexpectedly in NeedUpdate", node.ID, groupName)
+		}
+	}
+	for nodeID, v := range notFound {
+		if v {
+			t.Errorf("node %s of group %s is missing from NeedUpdate", nodeID, groupName)
+		}
+	}
+}
+
 func TestRollingUpdateTaintAllButOneNeedUpdate(t *testing.T) {
 	c, cloud, cluster := getTestSetup()