K8SPSMDB-1003: Kubernetes node zone/region tag (#1360)

* K8SPSMDB-1003 - kubernetes node tags zone/region Add kubernetes node tags zone/region to the monogo nodes. * Remove worning message if we do not have special permission. * fix test * fix cross-site test * fix image * update test * delete unsused * update cross-site test * fix PR comments * fix * fix * fix * fix --------- Co-authored-by: Viacheslav Sarzhan <[email protected]> Co-authored-by: Natalia Marukovich <[email protected]> Co-authored-by: Natalia Marukovich <[email protected]> Co-authored-by: Inel Pandzic <[email protected]>
percona · Apr 24, 2024 · 028e8c3 · 028e8c3
1 parent 282394a
commit 028e8c3
Show file tree

Hide file tree

Showing 6 changed files with 88 additions and 14 deletions.
diff --git a/deploy/cw-bundle.yaml b/deploy/cw-bundle.yaml
@@ -18336,6 +18336,14 @@ rules:
   - update
   - patch
   - delete
+- apiGroups:
+  - ""
+  resources:
+  - nodes
+  verbs:
+  - get
+  - list
+  - watch
 - apiGroups:
   - ""
   resources:

diff --git a/deploy/cw-rbac.yaml b/deploy/cw-rbac.yaml
@@ -35,6 +35,14 @@ rules:
   - update
   - patch
   - delete
+- apiGroups:
+  - ""
+  resources:
+  - nodes
+  verbs:
+  - get
+  - list
+  - watch
 - apiGroups:
   - ""
   resources:

diff --git a/e2e-tests/cross-site-sharded/run b/e2e-tests/cross-site-sharded/run
@@ -13,6 +13,26 @@ unset OPERATOR_NS
 main_cluster="cross-site-sharded-main"
 replica_cluster="cross-site-sharded-replica"
 
+wait_for_members() {
+  local endpoint="$1"
+  local rsName="$2"
+  local nodes_amount=0
+  until [[ ${nodes_amount} == 6 ]]; do
+    nodes_amount=$(run_mongos 'rs.conf().members.length' "clusterAdmin:clusterAdmin123456@$endpoint" "mongodb" ":27017" \
+     | egrep -v 'I NETWORK|W NETWORK|Error saving history file|Percona Server for MongoDB|connecting to:|Unable to reach primary for set|Implicit session:|versions do not match|Error saving history file:|bye' \
+     | $sed -re 's/ObjectId\("[0-9a-f]+"\)//; s/-[0-9]+.svc/-xxx.svc/')
+
+    echo "waiting for all members to be configured in ${rsName}"
+    let retry+=1
+    if [ $retry -ge 15 ]; then
+      echo "Max retry count $retry reached. something went wrong with mongo cluster. Config for endpoint $endpoint has $nodes_amount but expected 6."
+      exit 1
+    fi
+    echo -n .
+    sleep 10
+  done
+}
+
 desc "create main cluster"
 create_infra "$namespace"
 
@@ -118,7 +138,10 @@ sleep 30
 
 desc "create replica PSMDB cluster $cluster"
 apply_cluster "$test_dir/conf/${replica_cluster}.yml"
-sleep 300
+
+wait_for_running $replica_cluster-rs0 3 "false"
+wait_for_running $replica_cluster-rs1 3 "false"
+wait_for_running $replica_cluster-cfg 3 "false"
 
 replica_cfg_0_endpoint=$(get_service_ip cross-site-sharded-replica-cfg-0 'cfg')
 replica_cfg_1_endpoint=$(get_service_ip cross-site-sharded-replica-cfg-1 'cfg')
@@ -141,7 +164,10 @@ kubectl_bin patch psmdb ${main_cluster} --type=merge --patch '{
 			  }
 		}'
 
-sleep 60
+wait_for_members $replica_cfg_0_endpoint cfg
+wait_for_members $replica_rs0_0_endpoint rs0
+wait_for_members $replica_rs1_0_endpoint rs1
+
 kubectl_bin config set-context $(kubectl_bin config current-context) --namespace="$replica_namespace"
 
 desc 'check if all 3 Pods started'
@@ -165,8 +191,8 @@ compare_mongos_cmd "find" "myApp:myPass@$main_cluster-mongos.$namespace"
 
 desc 'test failover'
 kubectl_bin config set-context $(kubectl_bin config current-context) --namespace="$namespace"
+
 kubectl_bin delete psmdb $main_cluster
-sleep 60
 
 desc 'run disaster recovery script for replset: cfg'
 run_script_mongos "${test_dir}/disaster_recovery.js" "clusterAdmin:clusterAdmin123456@$replica_cfg_0_endpoint" "mongodb" ":27017"
@@ -180,7 +206,9 @@ run_script_mongos "${test_dir}/disaster_recovery.js" "clusterAdmin:clusterAdmin1
 desc 'make replica cluster managed'
 kubectl_bin config set-context $(kubectl_bin config current-context) --namespace="$replica_namespace"
 kubectl_bin patch psmdb ${replica_cluster} --type=merge --patch '{"spec":{"unmanaged": false}}'
-sleep 120
+
+wait_for_running $replica_cluster-rs0 3
+wait_for_running $replica_cluster-cfg 3
 
 desc "check failover status"
 compare_mongos_cmd "find" "myApp:myPass@$replica_cluster-mongos.$replica_namespace"

diff --git a/e2e-tests/serviceless-external-nodes/run b/e2e-tests/serviceless-external-nodes/run
@@ -46,6 +46,8 @@ cat $tmp_dir/psmdb.yaml \
 
 wait_cluster_consistency ${cluster}
 
+# waiting the config will be ready.
+sleep 30
 run_mongo 'rs.status().members.forEach(function(z){printjson(z.name);printjson(z.stateStr); })' "clusterAdmin:clusterAdmin123456@${cluster}-rs0-0.${cluster}-rs0.${namespace}" "mongodb" | egrep -v 'I NETWORK|W NETWORK|Error saving history file|Percona Server for MongoDB|connecting to:|Unable to reach primary for set|Implicit session:|versions do not match|Error saving history file:|bye' >"$tmp_dir/rs.txt"
 
 cat "${test_dir}/compare/rs.txt" \

diff --git a/pkg/controller/perconaservermongodb/mgo.go b/pkg/controller/perconaservermongodb/mgo.go
@@ -19,6 +19,7 @@ import (
 	api "github.com/percona/percona-server-mongodb-operator/pkg/apis/psmdb/v1"
 	"github.com/percona/percona-server-mongodb-operator/pkg/psmdb"
 	"github.com/percona/percona-server-mongodb-operator/pkg/psmdb/mongo"
+	"github.com/percona/percona-server-mongodb-operator/pkg/util"
 )
 
 var errReplsetLimit = fmt.Errorf("maximum replset member (%d) count reached", mongo.MaxMembers)
@@ -267,6 +268,20 @@ func (r *ReconcilePerconaServerMongoDB) updateConfigMembers(ctx context.Context,
 			return 0, fmt.Errorf("get host for pod %s: %v", pod.Name, err)
 		}
 
+		nodeLabels := mongo.ReplsetTags{
+			"nodeName":    pod.Spec.NodeName,
+			"podName":     pod.Name,
+			"serviceName": cr.Name,
+		}
+
+		labels, err := psmdb.GetNodeLabels(ctx, r.client, cr, pod)
+		if err == nil {
+			nodeLabels = util.MapMerge(nodeLabels, mongo.ReplsetTags{
+				"region": labels[corev1.LabelTopologyRegion],
+				"zone":   labels[corev1.LabelTopologyZone],
+			})
+		}
+
 		member := mongo.ConfigMember{
 			ID:           key,
 			Host:         host,
@@ -293,16 +308,11 @@ func (r *ReconcilePerconaServerMongoDB) updateConfigMembers(ctx context.Context,
 			member.ArbiterOnly = true
 			member.Priority = 0
 		case "mongod", "cfg":
-			member.Tags = mongo.ReplsetTags{
-				"podName":     pod.Name,
-				"serviceName": cr.Name,
-			}
+			member.Tags = nodeLabels
 		case "nonVoting":
-			member.Tags = mongo.ReplsetTags{
-				"podName":     pod.Name,
-				"serviceName": cr.Name,
-				"nonVoting":   "true",
-			}
+			member.Tags = util.MapMerge(mongo.ReplsetTags{
+				"nonVoting": "true",
+			}, nodeLabels)
 			member.Priority = 0
 			member.Votes = 0
 		}
@@ -597,7 +607,7 @@ func (r *ReconcilePerconaServerMongoDB) handleReplsetInit(ctx context.Context, c
 			"sh", "-c",
 			fmt.Sprintf(
 				`
-				cat <<-EOF | %s 
+				cat <<-EOF | %s
 				rs.initiate(
 					{
 						_id: '%s',

diff --git a/pkg/psmdb/getters.go b/pkg/psmdb/getters.go
@@ -3,6 +3,7 @@ package psmdb
 import (
 	"context"
 	"sort"
+	"time"
 
 	appsv1 "k8s.io/api/apps/v1"
 	corev1 "k8s.io/api/core/v1"
@@ -165,3 +166,20 @@ func GetExportedServices(ctx context.Context, cl client.Client, cr *api.PerconaS
 
 	return seList, nil
 }
+
+func GetNodeLabels(ctx context.Context, cl client.Client, cr *api.PerconaServerMongoDB, pod corev1.Pod) (map[string]string, error) {
+	// Set a timeout for the request, to avoid hanging forever
+	ctx, cancel := context.WithTimeout(ctx, 5*time.Second)
+	defer cancel()
+
+	node := &corev1.Node{}
+
+	err := cl.Get(ctx, client.ObjectKey{
+		Name: pod.Spec.NodeName,
+	}, node)
+	if err != nil {
+		return nil, errors.Wrapf(err, "failed to get node %s", pod.Spec.NodeName)
+	}
+
+	return node.Labels, nil
+}