diff --git a/CHANGELOG.md b/CHANGELOG.md
index c46a66af9..960a19c17 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -8,6 +8,7 @@ adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 ### Added
  * `MysqlDatabase` `MysqlUser` Add delete policy
  * Add `PtHeartbeatResources` in `.Spec.PodSpec` to allow the user specifying resources for pt-heartbeat.
+ * Add `CandidateNode` field in `MysqlBackup.Spec` to allow the user specifying candidate node for backup.
  * Set `MysqlCluter.Spec.BackupSchedule` to empty string to disable recurrent backups
 ### Changed
  * Set default MySQL server version to `5.7.35`
diff --git a/config/crd/bases/mysql.presslabs.org_mysqlbackups.yaml b/config/crd/bases/mysql.presslabs.org_mysqlbackups.yaml
index 20a793935..0a7e86069 100644
--- a/config/crd/bases/mysql.presslabs.org_mysqlbackups.yaml
+++ b/config/crd/bases/mysql.presslabs.org_mysqlbackups.yaml
@@ -36,6 +36,9 @@ spec:
                 backupURL:
                   description: BackupURL represents the URL to the backup location, this can be partially specifyied. Default is used the one specified in the cluster.
                   type: string
+                candidateNode:
+                  description: CandidateNode is the node host that will be used to take the backup. If not set or set to wrong node, the operator will calculate candidate node by itself.
+                  type: string
                 clusterName:
                   description: ClustterName represents the cluster for which to take backup
                   type: string
diff --git a/deploy/charts/mysql-operator/crds/mysql.presslabs.org_mysqlbackups.yaml b/deploy/charts/mysql-operator/crds/mysql.presslabs.org_mysqlbackups.yaml
index 0b5cc05f6..5e3a08209 100644
--- a/deploy/charts/mysql-operator/crds/mysql.presslabs.org_mysqlbackups.yaml
+++ b/deploy/charts/mysql-operator/crds/mysql.presslabs.org_mysqlbackups.yaml
@@ -37,6 +37,9 @@ spec:
                 backupURL:
                   description: BackupURL represents the URL to the backup location, this can be partially specifyied. Default is used the one specified in the cluster.
                   type: string
+                candidateNode:
+                  description: CandidateNode is the node host that will be used to take the backup. If not set or set to wrong node, the operator will calculate candidate node by itself.
+                  type: string
                 clusterName:
                   description: ClustterName represents the cluster for which to take backup
                   type: string
diff --git a/pkg/apis/mysql/v1alpha1/mysqlbackup_types.go b/pkg/apis/mysql/v1alpha1/mysqlbackup_types.go
index 330eed777..61bffe799 100644
--- a/pkg/apis/mysql/v1alpha1/mysqlbackup_types.go
+++ b/pkg/apis/mysql/v1alpha1/mysqlbackup_types.go
@@ -44,6 +44,11 @@ type MysqlBackupSpec struct {
 	// default it's used softDelete.
 	// +optional
 	RemoteDeletePolicy DeletePolicy `json:"remoteDeletePolicy,omitempty"`
+
+	// CandidateNode is the node host that will be used to take the backup.
+	// If not set or set to wrong node, the operator will calculate candidate node by itself.
+	// +optional
+	CandidateNode string `json:"candidateNode,omitempty"`
 }
 
 // BackupCondition defines condition struct for backup resource
@@ -94,7 +99,6 @@ type MysqlBackupStatus struct {
 
 // MysqlBackup is the Schema for the mysqlbackups API
 // +kubebuilder:object:root=true
-//
 type MysqlBackup struct {
 	metav1.TypeMeta   `json:",inline"`
 	metav1.ObjectMeta `json:"metadata,omitempty"`
@@ -105,7 +109,6 @@ type MysqlBackup struct {
 
 // MysqlBackupList contains a list of MysqlBackup
 // +kubebuilder:object:root=true
-//
 type MysqlBackupList struct {
 	metav1.TypeMeta `json:",inline"`
 	metav1.ListMeta `json:"metadata,omitempty"`
diff --git a/pkg/controller/mysqlbackup/internal/syncer/job.go b/pkg/controller/mysqlbackup/internal/syncer/job.go
index af13de3dc..33ae9a15f 100644
--- a/pkg/controller/mysqlbackup/internal/syncer/job.go
+++ b/pkg/controller/mysqlbackup/internal/syncer/job.go
@@ -88,9 +88,17 @@ func (s *jobSyncer) SyncFn() error {
 	return nil
 }
 
-// getBackupCandidate returns the hostname of the first not-lagged and
+// getBackupCandidate returns candidate node in mysqlbackup spec at first,
+// if not, it will return the hostname of the first not-lagged and
 // replicating slave node, else returns the master node.
 func (s *jobSyncer) getBackupCandidate() string {
+	if s.backup.Spec.CandidateNode != "" {
+		if err := s.backup.Validate(s.cluster); err == nil {
+			return s.backup.Spec.CandidateNode
+		}
+		log.Info("backup's candidate node is not valid, will try to calculate candidate node")
+	}
+
 	for _, node := range s.cluster.Status.Nodes {
 		master := s.cluster.GetNodeCondition(node.Name, api.NodeConditionMaster)
 		replicating := s.cluster.GetNodeCondition(node.Name, api.NodeConditionReplicating)
diff --git a/pkg/controller/mysqlbackup/mysqlbackup_controller_test.go b/pkg/controller/mysqlbackup/mysqlbackup_controller_test.go
index 8744cd8b6..6e1cbaa31 100644
--- a/pkg/controller/mysqlbackup/mysqlbackup_controller_test.go
+++ b/pkg/controller/mysqlbackup/mysqlbackup_controller_test.go
@@ -119,16 +119,7 @@ var _ = Describe("MysqlBackup controller", func() {
 		BeforeEach(func() {
 			// create a cluster with 2 nodes
 			Expect(c.Create(context.TODO(), cluster.Unwrap())).To(Succeed())
-			cluster.Status.Nodes = []api.NodeStatus{
-				{
-					Name:       cluster.GetPodHostname(0),
-					Conditions: testutil.NodeConditions(true, false, false, false),
-				},
-				{
-					Name:       cluster.GetPodHostname(1),
-					Conditions: testutil.NodeConditions(false, true, false, true),
-				},
-			}
+			cluster.Status.Nodes = getHealthyNodeStatus(cluster, 2)
 			Expect(c.Status().Update(context.TODO(), cluster.Unwrap())).To(Succeed())
 			// create the backup
 			Expect(c.Create(context.TODO(), backup.Unwrap())).To(Succeed())
@@ -316,8 +307,74 @@ var _ = Describe("MysqlBackup controller", func() {
 			Expect(c.Delete(context.TODO(), cluster.Unwrap())).To(Succeed())
 		})
 	})
+
+	When("candidate node is setted to wrong node", func() {
+		BeforeEach(func() {
+			backup.Spec.CandidateNode = cluster.GetPodHostname(3)
+			Expect(c.Create(context.TODO(), backup.Unwrap())).To(Succeed())
+
+			Expect(c.Create(context.TODO(), cluster.Unwrap())).To(Succeed())
+			cluster.Status.Nodes = getHealthyNodeStatus(cluster, 2)
+			Expect(c.Status().Update(context.TODO(), cluster.Unwrap())).To(Succeed())
+
+			Eventually(requests, timeout).Should(Receive(Equal(expectedRequest)))
+			Eventually(requests, timeout).Should(Receive(Equal(expectedRequest)))
+			testutil.DrainChan(requests)
+		})
+		AfterEach(func() {
+			Expect(c.Delete(context.TODO(), backup.Unwrap())).To(Succeed())
+			Expect(c.Delete(context.TODO(), cluster.Unwrap())).To(Succeed())
+		})
+
+		It("should take backup from replica 1", func() {
+			job := &batch.Job{}
+			Expect(c.Get(context.TODO(), jobKey, job)).To(Succeed())
+			Expect(job.Spec.Template.Spec.Containers[0].Args).To(ContainElement(Equal(cluster.GetPodHostname(1))))
+		})
+	})
+
+	When("candidate node is setted to master", func() {
+		BeforeEach(func() {
+			backup.Spec.CandidateNode = cluster.GetPodHostname(0)
+			Expect(c.Create(context.TODO(), backup.Unwrap())).To(Succeed())
+
+			Expect(c.Create(context.TODO(), cluster.Unwrap())).To(Succeed())
+			cluster.Status.Nodes = getHealthyNodeStatus(cluster, 2)
+			Expect(c.Status().Update(context.TODO(), cluster.Unwrap())).To(Succeed())
+
+			Eventually(requests, timeout).Should(Receive(Equal(expectedRequest)))
+			Eventually(requests, timeout).Should(Receive(Equal(expectedRequest)))
+			testutil.DrainChan(requests)
+		})
+		AfterEach(func() {
+			Expect(c.Delete(context.TODO(), backup.Unwrap())).To(Succeed())
+			Expect(c.Delete(context.TODO(), cluster.Unwrap())).To(Succeed())
+		})
+
+		It("should take backup from master", func() {
+			job := &batch.Job{}
+			Expect(c.Get(context.TODO(), jobKey, job)).To(Succeed())
+			Expect(job.Spec.Template.Spec.Containers[0].Args).To(ContainElement(Equal(cluster.GetPodHostname(0))))
+		})
+	})
 })
 
+func getHealthyNodeStatus(cluster *mysqlcluster.MysqlCluster, count int) []api.NodeStatus {
+	status := []api.NodeStatus{
+		{
+			Name:       cluster.GetPodHostname(0),
+			Conditions: testutil.NodeConditions(true, false, false, false),
+		},
+	}
+	for i := 1; i < count; i++ {
+		status = append(status, api.NodeStatus{
+			Name:       cluster.GetPodHostname(i),
+			Conditions: testutil.NodeConditions(false, true, false, false),
+		})
+	}
+	return status
+}
+
 func refreshFn(c client.Client, backupKey types.NamespacedName) func() *api.MysqlBackup {
 	return func() *api.MysqlBackup {
 		backup := &api.MysqlBackup{}
diff --git a/pkg/internal/mysqlbackup/validation.go b/pkg/internal/mysqlbackup/validation.go
new file mode 100644
index 000000000..a51a28fbf
--- /dev/null
+++ b/pkg/internal/mysqlbackup/validation.go
@@ -0,0 +1,42 @@
+/*
+Copyright 2018 Pressinfra SRL
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package mysqlbackup
+
+import (
+	"fmt"
+
+	"github.com/bitpoke/mysql-operator/pkg/internal/mysqlcluster"
+)
+
+// Validate checks if the backup spec is validated
+func (c *MysqlBackup) Validate(cluster *mysqlcluster.MysqlCluster) error {
+	// TODO: this validation should be done in an admission web-hook
+
+	if c.Spec.CandidateNode != "" {
+		validate := false
+		for i := 0; i < int(*cluster.Spec.Replicas); i++ {
+			if c.Spec.CandidateNode == cluster.GetPodHostname(i) {
+				validate = true
+				break
+			}
+		}
+		if !validate {
+			return fmt.Errorf("spec.candidateNode is not a valid node")
+		}
+	}
+	return nil
+}