diff --git a/src/common/meta/src/key/tombstone.rs b/src/common/meta/src/key/tombstone.rs
index 470482c75b..22039142f5 100644
--- a/src/common/meta/src/key/tombstone.rs
+++ b/src/common/meta/src/key/tombstone.rs
@@ -59,6 +59,8 @@ pub struct TombstoneManager {
 }
 
 const TOMBSTONE_PREFIX: &str = "__tombstone/";
+const MOVE_VALUE_TXN_OPS_PER_KEY: usize = 4;
+const RESTORE_VALUE_TXN_OPS_PER_KEY: usize = 6;
 
 impl TombstoneManager {
     /// Returns [TombstoneManager].
@@ -292,7 +294,12 @@ impl TombstoneManager {
         if keys.is_empty() {
             return Ok(0);
         }
-        let chunk_size = self.max_txn_ops() / 2;
+        let txn_ops_per_key = if require_dest_not_exists {
+            RESTORE_VALUE_TXN_OPS_PER_KEY
+        } else {
+            MOVE_VALUE_TXN_OPS_PER_KEY
+        };
+        let chunk_size = (self.max_txn_ops() / txn_ops_per_key).max(1);
         if keys.len() > chunk_size {
             debug!(
                 "Moving values with multiple chunks, keys len: {}, chunk_size: {}",
@@ -374,14 +381,84 @@ impl TombstoneManager {
 #[cfg(test)]
 mod tests {
 
+    use std::any::Any;
     use std::collections::HashMap;
     use std::sync::Arc;
 
-    use crate::error::Error;
+    use crate::error::{Error, Result};
     use crate::key::tombstone::TombstoneManager;
-    use crate::kv_backend::KvBackend;
     use crate::kv_backend::memory::MemoryKvBackend;
-    use crate::rpc::store::PutRequest;
+    use crate::kv_backend::txn::{Txn, TxnRequest, TxnResponse};
+    use crate::kv_backend::{KvBackend, TxnService};
+    use crate::rpc::store::{
+        BatchDeleteRequest, BatchDeleteResponse, BatchGetRequest, BatchGetResponse,
+        BatchPutRequest, BatchPutResponse, DeleteRangeRequest, DeleteRangeResponse, PutRequest,
+        PutResponse, RangeRequest, RangeResponse,
+    };
+
+    struct TxnOpLimitKvBackend {
+        inner: Arc<MemoryKvBackend<Error>>,
+        max_txn_ops: usize,
+    }
+
+    #[async_trait::async_trait]
+    impl TxnService for TxnOpLimitKvBackend {
+        type Error = Error;
+
+        async fn txn(&self, txn: Txn) -> Result<TxnResponse> {
+            let TxnRequest {
+                compare,
+                success,
+                failure,
+            } = txn.req();
+            let txn_ops = compare.len() + success.len() + failure.len();
+            assert!(
+                txn_ops <= self.max_txn_ops,
+                "txn ops {txn_ops} exceeds limit {}",
+                self.max_txn_ops
+            );
+            self.inner.txn(txn).await
+        }
+
+        fn max_txn_ops(&self) -> usize {
+            self.max_txn_ops
+        }
+    }
+
+    #[async_trait::async_trait]
+    impl KvBackend for TxnOpLimitKvBackend {
+        fn name(&self) -> &str {
+            "txn_op_limit"
+        }
+
+        fn as_any(&self) -> &dyn Any {
+            self
+        }
+
+        async fn range(&self, req: RangeRequest) -> Result<RangeResponse> {
+            self.inner.range(req).await
+        }
+
+        async fn put(&self, req: PutRequest) -> Result<PutResponse> {
+            self.inner.put(req).await
+        }
+
+        async fn batch_put(&self, req: BatchPutRequest) -> Result<BatchPutResponse> {
+            self.inner.batch_put(req).await
+        }
+
+        async fn batch_get(&self, req: BatchGetRequest) -> Result<BatchGetResponse> {
+            self.inner.batch_get(req).await
+        }
+
+        async fn delete_range(&self, req: DeleteRangeRequest) -> Result<DeleteRangeResponse> {
+            self.inner.delete_range(req).await
+        }
+
+        async fn batch_delete(&self, req: BatchDeleteRequest) -> Result<BatchDeleteResponse> {
+            self.inner.batch_delete(req).await
+        }
+    }
 
     #[derive(Debug, Clone)]
     struct MoveValue {
@@ -654,6 +731,70 @@ mod tests {
         check_moved_values(kv_backend.clone(), &move_values).await;
     }
 
+    #[tokio::test]
+    async fn test_restore_chunks_by_total_txn_ops_limit() {
+        let inner = Arc::new(MemoryKvBackend::default());
+        let kv_backend = Arc::new(TxnOpLimitKvBackend {
+            inner: inner.clone(),
+            max_txn_ops: 6,
+        });
+        let tombstone_manager = TombstoneManager::new(kv_backend);
+        let kvs = HashMap::from([
+            (b"bar".to_vec(), b"baz".to_vec()),
+            (b"foo".to_vec(), b"hi".to_vec()),
+            (b"baz".to_vec(), b"hello".to_vec()),
+        ]);
+        for (key, value) in &kvs {
+            inner
+                .put(
+                    PutRequest::new()
+                        .with_key(tombstone_manager.to_tombstone(key))
+                        .with_value(value.clone()),
+                )
+                .await
+                .unwrap();
+        }
+
+        let restored = tombstone_manager
+            .restore(kvs.keys().cloned().collect())
+            .await
+            .unwrap();
+
+        assert_eq!(kvs.len(), restored);
+    }
+
+    #[tokio::test]
+    async fn test_create_chunks_by_total_txn_ops_limit() {
+        let inner = Arc::new(MemoryKvBackend::default());
+        let kv_backend = Arc::new(TxnOpLimitKvBackend {
+            inner: inner.clone(),
+            max_txn_ops: 4,
+        });
+        let tombstone_manager = TombstoneManager::new(kv_backend);
+        let kvs = HashMap::from([
+            (b"bar".to_vec(), b"baz".to_vec()),
+            (b"foo".to_vec(), b"hi".to_vec()),
+            (b"baz".to_vec(), b"hello".to_vec()),
+        ]);
+        for (key, value) in &kvs {
+            inner
+                .put(
+                    PutRequest::new()
+                        .with_key(key.clone())
+                        .with_value(value.clone()),
+                )
+                .await
+                .unwrap();
+        }
+
+        let moved = tombstone_manager
+            .create(kvs.keys().cloned().collect())
+            .await
+            .unwrap();
+
+        assert_eq!(kvs.len(), moved);
+    }
+
     #[tokio::test]
     async fn test_move_values_with_non_exists_values() {
         let kv_backend = Arc::new(MemoryKvBackend::default());