diff --git a/src/meta/src/stream/stream_graph/schedule.rs b/src/meta/src/stream/stream_graph/schedule.rs
index 8361e6b8ef7c7..11272a6e56d7b 100644
--- a/src/meta/src/stream/stream_graph/schedule.rs
+++ b/src/meta/src/stream/stream_graph/schedule.rs
@@ -38,7 +38,7 @@ use crate::manager::{WorkerId, WorkerLocations};
 use crate::model::ActorId;
 use crate::stream::stream_graph::fragment::CompleteStreamFragmentGraph;
 use crate::stream::stream_graph::id::GlobalFragmentId as Id;
-use crate::MetaResult;
+use crate::{MetaError, MetaResult};
 
 type HashMappingId = usize;
 
@@ -210,6 +210,12 @@ impl Scheduler {
                 .push(p);
         }
 
+        if parallel_units_map.is_empty() {
+            return Err(MetaError::unavailable(
+                "No available parallel units to schedule".to_string(),
+            ));
+        }
+
         // Use all parallel units if no default parallelism is specified.
         let default_parallelism = default_parallelism.map_or_else(
             || parallel_units_map.values().map(|p| p.len()).sum::<usize>(),
@@ -236,10 +242,10 @@ impl Scheduler {
         round_robin.truncate(default_parallelism);
 
         if round_robin.len() < default_parallelism {
-            bail!(
+            return Err(MetaError::unavailable(format!(
                 "Not enough parallel units to schedule {} parallelism",
                 default_parallelism
-            );
+            )));
         }
 
         // Sort all parallel units by ID to achieve better vnode locality.