diff --git a/deployments/gcp-uscentral1b/config/prod.yaml b/deployments/gcp-uscentral1b/config/prod.yaml
index de665a06..c0e49e3d 100644
--- a/deployments/gcp-uscentral1b/config/prod.yaml
+++ b/deployments/gcp-uscentral1b/config/prod.yaml
@@ -11,3 +11,15 @@ daskhub:
     hub:
       extraEnv:
         OAUTH_CALLBACK_URL: "https://us-central1-b.gcp.pangeo.io/hub/oauth_callback"
+
+  dask-gateway:
+    backend:
+      # Ensure the scheduler and worker pods end up on 
+      # nodes that are already being used, to avoid scheduling on nodes
+      # that would otherwise be reclaimed by the cluster autoscaler.
+      scheduler:
+        extraPodConfig:
+          schedulerName: gcp-uscentral1b-prod-user-scheduler
+      worker:
+        extraPodConfig:
+          schedulerName: gcp-uscentral1b-prod-user-scheduler
diff --git a/deployments/gcp-uscentral1b/config/staging.yaml b/deployments/gcp-uscentral1b/config/staging.yaml
index 93aabc64..82fecb73 100644
--- a/deployments/gcp-uscentral1b/config/staging.yaml
+++ b/deployments/gcp-uscentral1b/config/staging.yaml
@@ -19,6 +19,17 @@ daskhub:
     scheduling:
       userScheduler:
         enabled: false
+  dask-gateway:
+    backend:
+      # Ensure the scheduler and worker pods end up on
+      # nodes that are already being used, to avoid scheduling on nodes
+      # that would otherwise be reclaimed by the cluster autoscaler.
+      scheduler:
+        extraPodConfig:
+          schedulerName: gcp-uscentral1b-staging-user-scheduler
+      worker:
+        extraPodConfig:
+          schedulerName: gcp-uscentral1b-staging-user-scheduler
 
 metrics:
   enabled: true
diff --git a/deployments/icesat2/config/prod.yaml b/deployments/icesat2/config/prod.yaml
index e3357d43..9fa2e112 100644
--- a/deployments/icesat2/config/prod.yaml
+++ b/deployments/icesat2/config/prod.yaml
@@ -33,3 +33,15 @@ daskhub:
           - aws-uswest2.pangeo.io
         letsencrypt:
           contactEmail: scottyh@uw.edu
+
+  dask-gateway:
+    backend:
+      # Ensure the scheduler and worker pods end up on
+      # nodes that are already being used, to avoid scheduling on nodes
+      # that would otherwise be reclaimed by the cluster autoscaler.
+      scheduler:
+        extraPodConfig:
+          schedulerName: icesat2-prod-user-scheduler
+      worker:
+        extraPodConfig:
+          schedulerName: icesat2-prod-staging-user-scheduler
diff --git a/deployments/icesat2/config/staging.yaml b/deployments/icesat2/config/staging.yaml
index 34679487..fdfad7c2 100644
--- a/deployments/icesat2/config/staging.yaml
+++ b/deployments/icesat2/config/staging.yaml
@@ -45,3 +45,15 @@ daskhub:
           - staging.aws-uswest2.pangeo.io
         letsencrypt:
           contactEmail: scottyh@uw.edu
+
+  dask-gateway:
+    backend:
+      # Ensure the scheduler and worker pods end up on
+      # nodes that are already being used, to avoid scheduling on nodes
+      # that would otherwise be reclaimed by the cluster autoscaler.
+      scheduler:
+        extraPodConfig:
+          schedulerName: icesat2-staging-user-scheduler
+      worker:
+        extraPodConfig:
+          schedulerName: icesat2-staging-user-scheduler