diff --git a/solr/CHANGES.txt b/solr/CHANGES.txt
index b006da2b850..96ce5d12ea7 100644
--- a/solr/CHANGES.txt
+++ b/solr/CHANGES.txt
@@ -232,6 +232,9 @@ Other Changes
 
 * SOLR-9331: Remove ReRankQuery's length constructor argument and member. (Christine Poerschke)
 
+* SOLR-9092: For the delete replica command we attempt to send the core admin delete request only
+  if that node is actually up. (Jessica Cheng Mallet, Varun Thacker)
+
 ==================  6.1.0 ==================
 
 Consult the LUCENE_CHANGES.txt file for additional, low level, changes in this release.
diff --git a/solr/core/src/java/org/apache/solr/cloud/OverseerCollectionMessageHandler.java b/solr/core/src/java/org/apache/solr/cloud/OverseerCollectionMessageHandler.java
index 0588446c653..49e094284b3 100644
--- a/solr/core/src/java/org/apache/solr/cloud/OverseerCollectionMessageHandler.java
+++ b/solr/core/src/java/org/apache/solr/cloud/OverseerCollectionMessageHandler.java
@@ -662,14 +662,19 @@ public class OverseerCollectionMessageHandler implements OverseerMessageHandler
     params.set(CoreAdminParams.DELETE_INSTANCE_DIR, message.getBool(CoreAdminParams.DELETE_INSTANCE_DIR, true));
     params.set(CoreAdminParams.DELETE_DATA_DIR, message.getBool(CoreAdminParams.DELETE_DATA_DIR, true));
 
-    sendShardRequest(replica.getNodeName(), params, shardHandler, asyncId, requestMap.get());
+    boolean isLive = zkStateReader.getClusterState().getLiveNodes().contains(replica.getNodeName());
+    if (isLive) {
+      sendShardRequest(replica.getNodeName(), params, shardHandler, asyncId, requestMap.get());
+    }
 
     Callable<Boolean> callable = () -> {
       try {
-        processResponses(results, shardHandler, false, null, asyncId, requestMap.get());
+        if (isLive) {
+          processResponses(results, shardHandler, false, null, asyncId, requestMap.get());
 
-        //check if the core unload removed the corenode zk entry
-        if (waitForCoreNodeGone(collectionName, shard, replicaName, 5000)) return Boolean.TRUE;
+          //check if the core unload removed the corenode zk entry
+          if (waitForCoreNodeGone(collectionName, shard, replicaName, 5000)) return Boolean.TRUE;
+        }
 
         // try and ensure core info is removed from cluster state
         deleteCoreNode(collectionName, replicaName, replica, core);