add a flag to distinguish duplicate remote model auto deploy and tran… (opensearch-project#2410) (opensearch-project#2412)

* add a flag to distinguish duplicate remote model auto deploy and transport deploy

Signed-off-by: Xun Zhang <xunzh@amazon.com>

* check for NPE for getIsAutoDeploying

Signed-off-by: Xun Zhang <xunzh@amazon.com>

---------

Signed-off-by: Xun Zhang <xunzh@amazon.com>
(cherry picked from commit f82e148)

Co-authored-by: Xun Zhang <xunzh@amazon.com>

Author: opensearch-trigger-bot[bot]

plugin/src/main/java/org/opensearch/ml/model/MLModelCache.java

@@ -54,6 +54,7 @@ public class MLModelCache {
     @Setter
     private Boolean deployToAllNodes;
     private @Setter(AccessLevel.PROTECTED) @Getter(AccessLevel.PROTECTED) Instant lastAccessTime;
+    private @Setter(AccessLevel.PROTECTED) @Getter(AccessLevel.PROTECTED) Boolean isAutoDeploying;
 
     public MLModelCache() {
         targetWorkerNodes = ConcurrentHashMap.newKeySet();

plugin/src/main/java/org/opensearch/ml/model/MLModelCacheHelper.java

@@ -17,6 +17,7 @@
 import java.util.concurrent.ConcurrentHashMap;
 import java.util.stream.Collectors;
 
+import org.apache.commons.lang3.BooleanUtils;
 import org.opensearch.cluster.service.ClusterService;
 import org.opensearch.common.settings.Settings;
 import org.opensearch.common.util.TokenBucket;
@@ -61,7 +62,7 @@ public synchronized void initModelState(
         List<String> targetWorkerNodes,
         boolean deployToAllNodes
     ) {
-        if (isModelRunningOnNode(modelId)) {
+        if (isModelRunningOnNode(modelId) && !isAutoDeploying(modelId)) {
             throw new MLLimitExceededException("Duplicate deploy model task");
         }
         log.debug("init model state for model {}, state: {}", modelId, state);
@@ -74,7 +75,7 @@ public synchronized void initModelState(
         modelCaches.put(modelId, modelCache);
     }
 
-    public synchronized void initModelStateLocal(
+    public synchronized void initModelStateAutoDeploy(
         String modelId,
         MLModelState state,
         FunctionName functionName,
@@ -92,6 +93,7 @@ public synchronized void initModelStateLocal(
         modelCache.setDeployToAllNodes(false);
         modelCache.setLastAccessTime(Instant.now());
         modelCaches.put(modelId, modelCache);
+        setIsAutoDeploying(modelId, true);
     }
 
     /**
@@ -279,6 +281,28 @@ public Boolean getIsModelEnabled(String modelId) {
         return modelCache.getIsModelEnabled();
     }
 
+    /**
+     * Set a flag to show if model is in auto deploying status
+     *
+     * @param modelId        model id
+     * @param isModelAutoDeploying auto deploy flag
+     */
+    public synchronized void setIsAutoDeploying(String modelId, Boolean isModelAutoDeploying) {
+        log.debug("Setting the auto deploying flag for Model {}", modelId);
+        getExistingModelCache(modelId).setIsAutoDeploying(isModelAutoDeploying);
+    }
+
+    /**
+     * Check if model is in auto deploying.
+     *
+     * @param modelId model id
+     * @return true if model is auto deploying.
+     */
+    public boolean isAutoDeploying(String modelId) {
+        MLModelCache modelCache = modelCaches.get(modelId);
+        return modelCache != null && BooleanUtils.isTrue(modelCache.getIsAutoDeploying());
+    }
+
     /**
      * Set memory size estimation CPU/GPU
      * 

plugin/src/main/java/org/opensearch/ml/model/MLModelManager.java

@@ -983,13 +983,14 @@ public void deployModel(
         if (!autoDeployModel) {
             modelCacheHelper.initModelState(modelId, MLModelState.DEPLOYING, functionName, workerNodes, deployToAllNodes);
         } else {
-            modelCacheHelper.initModelStateLocal(modelId, MLModelState.DEPLOYING, functionName, workerNodes);
+            modelCacheHelper.initModelStateAutoDeploy(modelId, MLModelState.DEPLOYING, functionName, workerNodes);
         }
 
         try (ThreadContext.StoredContext context = client.threadPool().getThreadContext().stashContext()) {
             ActionListener<String> wrappedListener = ActionListener.runBefore(listener, () -> {
                 context.restore();
                 modelCacheHelper.removeAutoDeployModel(modelId);
+                modelCacheHelper.setIsAutoDeploying(modelId, false);
             });
             if (!autoDeployModel) {
                 checkAndAddRunningTask(mlTask, maxDeployTasksPerNode);