Limit RTG exceptions to retry on (#105003)

Limit RTG exceptions to retry on and do not throw in `getCurrentNodeOfPrimary`. Follow up to #104579 (comment). Relates ES-5727
elastic · Feb 1, 2024 · e2d68ae · e2d68ae
1 parent 149ec37
commit e2d68ae
Showing 1 changed file with 9 additions and 5 deletions.
diff --git a/server/src/main/java/org/elasticsearch/action/get/TransportGetAction.java b/server/src/main/java/org/elasticsearch/action/get/TransportGetAction.java
@@ -17,7 +17,6 @@
 import org.elasticsearch.action.ActionRunnable;
 import org.elasticsearch.action.ActionType;
 import org.elasticsearch.action.NoShardAvailableActionException;
-import org.elasticsearch.action.UnavailableShardsException;
 import org.elasticsearch.action.admin.indices.refresh.TransportShardRefreshAction;
 import org.elasticsearch.action.support.ActionFilters;
 import org.elasticsearch.action.support.replication.BasicReplicationRequest;
@@ -194,6 +193,10 @@ private void handleGetOnUnpromotableShard(GetRequest request, IndexShard indexSh
         ShardId shardId = indexShard.shardId();
         if (request.refresh()) {
             var node = getCurrentNodeOfPrimary(clusterService.state(), shardId);
+            if (node == null) {
+                listener.onFailure(new NoShardAvailableActionException(shardId, "primary shard is not active"));
+                return;
+            }
             logger.trace("send refresh action for shard {} to node {}", shardId, node.getId());
             var refreshRequest = new BasicReplicationRequest(shardId);
             refreshRequest.setParentTask(request.getParentTask());
@@ -230,10 +233,7 @@ private void getFromTranslog(
         tryGetFromTranslog(request, indexShard, state, listener.delegateResponse((l, e) -> {
             final var cause = ExceptionsHelper.unwrapCause(e);
             logger.debug("get_from_translog failed", cause);
-            if (cause instanceof ShardNotFoundException
-                || cause instanceof IndexNotFoundException
-                || cause instanceof NoShardAvailableActionException
-                || cause instanceof UnavailableShardsException) {
+            if (cause instanceof ShardNotFoundException || cause instanceof IndexNotFoundException) {
                 logger.debug("retrying get_from_translog");
                 observer.waitForNextChange(new ClusterStateObserver.Listener() {
                     @Override
@@ -260,6 +260,10 @@ public void onTimeout(TimeValue timeout) {
     private void tryGetFromTranslog(GetRequest request, IndexShard indexShard, ClusterState state, ActionListener<GetResponse> listener) {
         ShardId shardId = indexShard.shardId();
         var node = getCurrentNodeOfPrimary(state, shardId);
+        if (node == null) {
+            listener.onFailure(new NoShardAvailableActionException(shardId, "primary shard is not active"));
+            return;
+        }
         TransportGetFromTranslogAction.Request getFromTranslogRequest = new TransportGetFromTranslogAction.Request(request, shardId);
         getFromTranslogRequest.setParentTask(request.getParentTask());
         transportService.sendRequest(