IGNITE-23395 Remove excessive raft subsystem log spam with network ex…

…ceptions (apache#4761)
12rcu · Nov 21, 2024 · ee004ae · ee004ae
1 parent dcaad51
commit ee004ae
Show file tree

Hide file tree

Showing 2 changed files with 16 additions and 3 deletions.
diff --git a/modules/raft/src/main/java/org/apache/ignite/raft/jraft/core/ReplicatorGroupImpl.java b/modules/raft/src/main/java/org/apache/ignite/raft/jraft/core/ReplicatorGroupImpl.java
@@ -119,9 +119,14 @@ public boolean addReplicator(final PeerId peer, final ReplicatorType replicatorT
         assert client != null;
 
         if (!client.connect(peer)) {
-            LOG.error("Fail to check replicator connection to peer={}, replicatorType={}.", peer, replicatorType);
+            if (!failureReplicators.containsKey(peer)) {
+                LOG.error("Fail to check replicator connection to peer={}, replicatorType={}.", peer, replicatorType);
+            }
+
             this.failureReplicators.put(peer, replicatorType);
             return false;
+        } else {
+            failureReplicators.remove(peer);
         }
 
 //        if (!sync) {

diff --git a/modules/raft/src/main/java/org/apache/ignite/raft/jraft/rpc/impl/AbstractClientService.java b/modules/raft/src/main/java/org/apache/ignite/raft/jraft/rpc/impl/AbstractClientService.java
@@ -21,7 +21,7 @@
 import java.net.ConnectException;
 import java.util.Set;
 import java.util.concurrent.CompletableFuture;
-import java.util.concurrent.ExecutionException;
+import java.util.concurrent.ConcurrentHashMap;import java.util.concurrent.ExecutionException;
 import java.util.concurrent.Executor;
 import java.util.concurrent.ExecutorService;
 import org.apache.ignite.internal.logger.IgniteLogger;
@@ -57,6 +57,8 @@ public abstract class AbstractClientService implements ClientService, TopologyEv
     protected ExecutorService rpcExecutor;
     protected RpcOptions rpcOptions;
 
+    private Set<PeerId> deadPeers = ConcurrentHashMap.newKeySet();
+
     /**
      * The set of pinged consistent IDs.
      */
@@ -121,7 +123,11 @@ public boolean connect(final PeerId peerId) {
 
             LOG.error("Interrupted while connecting to {}, exception: {}.", peerId, e.getMessage());
         } catch (ExecutionException e) {
-            LOG.error("Fail to connect {}, exception: {}.", peerId, e.getMessage());
+            if (!deadPeers.contains(peerId)) {
+                deadPeers.add(peerId);
+
+                LOG.error("Fail to connect {}, exception: {}.", peerId, e.getMessage());
+            }
         }
 
         return false;
@@ -153,6 +159,8 @@ public CompletableFuture<Boolean> connectAsync(PeerId peerId) {
             if (resp != null && resp.errorCode() == 0) {
                 readyConsistentIds.add(peerId.getConsistentId());
 
+                deadPeers.remove(peerId);
+
                 return true;
             } else {
                 return false;