HBASE-24758 Avoid flooding replication source RSes logs when no sinks… (#2118)

Signed-off-by: Josh Elser <elserj@apache.org>
Signed-off-by: Viraj Jasani <vjasani@apache.org>

(cherry picked from commit 8c0d7fa)

Author: wchevreuil

hbase-server/src/main/java/org/apache/hadoop/hbase/replication/HBaseReplicationEndpoint.java

@@ -168,8 +168,8 @@ protected static List<ServerName> fetchSlavesAddresses(ZKWatcher zkw)
   }
 
   /**
-   * Get a list of all the addresses of all the region servers
-   * for this peer cluster
+   * Get a list of all the addresses of all the available region servers
+   * for this peer cluster, or an empty list if no region servers available at peer cluster.
    * @return list of addresses
    */
   // Synchronize peer cluster connection attempts to avoid races and rate

hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/HBaseInterClusterReplicationEndpoint.java

@@ -128,6 +128,8 @@ public class HBaseInterClusterReplicationEndpoint extends HBaseReplicationEndpoi
   private boolean dropOnDeletedTables;
   private boolean dropOnDeletedColumnFamilies;
   private boolean isSerial = false;
+  //Initialising as 0 to guarantee at least one logging message
+  private long lastSinkFetchTime = 0;
 
   /*
    * Some implementations of HBaseInterClusterReplicationEndpoint may require instantiating
@@ -518,8 +520,14 @@ public boolean replicate(ReplicateContext replicateContext) {
 
     int numSinks = replicationSinkMgr.getNumSinks();
     if (numSinks == 0) {
-      LOG.warn("{} No replication sinks found, returning without replicating. "
-        + "The source should retry with the same set of edits.", logPeerId());
+      if((System.currentTimeMillis() - lastSinkFetchTime) >= (maxRetriesMultiplier*1000)) {
+        LOG.warn(
+          "No replication sinks found, returning without replicating. "
+            + "The source should retry with the same set of edits. Not logging this again for "
+            + "the next {} seconds.", maxRetriesMultiplier);
+        lastSinkFetchTime = System.currentTimeMillis();
+      }
+      sleepForRetries("No sinks available at peer", sleepMultiplier);
       return false;
     }
 

hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/ReplicationSinkManager.java

@@ -160,6 +160,9 @@ public synchronized void reportSinkSuccess(SinkPeer sinkPeer) {
    */
   public synchronized void chooseSinks() {
     List<ServerName> slaveAddresses = endpoint.getRegionServers();
+    if(slaveAddresses.isEmpty()){
+      LOG.warn("No sinks available at peer. Will not be able to replicate");
+    }
     Collections.shuffle(slaveAddresses, random);
     int numSinks = (int) Math.ceil(slaveAddresses.size() * ratio);
     sinks = slaveAddresses.subList(0, numSinks);