From 97acb9ef24094042548a2981bdd99767156caeb3 Mon Sep 17 00:00:00 2001
From: Andrew Purtell <apurtell@apache.org>
Date: Fri, 24 Oct 2014 14:17:46 -0700
Subject: [PATCH] HBASE-11964 Improve spreading replication load from failed
 regionservers

---
 .../hbase/replication/regionserver/ReplicationSource.java  | 7 ++++---
 .../replication/regionserver/ReplicationSourceManager.java | 3 ++-
 .../hadoop/hbase/replication/TestReplicationBase.java      | 2 ++
 .../regionserver/TestReplicationSourceManager.java         | 2 ++
 4 files changed, 10 insertions(+), 4 deletions(-)

diff --git a/hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/ReplicationSource.java b/hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/ReplicationSource.java
index 04a2ad66237..02b97f2309f 100644
--- a/hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/ReplicationSource.java
+++ b/hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/ReplicationSource.java
@@ -161,7 +161,10 @@ public class ReplicationSource extends Thread
         this.conf.getLong("replication.source.size.capacity", 1024*1024*64);
     this.replicationQueueNbCapacity =
         this.conf.getInt("replication.source.nb.capacity", 25000);
-    this.maxRetriesMultiplier = this.conf.getInt("replication.source.maxretriesmultiplier", 10);
+    this.sleepForRetries =
+        this.conf.getLong("replication.source.sleepforretries", 1000);    // 1 second
+    this.maxRetriesMultiplier =
+        this.conf.getInt("replication.source.maxretriesmultiplier", 300); // 5 minutes @ 1 sec per
     this.queue =
         new PriorityBlockingQueue<Path>(
             this.conf.getInt("hbase.regionserver.maxlogs", 32),
@@ -171,8 +174,6 @@ public class ReplicationSource extends Thread
     this.replicationQueues = replicationQueues;
     this.replicationPeers = replicationPeers;
     this.manager = manager;
-    this.sleepForRetries =
-        this.conf.getLong("replication.source.sleepforretries", 1000);
     this.fs = fs;
     this.metrics = metrics;
     this.repLogReader = new ReplicationHLogReaderManager(this.fs, this.conf);
diff --git a/hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/ReplicationSourceManager.java b/hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/ReplicationSourceManager.java
index cb0f6ce68ff..9a095433d91 100644
--- a/hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/ReplicationSourceManager.java
+++ b/hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/ReplicationSourceManager.java
@@ -137,7 +137,8 @@ public class ReplicationSourceManager implements ReplicationListener {
     this.fs = fs;
     this.logDir = logDir;
     this.oldLogDir = oldLogDir;
-    this.sleepBeforeFailover = conf.getLong("replication.sleep.before.failover", 2000);
+    this.sleepBeforeFailover =
+        conf.getLong("replication.sleep.before.failover", 30000); // 30 seconds
     this.clusterId = clusterId;
     this.replicationTracker.registerListener(this);
     this.replicationPeers.getAllPeerIds();
diff --git a/hbase-server/src/test/java/org/apache/hadoop/hbase/replication/TestReplicationBase.java b/hbase-server/src/test/java/org/apache/hadoop/hbase/replication/TestReplicationBase.java
index 9fe2e0bcc2c..ea524a18896 100644
--- a/hbase-server/src/test/java/org/apache/hadoop/hbase/replication/TestReplicationBase.java
+++ b/hbase-server/src/test/java/org/apache/hadoop/hbase/replication/TestReplicationBase.java
@@ -99,6 +99,8 @@ public class TestReplicationBase {
     conf1.setLong(HConstants.THREAD_WAKE_FREQUENCY, 100);
     conf1.setInt("replication.stats.thread.period.seconds", 5);
     conf1.setBoolean("hbase.tests.use.shortcircuit.reads", false);
+    conf1.setLong("replication.sleep.before.failover", 2000);
+    conf1.setInt("replication.source.maxretriesmultiplier", 10);
 
     utility1 = new HBaseTestingUtility(conf1);
     utility1.startMiniZKCluster();
diff --git a/hbase-server/src/test/java/org/apache/hadoop/hbase/replication/regionserver/TestReplicationSourceManager.java b/hbase-server/src/test/java/org/apache/hadoop/hbase/replication/regionserver/TestReplicationSourceManager.java
index 084f8087d60..43b37d01f08 100644
--- a/hbase-server/src/test/java/org/apache/hadoop/hbase/replication/regionserver/TestReplicationSourceManager.java
+++ b/hbase-server/src/test/java/org/apache/hadoop/hbase/replication/regionserver/TestReplicationSourceManager.java
@@ -129,6 +129,8 @@ public class TestReplicationSourceManager {
         ReplicationSourceDummy.class.getCanonicalName());
     conf.setBoolean(HConstants.REPLICATION_ENABLE_KEY,
         HConstants.REPLICATION_ENABLE_DEFAULT);
+    conf.setLong("replication.sleep.before.failover", 2000);
+    conf.setInt("replication.source.maxretriesmultiplier", 10);
     utility = new HBaseTestingUtility(conf);
     utility.startMiniZKCluster();