From 14b68ca4c0b21ced1f4ff2bc484e9b2c0f1a17cb Mon Sep 17 00:00:00 2001
From: Michael Stack <stack@apache.org>
Date: Wed, 5 May 2010 22:14:05 +0000
Subject: [PATCH] HBASE-2513 hbase-2414 added bug where we'd tight-loop if no
 root available

git-svn-id: https://svn.apache.org/repos/asf/hadoop/hbase/trunk@941517 13f79535-47bb-0310-9956-ffa450edef68
---
 .../apache/hadoop/hbase/master/HMaster.java    |  9 ++++++---
 .../master/RegionServerOperationQueue.java     | 18 ++++++++++--------
 2 files changed, 16 insertions(+), 11 deletions(-)

diff --git a/core/src/main/java/org/apache/hadoop/hbase/master/HMaster.java b/core/src/main/java/org/apache/hadoop/hbase/master/HMaster.java
index ed31cbcd996..5bc98b9e1e8 100644
--- a/core/src/main/java/org/apache/hadoop/hbase/master/HMaster.java
+++ b/core/src/main/java/org/apache/hadoop/hbase/master/HMaster.java
@@ -445,14 +445,17 @@ public class HMaster extends Thread implements HConstants, HMasterInterface,
             break;
           }
         }
-        boolean doDelayQueue = this.regionManager.getRootRegionLocation() != null;
-        switch (this.regionServerOperationQueue.process(doDelayQueue)) {
+        final HServerAddress root = this.regionManager.getRootRegionLocation();
+        switch (this.regionServerOperationQueue.process(root)) {
         case FAILED:
+            // If FAILED op processing, bad. Exit.
           break FINISHED;
         case REQUEUED_BUT_PROBLEM:
           if (!checkFileSystem())
+              // If bad filesystem, exit.
             break FINISHED;
-        default: // PROCESSED, NOOP, REQUEUED:
+          default:
+            // Continue run loop if conditions are PROCESSED, NOOP, REQUEUED
           break;
         }
       }
diff --git a/core/src/main/java/org/apache/hadoop/hbase/master/RegionServerOperationQueue.java b/core/src/main/java/org/apache/hadoop/hbase/master/RegionServerOperationQueue.java
index 53b0bd787d0..c85c1414b08 100644
--- a/core/src/main/java/org/apache/hadoop/hbase/master/RegionServerOperationQueue.java
+++ b/core/src/main/java/org/apache/hadoop/hbase/master/RegionServerOperationQueue.java
@@ -33,6 +33,7 @@ import org.apache.commons.logging.LogFactory;
 import org.apache.hadoop.conf.Configuration;
 import org.apache.hadoop.hbase.HMsg;
 import org.apache.hadoop.hbase.HServerInfo;
+import org.apache.hadoop.hbase.HServerAddress;
 import org.apache.hadoop.hbase.RemoteExceptionHandler;
 import org.apache.hadoop.hbase.util.Sleeper;
 import org.apache.hadoop.ipc.RemoteException;
@@ -109,19 +110,20 @@ public class RegionServerOperationQueue {
 
   /**
    * Try to get an operation off of the queue and process it.
-   * @param skipDelayedToDos If true, do not do delayed todos first but instead
-   * move straight to the current todos list.  This is set when we want to be
-   * sure that recently queued events are processed first such as the onlining
-   * of root region (Root region needs to be online before we can do meta
-   * onlining; meta onlining needs to be done before we can do... and so on). 
+   * @param rootRegionLocation Location of the root region.
    * @return {@link ProcessingResultCode#PROCESSED},
    * {@link ProcessingResultCode#REQUEUED},
    * {@link ProcessingResultCode#REQUEUED_BUT_PROBLEM}
    */ 
-  public synchronized ProcessingResultCode process(final boolean skipDelayedToDos) {
-    RegionServerOperation op = delayedToDoQueue.poll();
+  public synchronized ProcessingResultCode process(final HServerAddress rootRegionLocation) {
+    RegionServerOperation op = null;
+    // Only process the delayed queue if root region is online.  If offline,
+    // the operation to put it online is probably in the toDoQueue.  Process
+    // it first.
+    if (rootRegionLocation != null) {
+      op = delayedToDoQueue.poll();
+    } else {
     // if there aren't any todo items in the queue, sleep for a bit.
-    if (op == null) {
       try {
         op = toDoQueue.poll(threadWakeFrequency, TimeUnit.MILLISECONDS);
       } catch (InterruptedException e) {