HBASE-698 HLog recovery is not performed after master failure

git-svn-id: https://svn.apache.org/repos/asf/hadoop/hbase/trunk@792964 13f79535-47bb-0310-9956-ffa450edef68

CHANGES.txt

@@ -458,6 +458,7 @@ Release 0.20.0 - Unreleased
    HBASE-1637  Delete client class methods should return itself like Put, Get,
                Scan (Jon Gray via Nitay)
    HBASE-1640  Allow passing arguments to jruby script run when run by bin/hbase shell
+   HBASE-698   HLog recovery is not performed after master failure
 
   OPTIMIZATIONS
    HBASE-1412  Change values for delete column and column family in KeyValue

src/java/org/apache/hadoop/hbase/master/BaseScanner.java

@@ -42,7 +42,6 @@ import org.apache.hadoop.hbase.client.Delete;
 import org.apache.hadoop.hbase.client.Result;
 import org.apache.hadoop.hbase.client.Scan;
 import org.apache.hadoop.hbase.ipc.HRegionInterface;
-import org.apache.hadoop.hbase.regionserver.HLog;
 import org.apache.hadoop.hbase.regionserver.HRegion;
 import org.apache.hadoop.hbase.regionserver.Store;
 import org.apache.hadoop.hbase.regionserver.StoreFile;
@@ -372,33 +371,6 @@ abstract class BaseScanner extends Chore implements HConstants {
             " is not valid; " + " Server '" + serverAddress + "' startCode: " +
             startCode + " unknown.");
         }
-
-        // Recover the region server's log if there is one.
-        // This is only done from here if we are restarting and there is stale
-        // data in the meta region. Once we are on-line, dead server log
-        // recovery is handled by lease expiration and ProcessServerShutdown
-        if (!this.master.regionManager.isInitialMetaScanComplete() &&
-            serverName != null) {
-          Path logDir =
-            new Path(this.master.rootdir, HLog.getHLogDirectoryName(serverName));
-          try {
-            if (master.fs.exists(logDir)) {
-              this.master.regionManager.splitLogLock.lock();
-              try {
-                HLog.splitLog(master.rootdir, logDir, master.fs,
-                    master.getConfiguration());
-              } finally {
-                this.master.regionManager.splitLogLock.unlock();
-              }
-            }
-            if (LOG.isDebugEnabled()) {
-              LOG.debug("Split " + logDir.toString());
-            }
-          } catch (IOException e) {
-            LOG.warn("unable to split region server log because: ", e);
-            throw e;
-          }
-        }
         // Now get the region assigned
         this.master.regionManager.setUnassigned(info, true);
       }

src/java/org/apache/hadoop/hbase/master/HMaster.java

@@ -38,6 +38,7 @@ import java.util.concurrent.atomic.AtomicBoolean;
 
 import org.apache.commons.logging.Log;
 import org.apache.commons.logging.LogFactory;
+import org.apache.hadoop.fs.FileStatus;
 import org.apache.hadoop.fs.FileSystem;
 import org.apache.hadoop.fs.Path;
 import org.apache.hadoop.hbase.ClusterStatus;
@@ -71,6 +72,7 @@ import org.apache.hadoop.hbase.ipc.HMasterInterface;
 import org.apache.hadoop.hbase.ipc.HMasterRegionInterface;
 import org.apache.hadoop.hbase.ipc.HRegionInterface;
 import org.apache.hadoop.hbase.master.metrics.MasterMetrics;
+import org.apache.hadoop.hbase.regionserver.HLog;
 import org.apache.hadoop.hbase.regionserver.HRegion;
 import org.apache.hadoop.hbase.util.Bytes;
 import org.apache.hadoop.hbase.util.FSUtils;
@@ -543,6 +545,7 @@ public class HMaster extends Thread implements HConstants, HMasterInterface,
       // Check if this is a fresh start of the cluster
       if(addresses.size() == 0) {
         LOG.debug("This is a fresh start, proceeding with normal startup");
+        splitLogAfterStartup();
         return;
       }
       LOG.info("This is a failover, ZK inspection begins...");
@@ -579,11 +582,46 @@ public class HMaster extends Thread implements HConstants, HMasterInterface,
       }
       LOG.info("Inspection found " + assignedRegions.size() + " regions, " + 
           (isRootRegionAssigned ? "with -ROOT-" : "but -ROOT- was MIA"));
+      splitLogAfterStartup();
     } catch(IOException ex) {
       ex.printStackTrace();
     }
   }
   
+  /**
+   * Inspect the log directory to recover any log file without
+   * and active region server.
+   * @throws IOException
+   */
+  private void splitLogAfterStartup() throws IOException {
+    Path logsDirPath =
+      new Path(this.rootdir, HConstants.HREGION_LOGDIR_NAME);
+    FileStatus [] logFolders = this.fs.listStatus(logsDirPath);
+    if (logFolders == null || logFolders.length == 0) {
+      LOG.debug("No log files to split, proceeding...");
+      return;
+    }
+    for (FileStatus status : logFolders) {
+      String serverName = status.getPath().getName();
+      LOG.info("Found log folder : " + serverName);
+      if(this.serverManager.getServerInfo(serverName) == null) {
+        LOG.info("Log folder doesn't belong " +
+            "to a known region server, splitting");
+        this.regionManager.splitLogLock.lock();
+        Path logDir =
+          new Path(this.rootdir, HLog.getHLogDirectoryName(serverName));
+        try {
+          HLog.splitLog(this.rootdir, logDir, this.fs,
+              getConfiguration());
+        } finally {
+          this.regionManager.splitLogLock.unlock();
+        }
+      } else {
+        LOG.info("Log folder belongs to an existing region server");
+      }
+    }
+  }
+
   /*
    * Start up all services. If any of these threads gets an unhandled exception
    * then they just die with a logged message.  This should be fine because