HBASE-7720 Improve logging messages of failed snapshot attempts.

git-svn-id: https://svn.apache.org/repos/asf/hbase/branches/hbase-7290@1445867 13f79535-47bb-0310-9956-ffa450edef68
ndimiduk · Feb 13, 2013 · 81a4387 · 81a4387
1 parent f60820c
commit 81a4387
Show file tree

Hide file tree

Showing 4 changed files with 43 additions and 2 deletions.
diff --git a/hbase-server/src/main/java/org/apache/hadoop/hbase/master/snapshot/SnapshotManager.java b/hbase-server/src/main/java/org/apache/hadoop/hbase/master/snapshot/SnapshotManager.java
@@ -51,6 +51,7 @@
 import org.apache.hadoop.hbase.master.SnapshotSentinel;
 import org.apache.hadoop.hbase.master.cleaner.HFileCleaner;
 import org.apache.hadoop.hbase.master.cleaner.HFileLinkCleaner;
+import org.apache.hadoop.hbase.procedure.Procedure;
 import org.apache.hadoop.hbase.procedure.ProcedureCoordinator;
 import org.apache.hadoop.hbase.procedure.ProcedureCoordinatorRpcs;
 import org.apache.hadoop.hbase.procedure.ZKProcedureCoordinatorRpcs;
@@ -322,7 +323,15 @@ public boolean isSnapshotDone(SnapshotDescription expected) throws IOException {
     try {
       handler.rethrowException();
     } catch (ForeignException e) {
-      throw new HBaseSnapshotException("Snapshot " + ssString +  " had an error from RS", e,
+      // Give some procedure info on an exception.
+      String status;
+      Procedure p = coordinator.getProcedure(expected.getName());
+      if (p != null) {
+        status = p.getStatus();
+      } else {
+        status = expected.getName() + " not found in proclist " + coordinator.getProcedureNames();
+      }
+      throw new HBaseSnapshotException("Snapshot " + ssString +  " had an error.  " + status, e,
           expected);
     }
 

diff --git a/hbase-server/src/main/java/org/apache/hadoop/hbase/procedure/Procedure.java b/hbase-server/src/main/java/org/apache/hadoop/hbase/procedure/Procedure.java
@@ -164,6 +164,19 @@ public String getName() {
     return procName;
   }
 
+  /**
+   * Returns a copy of the procedure members still trying to enter the barrier.
+   * @return
+   */
+  public String getStatus() {
+    String waiting, done;
+    synchronized (joinBarrierLock) {
+      waiting = acquiringMembers.toString();
+      done = inBarrierMembers.toString();
+    }
+    return "Procedure " + procName + " { waiting=" + waiting + " done="+ done + " }";
+  }
+
   /**
    * Get the ExternalErrorDispatcher
    * @return the Procedure's monitor.

diff --git a/hbase-server/src/main/java/org/apache/hadoop/hbase/procedure/ProcedureCoordinator.java b/hbase-server/src/main/java/org/apache/hadoop/hbase/procedure/ProcedureCoordinator.java
@@ -19,7 +19,9 @@
 
 import java.io.IOException;
 import java.util.Collection;
+import java.util.HashSet;
 import java.util.List;
+import java.util.Set;
 import java.util.concurrent.ConcurrentMap;
 import java.util.concurrent.ExecutorService;
 import java.util.concurrent.Future;
@@ -246,4 +248,21 @@ void memberFinishedBarrier(String procName, final String member) {
   ProcedureCoordinatorRpcs getRpcs() {
     return rpcs;
   }
+
+  /**
+   * Returns the procedure.  This Procedure is a live instance so should not be modified but can
+   * be inspected.
+   * @param name Name of the procedure
+   * @return Procedure or null if not present any more
+   */
+  public Procedure getProcedure(String name) {
+    return procedures.get(name);
+  }
+
+  /**
+   * @return Return set of all procedure names.
+   */
+  public Set<String> getProcedureNames() {
+    return new HashSet<String>(procedures.keySet());
+  }
 }
diff --git a/hbase-server/src/main/java/org/apache/hadoop/hbase/procedure/ProcedureMember.java b/hbase-server/src/main/java/org/apache/hadoop/hbase/procedure/ProcedureMember.java
@@ -76,7 +76,7 @@ public static ThreadPoolExecutor defaultPool(long wakeFrequency, long keepAlive,
       int procThreads, String memberName) {
     return new ThreadPoolExecutor(1, procThreads, keepAlive, TimeUnit.SECONDS,
         new SynchronousQueue<Runnable>(),
-        new DaemonThreadFactory("( member-" + memberName + ") subprocedure-pool"));
+        new DaemonThreadFactory("member: '" + memberName + "' subprocedure-pool"));
   }
 
   /**