HDFS-5591. Checkpointing should use monotonic time when calculating period. Contributed by Charles Lamb.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1583926 13f79535-47bb-0310-9956-ffa450edef68

Author: umbrant

hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt

@@ -299,6 +299,9 @@ Release 2.5.0 - UNRELEASED
     HDFS-6173. Move the default processor from Ls to Web in OfflineImageViewer.
     (Akira Ajisaka via wheat9)
 
+    HDFS-5591. Checkpointing should use monotonic time when calculating period.
+    (Charles Lamb via wang)
+
 Release 2.4.1 - UNRELEASED
 
   INCOMPATIBLE CHANGES

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/SecondaryNameNode.java

@@ -129,7 +129,9 @@ public String toString() {
     return getClass().getSimpleName() + " Status" 
       + "\nName Node Address    : " + nameNodeAddr   
       + "\nStart Time           : " + new Date(starttime)
-      + "\nLast Checkpoint Time : " + (lastCheckpointTime == 0? "--": new Date(lastCheckpointTime))
+      + "\nLast Checkpoint      : " + (lastCheckpointTime == 0? "--":
+				       ((Time.monotonicNow() - lastCheckpointTime) / 1000))
+	                            + " seconds ago"
       + "\nCheckpoint Period    : " + checkpointConf.getPeriod() + " seconds"
       + "\nCheckpoint Size      : " + StringUtils.byteDesc(checkpointConf.getTxnCount())
                                     + " (= " + checkpointConf.getTxnCount() + " bytes)" 
@@ -376,7 +378,7 @@ public void doWork() {
         if(UserGroupInformation.isSecurityEnabled())
           UserGroupInformation.getCurrentUser().checkTGTAndReloginFromKeytab();
 
-        long now = Time.now();
+        final long now = Time.monotonicNow();
 
         if (shouldCheckpointBasedOnCount() ||
             now >= lastCheckpointTime + 1000 * checkpointConf.getPeriod()) {

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/ha/StandbyCheckpointer.java

@@ -17,7 +17,7 @@
  */
 package org.apache.hadoop.hdfs.server.namenode.ha;
 
-import static org.apache.hadoop.util.Time.now;
+import static org.apache.hadoop.util.Time.monotonicNow;
 
 import java.io.IOException;
 import java.net.URI;
@@ -277,14 +277,14 @@ public Object run() {
      * prevented
      */
     private void preventCheckpointsFor(long delayMs) {
-      preventCheckpointsUntil = now() + delayMs;
+      preventCheckpointsUntil = monotonicNow() + delayMs;
     }
 
     private void doWork() {
       final long checkPeriod = 1000 * checkpointConf.getCheckPeriod();
       // Reset checkpoint time so that we don't always checkpoint
       // on startup.
-      lastCheckpointTime = now();
+      lastCheckpointTime = monotonicNow();
       while (shouldRun) {
         boolean needRollbackCheckpoint = namesystem.isNeedRollbackFsImage();
         if (!needRollbackCheckpoint) {
@@ -302,9 +302,9 @@ private void doWork() {
             UserGroupInformation.getCurrentUser().checkTGTAndReloginFromKeytab();
           }
 
-          long now = now();
-          long uncheckpointed = countUncheckpointedTxns();
-          long secsSinceLast = (now - lastCheckpointTime)/1000;
+          final long now = monotonicNow();
+          final long uncheckpointed = countUncheckpointedTxns();
+          final long secsSinceLast = (now - lastCheckpointTime) / 1000;
 
           boolean needCheckpoint = needRollbackCheckpoint;
           if (needCheckpoint) {

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestSecondaryWebUi.java

@@ -62,7 +62,8 @@ public static void shutDownCluster() {
   public void testSecondaryWebUi() throws IOException {
     String pageContents = DFSTestUtil.urlGet(new URL("http://localhost:" +
         SecondaryNameNode.getHttpAddress(conf).getPort() + "/status.jsp"));
-    assertTrue(pageContents.contains("Last Checkpoint Time"));
+    assertTrue("Didn't find \"Last Checkpoint\"",
+        pageContents.contains("Last Checkpoint"));
   }
 
   @Test