Merged #132 from twitter_only to master (#134).

hraven-core/src/main/java/com/twitter/hraven/Constants.java

@@ -187,6 +187,10 @@ public class Constants {
   public static final String SLOTS_MILLIS_REDUCES = "SLOTS_MILLIS_REDUCES";
   public static final String REDUCE_SHUFFLE_BYTES = "REDUCE_SHUFFLE_BYTES";
 
+  /** MB MILLIS counters in hadoop2 */
+  public static final String MB_MILLIS_MAPS  = "MB_MILLIS_MAPS";
+  public static final String MB_MILLIS_REDUCES = "MB_MILLIS_REDUCES";
+
   /** used to indicate how expensive a job is in terms of memory and time taken*/
   public static final String MEGABYTEMILLIS = "megabytemillis" ;
   public static final byte[] MEGABYTEMILLIS_BYTES = Bytes.toBytes(MEGABYTEMILLIS);

hraven-etl/src/main/java/com/twitter/hraven/etl/JobHistoryFileParserHadoop2.java

@@ -87,6 +87,11 @@ public class JobHistoryFileParserHadoop2 extends JobHistoryFileParserBase {
    */
   public static final String JOB_STATUS_SUCCEEDED = "SUCCEEDED";
 
+  // initialize mb millis to NOTFOUND_VALUE
+  // since older versions of hadoop2 did not have this counter
+  private long mapMbMillis = Constants.NOTFOUND_VALUE;
+  private long reduceMbMillis = Constants.NOTFOUND_VALUE;
+
   private static final String LAUNCH_TIME_KEY_STR = JobHistoryKeys.LAUNCH_TIME.toString();
   private static final String FINISH_TIME_KEY_STR = JobHistoryKeys.FINISH_TIME.toString();
 
@@ -705,7 +710,18 @@ private void populatePut(Put p, byte[] family, String key, String groupName, Str
     byte[] groupPrefix = Bytes.add(counterPrefix, Bytes.toBytes(groupName), Constants.SEP_BYTES);
     byte[] qualifier = Bytes.add(groupPrefix, Bytes.toBytes(counterName));
 
-    /**
+    /*
+     * store the map and reduce mb millis counter value
+     */
+    if (Constants.JOB_COUNTER_HADOOP2.equals(groupName)) {
+      if (Constants.MB_MILLIS_MAPS.equals(counterName)) {
+        this.mapMbMillis = counterValue;
+      } else if (Constants.MB_MILLIS_REDUCES.equals(counterName)) {
+          this.reduceMbMillis = counterValue;
+      }
+    }
+
+    /*
      * correct and populate map and reduce slot millis
      */
     if ((Constants.SLOTS_MILLIS_MAPS.equals(counterName)) ||
@@ -739,10 +755,13 @@ private long getMemoryMb(String key) {
    * hadoop2ReportedMapSlotMillis * yarn.scheduler.minimum-allocation-mb
    *        / mapreduce.mapreduce.memory.mb
    * similarly for reduce slot millis
+   *
+   * Marking method as deprecated as noted in Pull Request #132
    * @param counterName
    * @param counterValue
    * @return corrected counter value
    */
+  @Deprecated
   private Long getStandardizedCounterValue(String counterName, Long counterValue) {
     if (jobConf == null) {
       throw new ProcessingException("While correcting slot millis, jobConf is null");
@@ -879,17 +898,30 @@ public Long getMegaByteMillis() {
           + " since " + Constants.AM_MEMORY_MB_CONF_KEY + " not found!");
     }
 
-    long mbMillis = 0L;
     long mapSlotMillis = this.jobDetails.getMapSlotMillis();
     long reduceSlotMillis = this.jobDetails.getReduceSlotMillis();
+
+    /* in case of older versions of hadoop2
+     *  the counter of mb millis is not available
+     * then use slot millis counter value
+     */
+    if (this.mapMbMillis == Constants.NOTFOUND_VALUE) {
+      this.mapMbMillis = (mapMb * mapSlotMillis);
+    }
+    if (this.reduceMbMillis == Constants.NOTFOUND_VALUE) {
+       this.reduceMbMillis = (reduceMb * reduceSlotMillis);
+    }
+
+    long mbMillis = 0L;
     if (uberized) {
       mbMillis = amMb * jobRunTime;
     } else {
-      mbMillis = (mapMb * mapSlotMillis) + (reduceMb * reduceSlotMillis) + (amMb * jobRunTime);
+      mbMillis = this.mapMbMillis + this.reduceMbMillis + (amMb * jobRunTime);
     }
 
-    LOG.debug("For " + jobKey.toString() + " " + Constants.MEGABYTEMILLIS + " is " + mbMillis
-        + " since \n uberized: " + uberized + " \n " + "mapMb: " + mapMb + " mapSlotMillis: "
+    LOG.debug("For " + jobKey.toString() + "\n" + Constants.MEGABYTEMILLIS + " is " + mbMillis
+        + " since \n uberized: " + uberized + " \n " + "mapMbMillis: " + mapMbMillis
+        + " reduceMbMillis:" + reduceMbMillis + "mapMb: " + mapMb + " mapSlotMillis: "
         + mapSlotMillis + " \n " + " reduceMb: " + reduceMb + " reduceSlotMillis: "
         + reduceSlotMillis + " \n " + " amMb: " + amMb + " jobRunTime: " + jobRunTime
         + " start time: " + startTime + " endtime " + endTime);
@@ -902,4 +934,20 @@ public Long getMegaByteMillis() {
   public JobDetails getJobDetails() {
     return this.jobDetails;
   }
+
+  public long getMapMbMillis() {
+    return mapMbMillis;
+  }
+
+  public void setMapMbMillis(long mapMbMillis) {
+    this.mapMbMillis = mapMbMillis;
+  }
+
+  public long getReduceMbMillis() {
+    return reduceMbMillis;
+  }
+
+  public void setReduceMbMillis(long reduceMbMillis) {
+    this.reduceMbMillis = reduceMbMillis;
+  }
 }

hraven-etl/src/test/java/com/twitter/hraven/etl/TestJobHistoryFileParserHadoop2.java

@@ -203,19 +203,24 @@ public void testMissingSlotMillis() throws IOException {
     Configuration jobConf = new Configuration();
     jobConf.addResource(new Path(JOB_CONF_FILE_NAME));
 
-    JobHistoryFileParser historyFileParser =
-        JobHistoryFileParserFactory.createJobHistoryFileParser(contents, jobConf);
+    JobHistoryFileParserHadoop2 historyFileParser =
+        new JobHistoryFileParserHadoop2(jobConf);
     assertNotNull(historyFileParser);
 
-    // confirm that we get back an object that can parse hadoop 2.0 files
-    assertTrue(historyFileParser instanceof JobHistoryFileParserHadoop2);
     JobKey jobKey = new JobKey("cluster1", "user", "Sleep", 1, "job_1329348432655_0001");
     historyFileParser.parse(contents, jobKey);
 
     // this history file has only map slot millis no reduce millis
+    Long mapMbMillis = historyFileParser.getMapMbMillis();
+    assertNotNull(mapMbMillis);
+    assertEquals(mapMbMillis, new Long(178169856L));
+    Long reduceMbMillis = historyFileParser.getReduceMbMillis();
+    assertNotNull(reduceMbMillis);
+    assertEquals(reduceMbMillis, Constants.NOTFOUND_VALUE);
+
     Long mbMillis = historyFileParser.getMegaByteMillis();
     assertNotNull(mbMillis);
-    Long expValue = 10402816L;
+    Long expValue = 188559872L;
     assertEquals(expValue, mbMillis);
   }
 
@@ -307,4 +312,5 @@ public void testVersion2_4() throws IOException {
     // ensure that we got the hadoop2 version put
     assertTrue(foundQueue);
   }
+
 }