[HADOOP-1012] OutOfMemoryError in reduce

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Won't Fix
• Affects Version/s: 0.11.1
• Assignee: Devaraj Das
• Reporter: Nigel Daley
• Created: 12/Feb/07 22:27
• Updated: 08/Jul/09 16:52
• Resolved: 22/Feb/07 11:08

Description

I'm seeing OutOfMemoryErrors from a reduce in each of DFSIO Benchmark and RandomWriter. No stack traces are given. Snipets from the TaskTracker logs are below. I believe I first saw this on February 3rd during tests that I run weekly.

=====
DFSIO
=====
...
2007-02-10 18:25:20,201 INFO org.apache.hadoop.mapred.TaskRunner: task_0005_r_000000_0 Copying of all map outputs complete. Initiating the last merge on the remaining files in ramfs://mapoutput9105104
2007-02-10 18:25:20,771 INFO org.apache.hadoop.mapred.TaskTracker: task_0005_r_000000_0 0.33333334% reduce > copy (9000 of 9000 at 0.00 MB/s)
2007-02-10 18:25:21,773 INFO org.apache.hadoop.mapred.TaskTracker: task_0005_r_000000_0 0.33333334% reduce > copy (9000 of 9000 at 0.00 MB/s)
2007-02-10 18:25:23,280 INFO org.apache.hadoop.mapred.TaskTracker: task_0005_r_000000_0 0.33333334% reduce > copy (9000 of 9000 at 0.00 MB/s)
2007-02-10 18:25:24,607 INFO org.apache.hadoop.mapred.TaskTracker: task_0005_r_000000_0 0.33333334% reduce > copy (9000 of 9000 at 0.00 MB/s)
2007-02-10 18:25:25,960 INFO org.apache.hadoop.mapred.TaskTracker: task_0005_r_000000_0 0.33333334% reduce > copy (9000 of 9000 at 0.00 MB/s)
2007-02-10 18:25:27,105 INFO org.apache.hadoop.mapred.TaskTracker: task_0005_r_000000_0 0.33333334% reduce > copy (9000 of 9000 at 0.00 MB/s)
2007-02-10 18:25:28,982 INFO org.apache.hadoop.mapred.TaskTracker: task_0005_r_000000_0 0.33333334% reduce > copy (9000 of 9000 at 0.00 MB/s)
2007-02-10 18:25:29,984 INFO org.apache.hadoop.mapred.TaskTracker: task_0005_r_000000_0 0.33333334% reduce > copy (9000 of 9000 at 0.00 MB/s)
2007-02-10 18:25:31,481 INFO org.apache.hadoop.mapred.TaskTracker: task_0005_r_000000_0 0.33333334% reduce > copy (9000 of 9000 at 0.00 MB/s)
2007-02-10 18:25:33,379 INFO org.apache.hadoop.mapred.TaskTracker: task_0005_r_000000_0 0.33333334% reduce > copy (9000 of 9000 at 0.00 MB/s)
2007-02-10 18:25:34,478 INFO org.apache.hadoop.mapred.TaskTracker: task_0005_r_000000_0 0.33333334% reduce > copy (9000 of 9000 at 0.00 MB/s)
2007-02-10 18:25:35,656 INFO org.apache.hadoop.mapred.TaskTracker: task_0005_r_000000_0 0.33333334% reduce > copy (9000 of 9000 at 0.00 MB/s)
2007-02-10 18:25:36,758 INFO org.apache.hadoop.mapred.TaskTracker: task_0005_r_000000_0 0.33333334% reduce > copy (9000 of 9000 at 0.00 MB/s)
2007-02-10 18:25:42,593 INFO org.apache.hadoop.mapred.TaskTracker: task_0005_r_000000_0 0.33333334% reduce > copy (9000 of 9000 at 0.00 MB/s)
2007-02-10 18:25:43,600 INFO org.apache.hadoop.mapred.TaskTracker: task_0005_r_000000_0 0.33333334% reduce > copy (9000 of 9000 at 0.00 MB/s)
2007-02-10 18:25:46,573 INFO org.apache.hadoop.mapred.TaskTracker: task_0005_r_000000_0 0.33333334% reduce > copy (9000 of 9000 at 0.00 MB/s)
2007-02-10 18:25:48,791 INFO org.apache.hadoop.mapred.TaskTracker: task_0005_r_000000_0 0.33333334% reduce > copy (9000 of 9000 at 0.00 MB/s)
2007-02-10 18:25:49,828 WARN org.apache.hadoop.mapred.TaskRunner: Merge of the inmemory files threw an exception: java.lang.OutOfMemoryError: Java heap space
...

============
RandomWriter
============
...
2007-02-11 03:58:00,887 INFO org.apache.hadoop.mapred.TaskRunner: task_0001_r_000000_3 Copying of all map outputs complete. Initiating the last merge on the remaining files in ramfs://mapoutput6576294
2007-02-11 03:58:01,681 INFO org.apache.hadoop.mapred.TaskTracker: task_0001_r_000000_3 0.33333334% reduce > copy (8890 of 8890 at 0.00 MB/s)
2007-02-11 03:58:02,921 INFO org.apache.hadoop.mapred.TaskTracker: task_0001_r_000000_3 0.33333334% reduce > copy (8890 of 8890 at 0.00 MB/s)
2007-02-11 03:58:03,923 INFO org.apache.hadoop.mapred.TaskTracker: task_0001_r_000000_3 0.33333334% reduce > copy (8890 of 8890 at 0.00 MB/s)
2007-02-11 03:58:05,375 INFO org.apache.hadoop.mapred.TaskTracker: task_0001_r_000000_3 0.33333334% reduce > copy (8890 of 8890 at 0.00 MB/s)
2007-02-11 03:58:06,742 INFO org.apache.hadoop.mapred.TaskTracker: task_0001_r_000000_3 0.33333334% reduce > copy (8890 of 8890 at 0.00 MB/s)
2007-02-11 03:58:08,818 INFO org.apache.hadoop.mapred.TaskTracker: task_0001_r_000000_3 0.33333334% reduce > copy (8890 of 8890 at 0.00 MB/s)
2007-02-11 03:58:09,821 INFO org.apache.hadoop.mapred.TaskTracker: task_0001_r_000000_3 0.33333334% reduce > copy (8890 of 8890 at 0.00 MB/s)
2007-02-11 03:58:11,406 INFO org.apache.hadoop.mapred.TaskTracker: task_0001_r_000000_3 0.33333334% reduce > copy (8890 of 8890 at 0.00 MB/s)
2007-02-11 03:58:13,277 INFO org.apache.hadoop.mapred.TaskTracker: task_0001_r_000000_3 0.33333334% reduce > copy (8890 of 8890 at 0.00 MB/s)
2007-02-11 03:58:14,280 INFO org.apache.hadoop.mapred.TaskTracker: task_0001_r_000000_3 0.33333334% reduce > copy (8890 of 8890 at 0.00 MB/s)
2007-02-11 03:58:15,282 INFO org.apache.hadoop.mapred.TaskTracker: task_0001_r_000000_3 0.33333334% reduce > copy (8890 of 8890 at 0.00 MB/s)
2007-02-11 03:58:16,284 INFO org.apache.hadoop.mapred.TaskTracker: task_0001_r_000000_3 0.33333334% reduce > copy (8890 of 8890 at 0.00 MB/s)
2007-02-11 03:58:18,401 INFO org.apache.hadoop.mapred.TaskTracker: task_0001_r_000000_3 0.33333334% reduce > copy (8890 of 8890 at 0.00 MB/s)
2007-02-11 03:58:19,403 INFO org.apache.hadoop.mapred.TaskTracker: task_0001_r_000000_3 0.33333334% reduce > copy (8890 of 8890 at 0.00 MB/s)
2007-02-11 03:58:20,636 INFO org.apache.hadoop.mapred.TaskTracker: task_0001_r_000000_3 0.33333334% reduce > copy (8890 of 8890 at 0.00 MB/s)
2007-02-11 03:58:37,860 INFO org.apache.hadoop.mapred.TaskTracker: task_0001_r_000000_3 0.33333334% reduce > copy (8890 of 8890 at 0.00 MB/s)
2007-02-11 03:58:37,898 WARN org.apache.hadoop.mapred.TaskRunner: task_0001_r_000000_3 Child Error
java.lang.OutOfMemoryError: Java heap space
...

Issue Links

• is part of: HADOOP-1027 Fix the RAM FileSystem/Merge problems (reported in HADOOP-1014)