cloudera hadoop mapreduce作业GC开销限制超出错误

我正在cloudera cdh4上运行一个canopy集群作业（使用mahout）。要集群的内容大约有1m条记录（每条记录的大小小于1k）。整个hadoop环境（包括所有节点）都运行在一个具有4G内存的vm中。默认情况下安装cdh4。我在运行作业时遇到以下异常

根据异常情况，作业客户端似乎需要更高的jvm堆大小。然而，cloudera manager中有很多jvm堆大小的配置选项。我将“客户端Java堆大小（字节）”从256MiB更改为512MiB。然而，情况并没有改善

有关于设置这些堆大小选项的提示/提示吗

13/07/03 17:12:45 INFO input.FileInputFormat: Total input paths to process : 1
13/07/03 17:12:46 INFO mapred.JobClient: Running job: job_201307031710_0001
13/07/03 17:12:47 INFO mapred.JobClient:  map 0% reduce 0%
13/07/03 17:13:06 INFO mapred.JobClient:  map 1% reduce 0%
13/07/03 17:13:27 INFO mapred.JobClient:  map 2% reduce 0%
13/07/03 17:14:01 INFO mapred.JobClient:  map 3% reduce 0%
13/07/03 17:14:50 INFO mapred.JobClient:  map 4% reduce 0%
13/07/03 17:15:50 INFO mapred.JobClient:  map 5% reduce 0%
13/07/03 17:17:06 INFO mapred.JobClient:  map 6% reduce 0%
13/07/03 17:18:44 INFO mapred.JobClient:  map 7% reduce 0%
13/07/03 17:20:24 INFO mapred.JobClient:  map 8% reduce 0%
13/07/03 17:22:20 INFO mapred.JobClient:  map 9% reduce 0%
13/07/03 17:25:00 INFO mapred.JobClient:  map 10% reduce 0%
13/07/03 17:28:08 INFO mapred.JobClient:  map 11% reduce 0%
13/07/03 17:31:46 INFO mapred.JobClient:  map 12% reduce 0%
13/07/03 17:35:57 INFO mapred.JobClient:  map 13% reduce 0%
13/07/03 17:40:52 INFO mapred.JobClient:  map 14% reduce 0%
13/07/03 17:46:55 INFO mapred.JobClient:  map 15% reduce 0%
13/07/03 17:55:02 INFO mapred.JobClient:  map 16% reduce 0%
13/07/03 18:08:42 INFO mapred.JobClient:  map 17% reduce 0%
13/07/03 18:59:11 INFO mapred.JobClient:  map 8% reduce 0%
13/07/03 18:59:13 INFO mapred.JobClient: Task Id : attempt_201307031710_0001_m_000001_0, Status : FAILED
Error: GC overhead limit exceeded
13/07/03 18:59:23 INFO mapred.JobClient:  map 9% reduce 0%
13/07/03 19:00:09 INFO mapred.JobClient:  map 10% reduce 0%
13/07/03 19:01:49 INFO mapred.JobClient:  map 11% reduce 0%
13/07/03 19:04:25 INFO mapred.JobClient:  map 12% reduce 0%
13/07/03 19:07:48 INFO mapred.JobClient:  map 13% reduce 0%
13/07/03 19:12:48 INFO mapred.JobClient:  map 14% reduce 0%
13/07/03 19:19:46 INFO mapred.JobClient:  map 15% reduce 0%
13/07/03 19:29:05 INFO mapred.JobClient:  map 16% reduce 0%
13/07/03 19:43:43 INFO mapred.JobClient:  map 17% reduce 0%
13/07/03 20:49:36 INFO mapred.JobClient:  map 8% reduce 0%
13/07/03 20:49:38 INFO mapred.JobClient: Task Id : attempt_201307031710_0001_m_000001_1, Status : FAILED
Error: GC overhead limit exceeded
13/07/03 20:49:48 INFO mapred.JobClient:  map 9% reduce 0%
13/07/03 20:50:31 INFO mapred.JobClient:  map 10% reduce 0%
13/07/03 20:52:08 INFO mapred.JobClient:  map 11% reduce 0%
13/07/03 20:54:38 INFO mapred.JobClient:  map 12% reduce 0%
13/07/03 20:58:01 INFO mapred.JobClient:  map 13% reduce 0%
13/07/03 21:03:01 INFO mapred.JobClient:  map 14% reduce 0%
13/07/03 21:10:10 INFO mapred.JobClient:  map 15% reduce 0%
13/07/03 21:19:54 INFO mapred.JobClient:  map 16% reduce 0%
13/07/03 21:31:35 INFO mapred.JobClient:  map 8% reduce 0%
13/07/03 21:31:37 INFO mapred.JobClient: Task Id : attempt_201307031710_0001_m_000000_0, Status : FAILED
java.lang.Throwable: Child Error
    at org.apache.hadoop.mapred.TaskRunner.run(TaskRunner.java:250)
Caused by: java.io.IOException: Task process exit with nonzero status of 65.
    at org.apache.hadoop.mapred.TaskRunner.run(TaskRunner.java:237)

13/07/03 21:32:09 INFO mapred.JobClient:  map 9% reduce 0%
13/07/03 21:33:31 INFO mapred.JobClient:  map 10% reduce 0%
13/07/03 21:35:42 INFO mapred.JobClient:  map 11% reduce 0%
13/07/03 21:38:41 INFO mapred.JobClient:  map 12% reduce 0%
13/07/03 21:42:27 INFO mapred.JobClient:  map 13% reduce 0%
13/07/03 21:48:20 INFO mapred.JobClient:  map 14% reduce 0%
13/07/03 21:56:12 INFO mapred.JobClient:  map 15% reduce 0%
13/07/03 22:07:20 INFO mapred.JobClient:  map 16% reduce 0%
13/07/03 22:26:36 INFO mapred.JobClient:  map 17% reduce 0%
13/07/03 23:35:30 INFO mapred.JobClient:  map 8% reduce 0%
13/07/03 23:35:32 INFO mapred.JobClient: Task Id : attempt_201307031710_0001_m_000000_1, Status : FAILED
Error: GC overhead limit exceeded
13/07/03 23:35:42 INFO mapred.JobClient:  map 9% reduce 0%
13/07/03 23:36:16 INFO mapred.JobClient:  map 10% reduce 0%
13/07/03 23:38:01 INFO mapred.JobClient:  map 11% reduce 0%
13/07/03 23:40:47 INFO mapred.JobClient:  map 12% reduce 0%
13/07/03 23:44:44 INFO mapred.JobClient:  map 13% reduce 0%
13/07/03 23:50:42 INFO mapred.JobClient:  map 14% reduce 0%
13/07/03 23:58:58 INFO mapred.JobClient:  map 15% reduce 0%
13/07/04 00:10:22 INFO mapred.JobClient:  map 16% reduce 0%
13/07/04 00:21:38 INFO mapred.JobClient:  map 7% reduce 0%
13/07/04 00:21:40 INFO mapred.JobClient: Task Id : attempt_201307031710_0001_m_000001_2, Status : FAILED
java.lang.Throwable: Child Error
    at org.apache.hadoop.mapred.TaskRunner.run(TaskRunner.java:250)
Caused by: java.io.IOException: Task process exit with nonzero status of 65.
    at org.apache.hadoop.mapred.TaskRunner.run(TaskRunner.java:237)

13/07/04 00:21:50 INFO mapred.JobClient:  map 8% reduce 0%
13/07/04 00:22:27 INFO mapred.JobClient:  map 9% reduce 0%
13/07/04 00:23:52 INFO mapred.JobClient:  map 10% reduce 0%
13/07/04 00:26:00 INFO mapred.JobClient:  map 11% reduce 0%
13/07/04 00:28:47 INFO mapred.JobClient:  map 12% reduce 0%
13/07/04 00:32:17 INFO mapred.JobClient:  map 13% reduce 0%
13/07/04 00:37:34 INFO mapred.JobClient:  map 14% reduce 0%
13/07/04 00:44:30 INFO mapred.JobClient:  map 15% reduce 0%
13/07/04 00:54:28 INFO mapred.JobClient:  map 16% reduce 0%
13/07/04 01:16:30 INFO mapred.JobClient:  map 17% reduce 0%
13/07/04 01:32:05 INFO mapred.JobClient:  map 8% reduce 0%
13/07/04 01:32:08 INFO mapred.JobClient: Task Id : attempt_201307031710_0001_m_000000_2, Status : FAILED
Error: GC overhead limit exceeded
13/07/04 01:32:21 INFO mapred.JobClient:  map 9% reduce 0%
13/07/04 01:33:26 INFO mapred.JobClient:  map 10% reduce 0%
13/07/04 01:35:37 INFO mapred.JobClient:  map 11% reduce 0%
13/07/04 01:38:48 INFO mapred.JobClient:  map 12% reduce 0%
13/07/04 01:43:06 INFO mapred.JobClient:  map 13% reduce 0%
13/07/04 01:49:58 INFO mapred.JobClient:  map 14% reduce 0%
13/07/04 01:59:07 INFO mapred.JobClient:  map 15% reduce 0%
13/07/04 02:12:00 INFO mapred.JobClient:  map 16% reduce 0%
13/07/04 02:37:56 INFO mapred.JobClient:  map 17% reduce 0%
13/07/04 03:31:55 INFO mapred.JobClient:  map 8% reduce 0%
13/07/04 03:32:00 INFO mapred.JobClient: Job complete: job_201307031710_0001
13/07/04 03:32:00 INFO mapred.JobClient: Counters: 7
13/07/04 03:32:00 INFO mapred.JobClient:   Job Counters 
13/07/04 03:32:00 INFO mapred.JobClient:     Failed map tasks=1
13/07/04 03:32:00 INFO mapred.JobClient:     Launched map tasks=8
13/07/04 03:32:00 INFO mapred.JobClient:     Data-local map tasks=8
13/07/04 03:32:00 INFO mapred.JobClient:     Total time spent by all maps in occupied slots (ms)=11443502
13/07/04 03:32:00 INFO mapred.JobClient:     Total time spent by all reduces in occupied slots (ms)=0
13/07/04 03:32:00 INFO mapred.JobClient:     Total time spent by all maps waiting after reserving slots (ms)=0
13/07/04 03:32:00 INFO mapred.JobClient:     Total time spent by all reduces waiting after reserving slots (ms)=0
Exception in thread "main" java.lang.RuntimeException: java.lang.InterruptedException: Canopy Job failed processing vector

---

您需要更改Hadoop的内存设置，因为为Hadoop分配的内存不足以满足您正在运行的作业要求，请尝试增加堆内存并验证，因为内存过度使用，操作系统可能会因为作业失败而终止进程。

Mahout作业非常占用内存。我不知道映射程序还是还原程序是罪魁祸首，但是，无论如何，你都必须告诉Hadoop给他们更多的内存。“超出GC开销限制”只是“内存不足”的一种说法——意味着JVM放弃了回收最后0.01%的可用RAM的尝试

如何设置这一点确实有点复杂，因为有几个属性，它们在Hadoop 2中发生了更改。CDH4可以支持Hadoop 1或2——您使用哪一种

---

如果让我猜的话：将

mapreduce.child.java.opts

设置为

-Xmx1g

。但是正确的答案实际上取决于您的版本和数据。

您的应用程序需要使用大量内存吗？如果没有，可能是应用程序中有一些bug占用了整个内存。它正在运行mahout Corporation集群，因此不应该是应用程序bug。我可以看到每个子客户机都分配了大约200MB，这在我的情况下可能不够。@zsxwing您应该将其写为“-Xmx1024M”，原因正是：您在那里放了太多的零。那是10.24克