Hi!
Tried 5 seconds. Less number of nodes get into storm, but still they do.
Additionaly update of ntp service helped a little.
Initially almost 50% got into storming each MR job. But after ntp update
and and increasing heart-beatto 5 seconds level is around 10%.
On 26/05/13 10:43, murali adireddy wrote:
Hi ,
Just try this one.
in the file "hdfs-site.xml" try to add the below property
"dfs.heartbeat.interval" and value in seconds.
Default value is '3' seconds. In your case increase value.
<property>
<name>dfs.heartbeat.interval</name>
<value>3</value>
</property>
You can find more properties and default values in the below link.
http://hadoop.apache.org/docs/current/hadoop-project-dist/hadoop-hdfs/hdfs-default.xml
Please let me know is the above solution worked for you ..?
On Fri, May 24, 2013 at 6:40 PM, Eremikhin Alexey
<[email protected] <mailto:[email protected]>> wrote:
Hi all,
I have 29 servers hadoop cluster in almost default configuration.
After installing Hadoop 1.0.4 I've noticed that JT and some TT
waste CPU.
I started stracing its behaviour and found that some TT send
heartbeats in an unlimited ways.
It means hundreds in a second.
Daemon restart solves the issue, but even easiest Hive MR returns
issue back.
Here is the filtered strace of heartbeating process
hadoop9.mlan:~$ sudo strace -tt -f -s 10000 -p 6032 2>&1 | grep
6065 | grep write
[pid 6065] 13:07:34.801106 write(70,
"\0\0\1\30\0:\316N\0\theartbeat\0\0\0\5\0*org.apache.hadoop.mapred.TaskTrackerStatus\0*org.apache.hadoop.mapred.TaskTrackerStatus.tracker_hadoop9.mlan:localhost/127.0.0.1:52355
<http://127.0.0.1:52355>\fhadoop9.mlan\0\0\303\214\0\0\0\0\0\0\0\2\0\0\0\2\213\1\367\373\200\0\214\367\223\220\0\213\1\341p\220\0\214\341\351\200\0\377\377\213\6\243\253\200\0\214q\r\33\300\215$\205\266\4B\16\333n\0\0\0\0\1\0\0\0\0\0\0\0\0\0\0\7boolean\0\0\7boolean\0\0\7boolean\1\0\5short\316\30",
284) = 284
[pid 6065] 13:07:34.807968 write(70,
"\0\0\1\30\0:\316O\0\theartbeat\0\0\0\5\0*org.apache.hadoop.mapred.TaskTrackerStatus\0*org.apache.hadoop.mapred.TaskTrackerStatus.tracker_hadoop9.mlan:localhost/127.0.0.1:52355
<http://127.0.0.1:52355>\fhadoop9.mlan\0\0\303\214\0\0\0\0\0\0\0\2\0\0\0\2\213\1\367\373\200\0\214\367\223\220\0\213\1\341p\220\0\214\341\351\200\0\377\377\213\6\243\253\200\0\214q\r\33\312\215$\205\266\4B\16\333n\0\0\0\0\1\0\0\0\0\0\0\0\0\0\0\7boolean\0\0\7boolean\0\0\7boolean\1\0\5short\316\31",
284 <unfinished ...>
[pid 6065] 13:07:34.808080 <... write resumed> ) = 284
[pid 6065] 13:07:34.814473 write(70,
"\0\0\1\30\0:\316P\0\theartbeat\0\0\0\5\0*org.apache.hadoop.mapred.TaskTrackerStatus\0*org.apache.hadoop.mapred.TaskTrackerStatus.tracker_hadoop9.mlan:localhost/127.0.0.1:52355
<http://127.0.0.1:52355>\fhadoop9.mlan\0\0\303\214\0\0\0\0\0\0\0\2\0\0\0\2\213\1\367\373\200\0\214\367\223\220\0\213\1\341p\220\0\214\341\351\200\0\377\377\213\6\243\253\200\0\214q\r\33\336\215$\205\266\4B\16\333n\0\0\0\0\1\0\0\0\0\0\0\0\0\0\0\7boolean\0\0\7boolean\0\0\7boolean\1\0\5short\316\32",
284 <unfinished ...>
[pid 6065] 13:07:34.814595 <... write resumed> ) = 284
[pid 6065] 13:07:34.820960 write(70,
"\0\0\1\30\0:\316Q\0\theartbeat\0\0\0\5\0*org.apache.hadoop.mapred.TaskTrackerStatus\0*org.apache.hadoop.mapred.TaskTrackerStatus.tracker_hadoop9.mlan:localhost/127.0.0.1:52355
<http://127.0.0.1:52355>\fhadoop9.mlan\0\0\303\214\0\0\0\0\0\0\0\2\0\0\0\2\213\1\367\373\200\0\214\367\223\220\0\213\1\341p\220\0\214\341\351\200\0\377\377\213\6\243\253\200\0\214q\r\33\336\215$\205\266\4B\16\333n\0\0\0\0\1\0\0\0\0\0\0\0\0\0\0\7boolean\0\0\7boolean\0\0\7boolean\1\0\5short\316\33",
284 <unfinished ...>
Please help me to stop this storming 8(
