Hello.
I am running hadoop-0.20.1.
conf/hadoop-env.sh have default setting, excluding one "JAVA_HOME".
----------------------
Success with 2 such nodes:
1) laptop, pentium M760, 2GB RAM
2) VirtualBox running on this laptop with 350MB allowed "RAM"
(all - gentoo linux)
--------------------
Failed with 2 such nodes:
1) VPS, 64bit, 3.5GB RAM
2) VPS, 64bit, 2.5GB RAM
(all - gentoo linux, hardware - 1 server, 16 Xeons, 16GB)
Many errors, all because of deficit of memory:
* "DrWho" (whoami available, but can't start because of memory deficit)
* "java.io.IOException: Cannot run program "chmod": java.io.IOException:
error=12, Cannot allocate memory"
* and other...
-----------------------
So, it can allocate memory when running on laptop and can't allocate
memory running on VPS with 3.5 GB.
I have write small C++ program, that do "char *p = new char [ 1024*1024
];" many times (ok, ~3500 times, little less than 3.5GB), and i see that
program can allocate about ~3.5GB successfully and then deallocate it, so
memory system is OK )
Any ideas? :)
Thank you.
P.S.
I played with -Xmx in HADOOP_*_OPTS in hadoop-env.sh - not successful.
(strange effects described in this mailing list earlier, subject was:
"Memory. -Xmx. error=12, chmod, JobTracker")
