On Mon, 07 Dec 2009 05:07:40 -0000, Todd Lipcon <[email protected]> wrote:
Thank you!
Interesting that previously i have launched hadoop perfectly in two
wariants:
1. MASTER (32bit cpu, 2GB ram, linux gentoo) + SLAVE (32bit cpu, 1GB ram,
linux gentoo).
2. MASTER (32bit cpu, 2GB ram, linux gentoo) + SLAVE (VirtualBox in the
MASTER's machine).
* Understand about swap / memory increase.
* Will play with "-Xmx" option.
13940 clone(child_stack=0,
flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD,
child_tidptr=0x7f51260459e0) = -1 ENOMEM (Cannot allocate memory)
(a few times)
Solutions: (pick one)
1) allocate more swap on your node
2) futz with /proc/sys/vm/overcommit_ratio (make it bigger) or
/proc/sys/vm/overcommit_memory (set to 1)
For option #2 you risk letting the linux OOM killer loose on your system,
which is usually no fun. So #1 is what I'd recommend.
-Todd
On Mon, Dec 7, 2009 at 1:00 AM, pavel kolodin
<[email protected]>wrote:
