- the JVM. Prior to 1.3, the Sun JVM's assumed a 2MB stack, imposing a hard limit of 1024 threads.
- your threads implementation. From the top of my head, prior to around 0.7, LinuxThreads also hard-coded a 2MB stack for threads
Note that whenever the JVM can't create a new thread, you'll get an OutOfMemoryError, regardless of whether that's due to a lack of memory, or permissions, or thread resources. (Again, my only experience is with Sun's JVMs).
On 6/13/2002 12:47 PM, Adam Majer wrote:
On Wed, Jun 12, 2002 at 06:44:47PM +0200, Philipp Meier wrote:
On Tue, Jun 11, 2002 at 10:07:35PM -0500, Adam Majer wrote:
On Thu, May 16, 2002 at 10:32:06AM -0700, A.J. Rossini wrote:
"tony" == tony mancill <[EMAIL PROTECTED]> writes:
tony> For both cases, would it possible to post the kernel version of the test tony> box? Maybe it's part of the equation.
(I actually did, privately, to Bill, forgot to cc the list).
I'm running 2.4.18, on a dual P2 1.0 Ghz box, 512Mb/2Gb swap.
turns out it might be a process limit problem.
Yes. By deafault the max number of threads is set to 320. Well, at least on my 486 box. On Athlon I have it as 8k.
/proc/sys/kernel/threads-max
If it says it can't fork, then it is threads-max that's the problem.
It even might be the shell, my standard debian zsh e.g. limits the number um threads to 256:
[EMAIL PROTECTED]:~/ > ulimit -u 256
It might help to set it higher:
[EMAIL PROTECTED]:~/ > ulimit -u 8192
ulimit is very useless. At least that's my opinion. The hadlimit for 2.2.x kernels is set in include/linux/tasks.h in the kernel source. The think is NR_TASKS which I think is set to 512. There is also 2 lines below that might be interresting ie. the ordinary user can only make NR_TASKS/2 no matter what umit is.
- Adam
-- Damian Morris
Director, Quality Assurance Gaming & Entertainment Technology
http://www.getsystems.com
Phone: (+61) 2 9419 2000 Mobile: (+61) 412 808 307
-- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
