[ILUG] anyone have any experience of java thread problems on linux ?

[Hunt, Bryan]

Actually guys I had ZERO as in 0 ( as in 1 less than 1 ) input into the
design 
of the product so I couldn't give a toss as to whether it is bad design or
not. 
I just get a request from my boss make this scale up past the 1018 thread
limit 
or else we ditch Linux and go with Windows or some other crap. I prefer to
use 
Linux (for various reasons) so I try to find a solution ... So thats there I
am 
coming from ... you dig ? 

As far as I can tell the correct answer to the question was 

cut=
The Solution Make the following changes:

In /usr/include/linux/limits.h:
 #define NR_OPEN 1024
 should be
 #define NR_OPEN 4096
 In /usr/include/linux/posix_types.h:
 #define __FD_SETSIZE 1024
 should be
 #define __FD_SETSIZE 4096
 In /usr/include/bits/types.h:
 #define __FD_SETSIZE 1024
 should be
 #define __FD_SET_SIZE 4096
=cut

Then go and recompile glibc .... hmmm looks like a whole load of fun 

--B

-----Original Message-----
From: Padraig Brady [mailto:padraig at antefacto.com]
Sent: Wednesday, May 08, 2002 4:08 PM
To: Nick Murtagh
Cc: Hunt, Bryan; Ilug (E-mail)
Subject: Re: [ILUG] anyone have any experience of java thread problems
on linux ?


Nick Murtagh wrote:
> On Wednesday 08 May 2002 15:50, Padraig Brady wrote:
> 
>>Using the same design logic, you could work around the problem
>>by adding more memory and using multiple copies of
>>linux (UML) talking to each other each running a JVM,
>>each handling one connection. Or perhaps run the JVM
>>on doze on VMware on Linux?
> 
> 
> Actually, the UML guy was talking about embedding UML in Apache,
> as a generic replacement for mod_perl etc. So you might not
> be far wrong :)

Ewww...
Actually UML has become self hosting in the last couple of days,
so you can have UML within UML ... ad infinitum.

Padraig.