----------------------------------------------------------------
BEFORE YOU POST, search the faq at <http://java.apache.org/faq/>
WHEN YOU POST, include all relevant version numbers, log files,
and configuration files. Don't make us guess your problem!!!
----------------------------------------------------------------
One other thing we needed to change:
edit the HARD_SERVER_LIMIT in Apache's httpd.h to be
2048 instead of 256 and recompile.
--- Nick Bauman <[EMAIL PROTECTED]> wrote:
>
> The problem we found was that between our
> application,
> the web process gang, the JVM, JServ and the OS...
> we
> were running out of file descriptors and their
> associated resources. The solution was to recomplie
> the Linux kernel and here are the step-by-step
> instructions for what we did:
>
> Change the following C header files.
> In /usr/include/linux/tasks.h, change:
>
> NR_TASKS 512 -> 4090
> MAX_TASKS_PER_USER (NR_TASKS/2) -> NR_TASKS
>
> In /usr/include/linux/limits.h, change:
>
> NR_OPEN 1024 -> 4096
> OPEN_MAX 256 -> 4096
>
> In /usr/include/linux/posix_types.h, change:
>
> __FD_SETSIZE 1024 -> 4096
>
> In /usr/include/bits/types.h, change:
>
> __FD_SETSIZE 1024 -> 4096
>
> To allow users to increase their file descriptor
> limits, change the following configuration files:
> In /etc/security/limits.conf, add the lines:
>
> * soft nofile 1024
> * hard nofile 4096
>
> In /etc/pam.d/login, add:
>
> session required /lib/security/pam_limits.so
>
> To increase the system-wide file descriptor limit,
> add
> the following three lines to the /etc/rc.d/rc.local
> startup script:
> # Increase system-wide file descriptor limit.
> echo 8192 > /proc/sys/fs/file-max
> echo 24576 > /proc/sys/fs/inode-max
>
> Change to the directory /usr/src/linux and set up
> the
> kernel sources with:
> make mrproper
>
> Make any other kernel configuration changes, such as
> setting the processor family type, by entering the
> command below. Save your changes and exit the
> configuration panel when you are done.
> make xconfig
>
> Prepare for the kernel build by entering:
> make dep
> make clean
>
> Rebuild your kernel with the commands:
> make bzImage
> make modules
>
> Backup your old modules directory by entering (for
> example):
> mv /lib/modules/2.2.5-15 /lib/modules/2.2.5-15.old
>
> Create a new set of kernel modules with the command:
>
> make modules_install
>
> Backup your old kernel by entering (for example):
> mv /boot/vmlinuz /boot/vmlinuz.old
>
> Copy over your new kernel image with:
> cp /usr/src/linux/arch/i386/boot/bzImage
> /boot/vmlinuz
>
> Edit the Linux loader configuration in
> /etc/lilo.conf
> so that your new kernel is the default and your old
> kernel is still available from the boot prompt, as
> in
> the example shown below:
>
> boot=/dev/hda
> map=/boot/map
> install=/boot/boot.b
> prompt
> timeout=50
> image=/boot/vmlinuz
> label=linux
> root=/dev/hda1
> read-only
> image=/boot/vmlinuz.old
> label=linux.old
> root=/dev/hda1
> read-only
>
> Generate the new Linux loader configuration by
> running:
> /sbin/lilo
>
> Shutdown and reboot with your new kernel by
> entering:
> shutdown -r now
>
> After rebooting, verify that the changes were
> effective.
>
> --- Nick Bauman <[EMAIL PROTECTED]> wrote:
> >
>
----------------------------------------------------------------
> > BEFORE YOU POST, search the faq at
> > <http://java.apache.org/faq/>
> > WHEN YOU POST, include all relevant version
> numbers,
> > log files,
> > and configuration files. Don't make us guess your
> > problem!!!
> >
>
----------------------------------------------------------------
> >
> > We have a very high traffic website that is
> > increasing
> > in it's traffic. Yesterday we had 41,000 unique
> > visitors for over 2.9 million "hits". That number
> is
> > expceted to increase over the coming weeks.
> >
> > The problem is we are seeing as many as 500-750 of
> > WAIT status in netstat -a, so naturally we wanted
> to
> > up the max children to 500-750 to accommodate this
> > observation. When we do this, we see a marked
> > increase
> > in the number of concurrent requests we can handle
> > and
> > much better performance of the app, but as we
> break
> > the 200 concurrent connections mark, JServ just
> > gives
> > up with the java process gang just disappearing
> from
> > the process table, leaving no core and nothing in
> > the
> > logs. Of course we are running using Native
> threads.
> > We are limping along right now under the load
> trying
> > to load balance servers right now with a Local
> > Director. We have tried:
> >
> > 1) increasing the max stack and max heap size to
> > 2/256MB
> > 2) switched VMs and OSs
> > 3) turned off JIT
> >
> > ENVIRONMENT:
> >
> > Software: Apache 1.3.9 with JServ 2.0 as a DSO
> >
> > Operating System: tried on Solaris 2.7 and Linux
> > RH6.0
> >
> > JVM: Sun's JDK 1.1.8 and IBM's JDK 1.1.8 for Linux
> >
> > Hardware: Sun E350 with 512k RAM and Compaq
> Proliant
> > PIII dual 500 with 512k RAM
> >
> >
> >
> > =====
> > # Nick Bauman
> > # Technical Programmer
> > # http://webhelp.com
> > # real people, real answers, real time
> > __________________________________________________
> > Do You Yahoo!?
> > Thousands of Stores. Millions of Products. All
> in
> > one place.
> > Yahoo! Shopping: http://shopping.yahoo.com
> >
> >
> > --
> >
>
--------------------------------------------------------------
>
=== message truncated ===
=====
# Nick Bauman
# Technical Programmer
# http://webhelp.com
# real people, real answers, real time
__________________________________________________
Do You Yahoo!?
Thousands of Stores. Millions of Products. All in one place.
Yahoo! Shopping: http://shopping.yahoo.com
--
--------------------------------------------------------------
Please read the FAQ! <http://java.apache.org/faq/>
To subscribe: [EMAIL PROTECTED]
To unsubscribe: [EMAIL PROTECTED]
Archives and Other: <http://java.apache.org/main/mail.html>
Problems?: [EMAIL PROTECTED]
