Hi!
First of all my special thanks to Nick Hill and Arrigo Triulzi.
The problem is that the default value for tha maximum virtual memory
space that can be allocated by one process is 1 Gbyte. So you should
change this value to lets say 2 Gbyte in /etc/sysconfigtab. However,
you should carefully follow the man pages and use sysconfigdb, NOT
simply edit /etc/sysconfigtab ;-).
So I created a stanza for the virtual memory subsystem (vm) (called
vm.stanza):
vm:
new-wire-method = 0
vm-maxvas = 2147483648
vm-ubcseqpercent = 2
vm-ubcseqstartpercent = 20
vm-vpagemax = 65536
vm-mapentries = 16384
The crucial entry is vm-maxvas, the maximum address space. The other
values are recommendations I got when running sys_check.
I updated /etc/sysconfigtab by running
#sysconfigdb -f vm.stanza -u vm
I further added some entries for the proc subsystem (by editing
/etc/sysconfigtab, which is not the propper way, but it worked ;-))
proc:
max-per-proc-address-space = 2147483648
per-proc-address-space = 2147483648
max-per-proc-data-size = 2147483648
max-per-proc-stack-size = 268435456
since the user (running a tcsh) is not allowed to unlimit the
vmemryuse (as pointed aout by Nick Hill).
Of course, I then had to reboot the systems to make the changes
valid for the kernel (you do naot actually have to rebuild it!)
/Andre
Received on Thu Sep 10 1998 - 09:35:13 NZST