Error: out of memory

[Erling Paulsen]

Hi all.

I have a problem of periodic "crashes" with my server. It can be 
anything from 1-3 weeks inbetween. Currently I'm running 1.0.5, but I 
also had the same crashes with 1.0.4. Right before it dies, these 
messages are recorded to the log:

<logsnip>
Wed Oct 26 19:23:04 2005 : Error: out of memory
Wed Oct 26 19:23:05 2005 : Error: no memory
</logsnip>

The general system management graphs/tools-history show no excessive or 
heightened memory usage on the machine before or during the time of 
these events. The server has 1GB of RAM. I would guess that installing 
more physical memory on the server would not fix this problem!

Generally about the server:
The server runs in chroot environment, is threaded, and mostly functions 
as a proxy server to remote ad/ldaps and other radiuses. It also 
authenticates some VPN/dial-in users towards local /etc/passwd. The main 
load is however proxying of about 40-60 PEAP requests a minute. It sends 
full accounting to a secondary server that pushes ACCT-messages to a 
Postgresql server.

Anyone has some ideas or clues to why this happens? Are there some hints 
on memory tuning for my server. I.e. what size should my thread-pools be 
etc for the load described above?

- Erling Paulsen