> I hope this is the right place to ask this, if not just say so...
It just might be (see below).
> We're trying to get the server side of Ultima Online (see http://www.owo.com/
> for info) running under Linux (the client has run under Linux for quite some
> time now).
>
> Currently, one world for UO consists of around 20 processes running on 6 dual
> Sparc 300mhz processor Sun Ultra 2 boxes with 512MB RAM.
>
> We're *this* close to everything running on 6 Dell dual processor Pentium
> II/300mhz boxes with 512MB RAM (we could run in 256MB RAM under Linux because of
> copy-on-write pages, but went with 512MB to be safe).
>
> Some info on our systems: 2.1.89, Red Hat 5.0 with errata updates, and all other
> 2.1 required program updates except NFS stuff since it didn't compile under
> glibc. We are trying 2.1.89 because of better SMP reliability.
>
> - Page starvation?
>
> There is plenty of free memory available (real and virtual), but we get
> occasional crashes in glibc while in a new or malloc. I've read some old info
> on page starvation and changing some settings in /proc/sys/vm/freepages.
The system is paging out too much memory at once... Currently,
there is a bug in the VM code. A function (free_memory_available)
is being called to test for enough free memory).
A quick workaround for a 512MB machine might be to:
# echo "512 1024 8192" > /proc/sys/vm/freepages
(this way, so much memory will be left unallocated, that the
system will always pass the free_memory_available() test...
It's not a fix, but it just might do the trick.
Currently, we're rewriting the VM subsystem (mainly
Stephen's and Ben's work) so we don't have to do this
dirty hack anymore RSN...
And as for the docs: There's new documentation on
/proc/sys/*/* in linux/Documentation/sysctl/
succes,
Rik.
+-----------------------------+------------------------------+
| For Linux mm-patches, go to | "I'm busy managing memory.." |
| my homepage (via LinuxHQ). | H.H.vanRiel@fys.ruu.nl |
| ...submissions welcome... | http://www.fys.ruu.nl/~riel/ |
+-----------------------------+------------------------------+
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu