Forest Bond wrote:
> A customer is seeing "file system full" messages in his logs:
> Jan 15 12:21:56 monogw-rmueznet /kernel: pid 104 (dhcpd), uid 0 on /: file system full
> Jan 15 12:25:15 monogw-rmueznet /kernel: pid 104 (dhcpd), uid 0 on /: file system full
> Jan 15 12:26:36 monogw-rmueznet /kernel: pid 104 (dhcpd), uid 0 on /: file system full
> At this point the m0n0wall box is unable to provide more DHCP leases.
> If I understand things correctly, the file system in this case is in memory.
> Would adding more memory to the machine solve this problem? It currently has
> He says:
> "What I would LIKE to do is accommodate 1000 active simultaneous connections with
> a default lease time of 20 minutes."
> Are these expectations reasonable?
Yes, the file system is in memory so adding memory should solve that
issue. However, look at the system status page and see if it is
actually using all memory to verify.
As far as his question is concerned I'd say it all depends on the
hardware. I feel 1000 users should be behind a very robust router, and
DHCP should be handled by a separate server in a network that big.
Maybe even 2 routers using carp in case 1 fails. 1000 people is a lot
of people to disappoint because you're cheap. Unless of course it's
free service. In that case let em suffer. :-)