Linux IPv6 router table size.

[George Bonser]

> On Behalf Of Ted
> Mittelstaedt
> Sent: Saturday, January 15, 2011 9:41 PM
> To: ipv6-ops at lists.cluenet.de
> Subject: Re: Linux IPv6 router table size.
> 
> On 1/15/2011 4:09 PM, Richard Hartmann wrote:
> > On Sat, Jan 15, 2011 at 22:23, Ted Mittelstaedt<tedm at ipinc.net>
> wrote:
> >
> >>   Keep in mind that in both Linux and FreeBSD (FreeBSD having the
> same
> >> problem) that the formula that calculates the max route size isn't
> >> assuming the system is going to be used as a BGP router.
> >
> > Intended purpose of any given system notwithstanding, if I put 4097
> > routes into my FIB, I expect 4097 routes to end up in my FIB. How or
> 
> Then why are you letting quagga off the hook here?  Why allow the
> quagga program get away with failing to add the 4097th route
> when quagga could easily have used the sysctl to see that the
> max route size had been hit and then proceeded to adjust it to add
> the route?

Seems to me that quagga could adopt something similar to the Brocade config settings:

system-max ip-route
system-max ipv6-route

and set the sysctl appropriately

That said, it brings up another point to combined v4-v6 table bloat with dual stacked machines. It would appear that my strategy will be to reduce the netmask length of accepted v4 routes when my total routing table comes under pressure due to a combination of v6 table growth and v4 fragmentation.  I will probably go from accepting out to a /24 in v4 to a /23 as my longest prefix accepted in PA space but will probably accept a /24 in PI space.

Anyone else thinking about this yet?

G