IPv6 routes too much

IPv6 routes too much

[Prashant Batra]

Hi,

I have a case wherein I have a route to a network connected to my
linux machine via an interface, which I added like-

route add -inet6 9aa::/16 gw 2001:db8:0:242::2 dev eth0

Now, when I send packets to ipv6 addresses in this network, from my
linux server, I see kernel adding routes for each of these addresses.
Is this behaviour correct. How can it be avoided?

My problem starts when I have millions of addresses in this network
that I am trying to reach.

A look of my routing table shows-
9aa:0:0:250::/128                           2001:db8:0:242::2
             UGC   0      23       0 eth0
9aa:0:0:251::/128                           2001:db8:0:242::2
             UGC   0      23       0 eth0
9aa:0:0:252::/128                           2001:db8:0:242::2
             UGC   0      23       0 eth0
9aa:0:0:253::/128                           2001:db8:0:242::2
             UGC   0      23       0 eth0
9aa:0:0:254::/128                           2001:db8:0:242::2
             UGC   0      23       0 eth0
9aa:0:0:255::/128                           2001:db8:0:242::2
             UGC   0      23       0 eth0
9aa:0:0:256::/128                           2001:db8:0:242::2
             UGC   0      23       0 eth0
9aa:0:0:257::/128                           2001:db8:0:242::2
             UGC   0      23       0 eth0
9aa:0:0:258::/128                           2001:db8:0:242::2
             UGC   0      23       0 eth0
9aa:0:0:259::/128                           2001:db8:0:242::2
             UGC   0      23       0 eth0
9aa:0:0:25a::/128                           2001:db8:0:242::2
             UGC   0      23       0 eth0
9aa:0:0:25b::/128                           2001:db8:0:242::2
             UGC   0      23       0 eth0
9aa:0:0:25c::/128                           2001:db8:0:242::2
             UGC   0      23       0 eth0

like this I get 10k routes
#cat /proc/net/ipv6_route | wc -l
10015


Any help is highly appreciated.

thanks,
Prashant
--
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Re: IPv6 routes too much

[Jan Engelhardt]

On Friday 2013-02-08 11:48, Prashant Batra wrote:

>I have a case wherein I have a route to a network connected to my
>linux machine via an interface, which I added like-
>
>route add -inet6 9aa::/16 gw 2001:db8:0:242::2 dev eth0

Forget net-tools already. It is not maintained.
There is no issue with iproute.

>like this I get 10k routes
>#cat /proc/net/ipv6_route | wc -l
>10015

It is correct - it dumps routes (for some definition of "route"), in 
particular all FIB entries. You will find the same for IPv4 in 
/proc/net/rt_cache on kernels where the cache has not already been 
removed. That said /proc/net only shows table 254, so the whole procfs 
business is pretty useless.

Just use iproute2, really.
--
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Re: IPv6 routes too much

[Prashant Batra]

On Fri, Feb 8, 2013 at 9:04 PM, Jan Engelhardt <jengelh@inai.de> wrote:
> On Friday 2013-02-08 11:48, Prashant Batra wrote:
>
>>I have a case wherein I have a route to a network connected to my
>>linux machine via an interface, which I added like-
>>
>>route add -inet6 9aa::/16 gw 2001:db8:0:242::2 dev eth0
>
> Forget net-tools already. It is not maintained.
> There is no issue with iproute.
>
>>like this I get 10k routes
>>#cat /proc/net/ipv6_route | wc -l
>>10015
>
> It is correct - it dumps routes (for some definition of "route"), in
> particular all FIB entries. You will find the same for IPv4 in
> /proc/net/rt_cache on kernels where the cache has not already been
> removed. That said /proc/net only shows table 254, so the whole procfs
> business is pretty useless.
>
So, if I understand correctly, the /proc/net/rt_cache and
/proc/net/ipv6_route display the routing cache for v4 and v6
respectively, which the kernel maintains.
And, in case my server is trying to send packets to millions of
addresses, kernel will cache the routes for them, until
"/proc/sys/net/ipv6/route/max_size" is reached after which it will
flush some old entries to make room for new ones.
So, any advice onto how the kernel configs specially under
/proc/sys/net/ipv6/route/* should be controlled to get best
performance.
> Just use iproute2, really.



--
Thanks
--
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Re: IPv6 routes too much

[Jan Engelhardt]

On Friday 2013-02-08 18:33, Prashant Batra wrote:
>
>So, any advice onto how the kernel configs specially under
>/proc/sys/net/ipv6/route/* should be controlled to get best performance.

The defaults should be good enough.

If they are not: Measure it.