[SOLVED] IPv6 works, then stops working, then works again after restarting networking service

woowoo · 05-11-2015, 09:45 PM

Hi there, and first of all thanks for reading!

So I saw some problems on my server as I am using IPv6 for some services such as DNS transfer for instance. I thus want to keep IPv6 active.

When I reboot my server, IPv6 works fine:

Code:

root@x:~# ping6 ipv6.google.com
PING ipv6.google.com(par10s22-in-x0e.1e100.net) 56 data bytes
64 bytes from par10s22-in-x0e.1e100.net: icmp_seq=1 ttl=57 time=4.61 ms
64 bytes from par10s22-in-x0e.1e100.net: icmp_seq=2 ttl=57 time=4.67 ms
64 bytes from par10s22-in-x0e.1e100.net: icmp_seq=3 ttl=57 time=4.81 ms
64 bytes from par10s22-in-x0e.1e100.net: icmp_seq=4 ttl=57 time=4.64 ms
64 bytes from par10s22-in-x0e.1e100.net: icmp_seq=5 ttl=57 time=4.56 ms
64 bytes from par10s22-in-x0e.1e100.net: icmp_seq=6 ttl=57 time=4.68 ms

But after a few minutes/hours:

Code:

root@x:~# ping6 ipv6.google.com
connect: Network is unreachable

But if I restart the networking service:

Code:

root@x:~# service networking restart
[warn] Running /etc/init.d/networking restart is deprecated because it may not re-enable some interfaces ... (warning).
[....] Reconfiguring network interfaces...SIOCDIFADDR: Cannot assign requested address
done.
root@x:~# ping6 ipv6.google.com
PING ipv6.google.com(par10s22-in-x0e.1e100.net) 56 data bytes
64 bytes from par10s22-in-x0e.1e100.net: icmp_seq=1 ttl=57 time=4.58 ms
64 bytes from par10s22-in-x0e.1e100.net: icmp_seq=2 ttl=57 time=4.59 ms
64 bytes from par10s22-in-x0e.1e100.net: icmp_seq=3 ttl=57 time=4.71 ms
64 bytes from par10s22-in-x0e.1e100.net: icmp_seq=4 ttl=57 time=4.62 ms

I am using Debian Wheezy:

Code:

root@x:~# cat /etc/debian_version
7.8

And my IPv6 configuration is done statically in /etc/network/interfaces:

Code:

iface lo inet6 loopback

iface eth0 inet6 static
	address 2001:x:x:x::42/64
	netmask 64

Do you have any idea why it stops working every so often?
Is there anything I can do to correct that behavior?

Thanks!

woowoo · 05-11-2015, 11:15 PM

I keep here updated of my research in case someone has an idea reading from my news.

So I looked up that:

Code:

root@x:~# ip -6 route

And I saw that the result is different whether I can ping an IPv6 IP or not.
Working case:

Code:

root@x:~# ip -6 route
2001:xxxx:xxxx:4b3a::/64 dev eth0  proto kernel  metric 256 
2001:xxxx:xxxx:4b00::/56 dev eth0  proto kernel  metric 256  expires 2591970sec
fe80::/64 dev eth0  proto kernel  metric 256 
default via fe80::xxxx:xxxx:xxxx:f080 dev eth0  proto kernel  metric 1024  expires 1608sec
default via fe80::xxxx:xxxx:xxxx:f100 dev eth0  proto kernel  metric 1024  expires 1608sec

Not working case:

Code:

root@x:~# ip -6 route
2001:xxxx:xxxx:4b3a::/64 dev eth0  proto kernel  metric 256 
2001:xxxx:xxxx:4b00::/56 dev eth0  proto kernel  metric 256  expires 2591970sec
fe80::/64 dev eth0  proto kernel  metric 256

So, it seems my two default gateway disappear.

I used a bash script to loop every few seconds and ping6 -c1 ipv6.google.com while the default route were still in the result of that command:

Code:

mROUTE="$(ip -6 route)"; while [ -n "$(echo "$mROUTE" | grep "default")" ]; do echo "$mROUTE"; ping6 -c1 ipv6.google.com; sleep 2; mROUTE="$(ip -6 route)"; done; echo "NOPE"; echo "NOPE"; echo "NOPE"; ping6 -c1 ipv6.google.com

And it seemed to loop even after the expiration, because as soon as it expired, I got these lines instead in the ip -6 route:

Code:

default via fe80::xxxx:xxxx:xxxx:f100 dev eth0  proto kernel  metric 1024  expires 8590471sec
default via fe80::xxxx:xxxx:xxxx:f080 dev eth0  proto kernel  metric 1024  expires 8590471sec

So I stopped my loop, and used ping6 ipv6.google.com manually: I got the error. I looked back at ip -6 route, and the lines were not here anymore.
I tried editing my /etc/network/interfaces to add the "gateway" line, then did service networking restart. I have the same default via appearing in the ip -6 route, with the same expiration. I'm now waiting for the delay to expire to see if ipv6 will be inaccessible again...

woowoo · 05-11-2015, 11:53 PM

Ok, problem solved!
So it was really a gateway problem. Adding the gateway as static work now like a charm!