[Olsr-dev] Bug in redistribution of IPv6 olsrd router into quagga/BGP

[ZioPRoTo (Saverio Proto)]

Hello,

looks like we managed to fix the problem with the following steps:

reboot the Linux machine (we had 410 days uptime)

Adding another patch on top of Quagga 0.99.20.1:
https://github.com/zioproto/quagga-manet/commits/openwrt-0.99.20.1

Upgrade the Kernel:
from:
Linux tecnopolo 2.6.32-24-generic-pae #39-Ubuntu SMP Wed Jul 28
07:39:26 UTC 2010 i686 GNU/Linux
to:
Linux tecnopolo 2.6.32-41-generic-pae #89-Ubuntu SMP Fri Apr 27
23:59:24 UTC 2012 i686 GNU/Linux

I dont think this patch we applied to Quagga made the difference. At
this point I think it was a Kernel related problem.

any feedback ?

we are going soon to test quagga 0.99.21:
https://github.com/zioproto/quagga-manet/commits/openwrt-0.99.21

ciao,

Saverio

2012/5/19 ZioPRoTo (Saverio Proto) <(spam-protected)>:
> Hello,
>
> looks like we found weird bug in olsrd quagga plugin when
> redistributing IPv6 routes to quagga. I am not really sure if the bug
> is from our plugin or from quagga, so feel free to forward this email
> where necessary to solve the bug.
>
> I attached the configuration. Quagga has an aggregate route /48 that
> is announced to upstream if any subnet of this aggregate is learnt
> from olsrd.
> quagga-0.99.20.1
> patched with this
> https://dev.openwrt.org/browser/packages/net/quagga/patches/120-quagga_manet.patch
>
> About the Linux machine: Linux tecnopolo 2.6.32-24-generic-pae
> #39-Ubuntu SMP Wed Jul 28 07:39:26 UTC 2010 i686 GNU/Linux
>
> we are using this git commit of olsrd stable:
> 1b5a2075bd90c0f2b96d00bdc03a7e0874b7922b
> the configuration is attached.
>
> *** olsr.org -
> pre-0.6.3-git_1b5a207-hash_98587ecc4fbbb15af6cf90a35c5aae23  - ***
>  Build date: 2012-03-31 16:18:20 on tecnopolo
>  http://www.olsr.org
>
> everything seems to work but about 20-30 times per day we have the
> Linux router unable to send out any packets in IPv6 (in the same time
> IPv4 works regularly). Sometimes this lock situation is short, like
> 5-6 seconds, other times is longer and it makes our BGP sessions flap
> and reset.
>
> we used this tool running in two instances to check the two interfaces
> of the router.
>
> https://github.com/ninuxorg/misc_tools/tree/master/watchdog-nocrontab
>
> here what we logged so far:
>
> Logs from our tool:
> 2012-05-14 00:58:49.213866problem
> 2012-05-14 00:59:32.185883problem
> 2012-05-14 06:17:36.569407problem
> 2012-05-14 06:18:27.385295problem
> 2012-05-14 08:59:10.212263problem
> 2012-05-14 08:59:42.716765problem
>
> Logs from kernel:
> May 14 00:59:07 ubuntu kernel: [34675043.052027] icmpv6_send: no reply
> to icmp error
> May 14 00:59:18 ubuntu kernel: [34675054.640013] icmpv6_send: no reply
> to icmp error
> May 14 00:59:30 ubuntu kernel: [34675066.152013] icmpv6_send: no reply
> to icmp error
> May 14 01:04:19 ubuntu kernel: [34675355.064014] icmpv6_send: no reply
> to icmp error
> May 14 06:17:57 ubuntu kernel: [34694172.976024] icmpv6_send: acast source
> May 14 06:17:57 ubuntu kernel: [34694172.976038] icmpv6_send: acast source
> May 14 09:17:27 ubuntu kernel: [34704943.168040] icmpv6_send: no reply
> to icmp error
> May 14 09:17:57 ubuntu kernel: [34704973.156010] icmpv6_send: no reply
> to icmp error
> May 14 09:18:05 ubuntu kernel: [34704981.216021] icmpv6_send: no reply
> to icmp error
>
> what you see in the command line:
> (spam-protected):~# python peer_watchdog.py
> unknown host
> connect: Network is unreachable
> connect: Network is unreachable
> unknown host
> connect: Network is unreachable
> unknown host
> connect: Network is unreachable
> connect: Network is unreachable
> unknown host
> unknown host
> connect: Network is unreachable
> connect: Network is unreachable
> unknown host
>
> Also we tested that if we stop olsrd and we redistribute a into BGP
> the static route of our /48 network, then quagga works ok without any
> lock.
>
> Vasilis can you give me some hints on how to debug further this problem ?
>
> thanks
>
> Saverio