[Olsr-users] Received message to big to be forwarded

[Jernej Kos]

Hello!

Probably because they are generated by a node that has a larger MTU when
they are distributed into smaller messages. But the problem occurs when
such messages are received by other nodes with smaller MTUs on some
interfaces and they can't just forward the messages.

They would need to do some splitting of their own, but if I understand
this correctly, only the originator splits the messages? Or can
forwarders also split messages?

We are not using any extensions on top of olsrd signalling if that's
what you are asking.


Jernej

On 17. 04. 2015 14:43, Ferry Huberts wrote:
> Then why are these packets too large?
> Do you piggy-back something over olsrd that uses large messages?
> 
> On 17/04/15 14:41, Henning Rogge wrote:
>> Yes,
>>
>> they can and will be distributed over multiple messages... if they
>> don't fit into the local MTU.
>>
>> Henning
>>
>> On Fri, Apr 17, 2015 at 2:39 PM, Jernej Kos <(spam-protected)> wrote:
>>> Hello!
>>>
>>> So the reason for these big messages is probably the fact that there are
>>> nodes that have a lot of neighbours? Can these be split over multiple
>>> messages or not?
>>>
>>>
>>> Jernej
>>>
>>> On 17. 04. 2015 14:37, Henning Rogge wrote:
>>>> Hi,
>>>>
>>>> I think olsrd (1) already can split packets along the messages... the
>>>> problem happens when a message is too large even for a UDP packet of
>>>> its own.
>>>>
>>>> Henning
>>>>
>>>> On Fri, Apr 17, 2015 at 2:34 PM, Ferry Huberts <(spam-protected)>
>>>> wrote:
>>>>> OTOH
>>>>>
>>>>> We should be able to split packets with multiple messages into
>>>>> multiple
>>>>> packets as long as the messages themselves aren't to big.
>>>>> And also the reverse: merge packets for larger MTUs.
>>>>>
>>>>> IMHO olsrd should/can adapt to different MTU sizes.
>>>>>
>>>>> This most probably not go into v1 unless someone has a nice patch
>>>>> set (well
>>>>> tested!). But I think it's a good idea for v2.
>>>>>
>>>>>
>>>>> On 17/04/15 14:30, Henning Rogge wrote:
>>>>>>
>>>>>> The only other option I see is to set the MTU size for olsrd messages
>>>>>> to 1280-40-8=1232 bytes... which should be guaranteed for all IPv6
>>>>>> capable networks.
>>>>>>
>>>>>> Henning
>>>>>>
>>>>>> On Fri, Apr 17, 2015 at 2:25 PM, Ferry Huberts
>>>>>> <(spam-protected)> wrote:
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> On 17/04/15 14:22, Henning Rogge wrote:
>>>>>>>>
>>>>>>>>
>>>>>>>> On Fri, Apr 17, 2015 at 2:07 PM, Jernej Kos <(spam-protected)> wrote:
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Hello!
>>>>>>>>>
>>>>>>>>> We are seeing a lot of the following messages in our logs:
>>>>>>>>>
>>>>>>>>> Received message to big to be forwarded on digger1226(1296 bytes)!
>>>>>>>>> Received message to big to be forwarded on digger1346(1328 bytes)!
>>>>>>>>>
>>>>>>>>> The message sizes indicate that the OLSR messages that are to be
>>>>>>>>> forwarded are probably larger than the MTU (the OLSR daemon has
>>>>>>>>> multiple
>>>>>>>>> interfaces with different MTU sizes). Does this warning
>>>>>>>>> indicate some
>>>>>>>>> problems with propagating routing information as messages are
>>>>>>>>> dropped
>>>>>>>>> instead of being forwarded?
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> The problem is that olsrd cannot fragment messages (which are
>>>>>>>> flooded
>>>>>>>> through the whole mesh) hop-by-hop... which means a network with
>>>>>>>> different MTU-sizes AND large neighborhoods (which produce huge
>>>>>>>> messages) is a problem.
>>>>>>>>
>>>>>>>
>>>>>>> Yes, you should really setup a lower MTU on all your nodes
>>>>>>>
>>>>>>>> Henning Rogge
>>>>>>>>
>>>>>>>
>>>>>>> -- 
>>>>>>> Ferry Huberts
>>>>>
>>>>>
>>>>> -- 
>>>>> Ferry Huberts
>>>
> 

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: OpenPGP digital signature
URL: <http://lists.olsr.org/pipermail/olsr-users/attachments/20150417/1f56b56e/attachment.sig>