PMTU = 1518 over local network at 1500 MTU

[Florent Bautista]

Thank you for your answer.

On 02/26/2014 11:33 AM, Guus Sliepen wrote:
> On Tue, Feb 25, 2014 at 01:53:51PM +0100, Florent Bautista wrote:
>
>> I have two nodes, connected to a switch, using Tinc 1.1 from git.
>>
>> They connect each other with sptps, and to other nodes in the Internet
>> with old protocol because they have Tinc 1.0.
>>
>> There is no problem with remote nodes, but between my 2 local nodes,
>> they see 1518 PMTU. But local network is 1500 MTU !!! So nodes can ping
>> each other but larger data does not go.
> That is strange, when the network has a MTU of 1500, tinc should detect a PMTU
> of 1451. I cannot reproduce the problem you have. Could it be that those two
> nodes are running inside a VM or that you have jumbo frames enabled on the
> local network?

nodes are physical nodes, not VM.

The switch maybe support jumbo frames but we do not use it.

eth0 of each node is connected in gigabit to a same switch, and eth0 is
set in a linux bridge "vmbr1" with MTU of... 1500 :)

vmbr1     Link encap:Ethernet  HWaddr d4:3d:7e:b2:67:8b 
          inet addr:192.168.0.202  Bcast:192.168.0.255  Mask:255.255.255.0
          inet6 addr: fe80::d63d:7eff:feb2:678b/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:405370568 errors:0 dropped:0 overruns:0 frame:0
          TX packets:404822575 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:0
          RX bytes:649040311034 (604.4 GiB)  TX bytes:422387052494
(393.3 GiB)



bridge name    bridge id        STP enabled    interfaces
vmbr1        8000.d43d7eb2678b    no        eth0
                            tap112i0
                            tap113i0
                            tap117i0
                            tap118i0
                            tap119i0
                            tap120i0
                            tap121i0
                            tap122i0
                            tap126i0


tap devices are VM. Each have 1500 MTU.


I really do not understand what could cause that...

When I set manually PMTU in Tinc config, it stays a few minutes then get
back to 1518...


>
>> I did -d5 in Tinc daemon but I see nothing interesting:
>>
>> root at test2:~# cat /var/log/tinc-coppint.log | grep PMTU
>> 2014-02-25 13:31:18 tinc.coppint[413140]: Increase in PMTU to sllm1
>> (192.168.0.200 port 655) detected, restarting PMTU discovery
> Hm, that is somewhat interesting, normally you would not expect the MTU of the
> local network to change.
>
>

It is changing...

I don't remember how exactly, but making some iperf inside VM of my
nodes, Tinc set a PMTU of 1451 once, but a few minutes later it goes
back to 1518..

What kind of information can I provide to you ?