[ovs-discuss] Running into MTU problem (2 extra bytes)

Johnson L. Wu johnson at snoopy.ORG
Wed Jun 24 21:43:36 UTC 2015


Thanks Jesse,

This is plain vanilla when all MTU left at default.  Dmesg showing 
[  752.962014] openvswitch: vnet5: dropped over-mtu packet: 1502 > 1500
[  752.962027] openvswitch: inter-1tom: dropped over-mtu packet: 1502 > 1500

* 10.128.0.12 is the VM on the virtual side
The others are the internet hosts
23.36.60.124 is apple.com trying to send it 1500Byte packets.
It stalls as I get the 1502 message in dmesg.

h# ovs-appctl   dpif/dump-flows switch-1
recirc_id(0),skb_priority(0),in_port(5),eth(src=20:e5:2a:07:0d:cf,dst=52:54:00:3d:a6:db),eth_type(0x0800),ipv4(src=23.36.60.124/255.255.255.255,dst=10.128.0.12/255.255.255.255,proto=6/0xff,tos=0x8/0,ttl=55/0,frag=no/0xff), packets:22, bytes:26040, used:1.388s, flags:SFP., actions:15,13
recirc_id(0),skb_priority(0),in_port(15),eth(src=52:54:00:3d:a6:db,dst=20:e5:2a:07:0d:cf),eth_type(0x0800),ipv4(src=10.128.0.12/255.255.255.255,dst=23.56.123.171/255.255.255.255,proto=6/0xff,tos=0/0,ttl=128/0,frag=no/0xff), packets:0, bytes:0, used:never, actions:5,13
recirc_id(0),skb_priority(0),in_port(5),eth(src=20:e5:2a:07:0d:cf,dst=52:54:00:3d:a6:db),eth_type(0x0800),ipv4(src=172.230.227.22/255.255.255.255,dst=10.128.0.12/255.255.255.255,proto=6/0xff,tos=0/0,ttl=57/0,frag=no/0xff), packets:0, bytes:0, used:never, actions:15,13
recirc_id(0),skb_priority(0),in_port(5),eth(src=20:e5:2a:07:0d:cf,dst=52:54:00:3d:a6:db),eth_type(0x0800),ipv4(src=23.56.123.171/255.255.255.255,dst=10.128.0.12/255.255.255.255,proto=6/0xff,tos=0/0,ttl=57/0,frag=no/0xff), packets:0, bytes:0, used:never, actions:15,13
recirc_id(0),skb_priority(0),in_port(5),eth(src=20:e5:2a:07:0d:cf,dst=ff:ff:ff:ff:ff:ff),eth_type(0x0800),ipv4(src=10.128.0.1/255.255.255.255,dst=255.255.255.255/255.255.255.255,proto=17/0xff,tos=0/0,ttl=64/0,frag=no/0xff),udp(src=32772/0x8000,dst=7423/0xf000), packets:0, bytes:0, used:never, actions:13

root at jwu-lenovo:/var/run/openvswitch# ovs-appctl   dpif/show      
system at ovs-system: hit:8304 missed:1447
        control-switch:
                control-switch 65534/2: (internal)
                vnet0 1/6: (system)
                vnet1 2/8: (system)
                vnet3 3/10: (system)
        mirror-switch:
                inter-mto1 2/14: (system)
                mirror-port 1/11: (system)
                mirror-switch 65534/1: (internal)
        switch-1:
                control-proxy-1 3/7: (system)
                inter-1tom 6/13: (system)
                svc-port-1 2/3: (internal)
                switch-1 65534/4: (internal)
                vmnic1 1/5: (system)
                vnet2 4/9: (system)
                vnet4 5/12: (system)
                vnet5 7/15: (system)
root at jwu-lenovo:/var/run/openvswitch#


-----Original Message-----
From: Jesse Gross [mailto:jesse at nicira.com] 
Sent: Wednesday, June 24, 2015 11:06 AM
To: Johnson L. Wu
Cc: Emil Flink; discuss at openvswitch.org
Subject: Re: [ovs-discuss] Running into MTU problem (2 extra bytes)

On Wed, Jun 24, 2015 at 10:34 AM, Johnson L. Wu <johnson at snoopy.org> wrote:
>
> Hi Jesse,
>
> I just checked to see if there's any vlan related settings in my OVS that might get it to insert a tag and I found nothing.
> I have a funny feeling that *might* be Debian 8 related vs OVS 2.3 (I am now using 2.3.2).  An old Wheezy machine using OVS OVS 1.11.1 doesnt have this problem when I compare them side-by-side.

Hmm, thanks for looking. I did notice that both of you are running Debian 8. I guess trying the Linux bridge is still the easiest check to make. One other thing: can you try running ovs-appctl dpif/dump-flows <BR> while the problem is occurring and post the output?



More information about the discuss mailing list