Docker Community Forums

Share and learn in the Docker community.

Communication between containers failing, dmesg vxlan messages


(Retardtank) #1

Hello,

Currently running a swarm on some bare-metal hosts on a bare-metal network. Everything has been running fine for a while, when (seemingly for no obvious reason) two containers on two different nodes (silently) weren’t able to communicate with each other anymore. The only thing I can find divergent in any logs is from dmesg on Node A, that had a container not being able to communicate with another container on Node B, (repeatedly):

[Thu Dec  7 18:32:16 2017] net_ratelimit: 45 callbacks suppressed
[Thu Dec  7 18:32:16 2017] vxlan: non-ECT from 10.112.182.4 with TOS=0x2
[Thu Dec  7 18:33:09 2017] vxlan: non-ECT from 10.112.182.4 with TOS=0x2
[Thu Dec  7 18:33:09 2017] vxlan: non-ECT from 10.112.182.4 with TOS=0x2
[Thu Dec  7 18:38:21 2017] vxlan: non-ECT from 10.112.182.4 with TOS=0x2
[Thu Dec  7 18:39:02 2017] vxlan: non-ECT from 10.112.182.4 with TOS=0x2
[Thu Dec  7 18:39:14 2017] vxlan: non-ECT from 10.112.182.4 with TOS=0x2
[Thu Dec  7 18:39:56 2017] vxlan: non-ECT from 10.112.182.4 with TOS=0x2
[Thu Dec  7 18:39:56 2017] vxlan: non-ECT from 10.112.182.4 with TOS=0x2

However, 10.112.182.4 is not Node B, it’s yet another host in the swarm, Node C.

Is it possible that this, although it concerns Node A and C, has anything to do with communication failing between containers on Node A and B?

Not being particularly familiar with the lower levels of vxlan, can anyone shed a light on what it means?

Thanks.