Large Packet Problem

The problem we are currently having on the kilo deployment is sending large packets.

Symptoms

  • ssh times out after first request, outside of cluster
  • ssh only works from the controller node hosting the router to the vm
  • Large packets (greater then 1480 bytes, 1472 + 8 byte ping header) get an icmp_timeout
  • Large packets can flow between compute and controller node, but not from either to the internet, or vice versa