tree 4d17beb6fcf3e31b715cdbe7120e07ed0cf179f5
parent 7bf837a5b667bffc19b25a4c841e0ab33d01d4dc
author Pavlin Radoslavov <pavlin@onlab.us> 1389310500 -0800
committer Pavlin Radoslavov <pavlin@onlab.us> 1389310500 -0800

Add two hacks to get around some initially discovered sub-optimal behavior:

* Fake it that the Flow Entries have been already pushed
  into the switches, so we don't push them again.

* Fake the topology read it by checking the cache with
  the old topology and ignoring topology events that don't make
  any impact to the topology.
  This is needed aa workaround: if a port is down, we get
  up to three additional "Port Down" or "Link Down" events.