1 of 6

  • New PoC deployment
  • Why can’t I ping my floating IP?
  • Among top ask.openstack.org questions

2 of 6

  • Can we automate some of the common troubleshooting steps?
  • Simplify Neutron, make it more usable
  • Improve Neutron’s perception
  • Four different RFEs

3 of 6

  • Not a monitoring system
  • First you have a problem, then you use the new tool for help
  • May be used as a response for “My network doesn’t work!”, or at least to rule out common issues

4 of 6

  • Example usage, ref. impl:

$neutron floatingip-health <fip>

ERROR, VM did not get a fixed IP

  • Example health checks, ref. impl:

Is router scheduled? Is the namespace there? Can the router ping the VM? Security group allows ICMP? Did the VM get an IP?

5 of 6

  • Example implementation design:
  • Add ‘health’ field to FIPs (Later possibly subnets, ports, …)
  • Each plugin/ml2 driver implements own health checks
  • For ref. impl: Add new ‘neutron-diagnostics’ agent, neutron-server(s) ask agent(s) to perform health checks via RPC

6 of 6

  • Discussion points in Etherpad:

etherpad.openstack.org/p/newton-neutron-troubleshooting