yahoo-eng-team team mailing list archive

[Mohammed Naser <2059716@xxxxxxxxxxxxxxxxxx>]

Public bug reported:

We've got an interesting scenario where one of the backends of a load
balancer is not reachable given the following test environment:

2x networks
- provider network, IPv4 + IPv6 subnets
- tenant network (Geneve), IPv4 + IPv6 subnets

3x VMs
- 2x single port, 2 IP addresses on the tenant network
- 1x single port, 2 IP addresses on the tenant network + floating IP (IPv4 only) attached

Load balancer:
- Using single tenant network, with floating IP (IPv4 only) attached
- OVN provider

With the setup above, the VM with the floating IP attached will not be
reachable by the load balancer (aka, hitting it multiple times will
timeout 1/3 of the time).  If you remove the floating IP and re-attach
it, it works.

In troubleshooting, we've noticed that when removing the IPv6 subnet
from the tenant network resolves this, so I suspect that it's somehow to
do with that.

** Affects: neutron
     Importance: Undecided
         Status: New

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to neutron.
https://bugs.launchpad.net/bugs/2059716

Title:
  [ovn] Multihomed backend (IPv4 + IPv6) with floating IP unreachable

Status in neutron:
  New

Bug description:
  We've got an interesting scenario where one of the backends of a load
  balancer is not reachable given the following test environment:

  2x networks
  - provider network, IPv4 + IPv6 subnets
  - tenant network (Geneve), IPv4 + IPv6 subnets

  3x VMs
  - 2x single port, 2 IP addresses on the tenant network
  - 1x single port, 2 IP addresses on the tenant network + floating IP (IPv4 only) attached

  Load balancer:
  - Using single tenant network, with floating IP (IPv4 only) attached
  - OVN provider

  With the setup above, the VM with the floating IP attached will not be
  reachable by the load balancer (aka, hitting it multiple times will
  timeout 1/3 of the time).  If you remove the floating IP and re-attach
  it, it works.

  In troubleshooting, we've noticed that when removing the IPv6 subnet
  from the tenant network resolves this, so I suspect that it's somehow
  to do with that.

To manage notifications about this bug go to:
https://bugs.launchpad.net/neutron/+bug/2059716/+subscriptions