yahoo-eng-team team mailing list archive

[Diego Lima <diegolima.br@xxxxxxxxx>]

Public bug reported:

The network interfaces in qlbaas namespaces always use 1500 as their
default MTU setting. This may cause trouble when communicating with
instances that have different MTU set (usually through dhcp options
using /etc/neutron/dnsmasq-neutron.conf). Different MTU values may break
communications between instances and load balancers and lead to weird
behaviours, such as some web pages timing out waiting for a reply while
others work.

 Manually setting an lbaas namespace to the same value as the instance
corrects any communication problems between balancer and instance.

Neutron should use options from their subnet or at least provide a way
to manually set the load balancer's MTU.

For reference, here is the output of a lbaas namespace settings showing
MTU 1500:

# ip netns exec qlbaas-a38fc742-9c0b-4503-a0b9-64d5a6adfe77 ip a
(...)
234: tape853dd58-66: <BROADCAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UNKNOWN group default 
    link/ether fa:16:3e:a2:e0:c7 brd ff:ff:ff:ff:ff:ff
    inet 10.134.54.151/22 brd 10.134.55.255 scope global tape853dd58-66
       valid_lft forever preferred_lft forever
    inet6 fe80::f816:3eff:fea2:e0c7/64 scope link 
       valid_lft forever preferred_lft forever

** 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/1376446

Title:
  Neutron LBAAS agent ignores MTU settings used by subnets

Status in OpenStack Neutron (virtual network service):
  New

Bug description:
  The network interfaces in qlbaas namespaces always use 1500 as their
  default MTU setting. This may cause trouble when communicating with
  instances that have different MTU set (usually through dhcp options
  using /etc/neutron/dnsmasq-neutron.conf). Different MTU values may
  break communications between instances and load balancers and lead to
  weird behaviours, such as some web pages timing out waiting for a
  reply while others work.

   Manually setting an lbaas namespace to the same value as the instance
  corrects any communication problems between balancer and instance.

  Neutron should use options from their subnet or at least provide a way
  to manually set the load balancer's MTU.

  For reference, here is the output of a lbaas namespace settings
  showing MTU 1500:

  # ip netns exec qlbaas-a38fc742-9c0b-4503-a0b9-64d5a6adfe77 ip a
  (...)
  234: tape853dd58-66: <BROADCAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UNKNOWN group default 
      link/ether fa:16:3e:a2:e0:c7 brd ff:ff:ff:ff:ff:ff
      inet 10.134.54.151/22 brd 10.134.55.255 scope global tape853dd58-66
         valid_lft forever preferred_lft forever
      inet6 fe80::f816:3eff:fea2:e0c7/64 scope link 
         valid_lft forever preferred_lft forever

To manage notifications about this bug go to:
https://bugs.launchpad.net/neutron/+bug/1376446/+subscriptions