OVSIntPort MTU set to 65000 between br-ex and br-floating

Bug #1663591 reported by Maël Kimmerlin
14
This bug affects 2 people
Affects Status Importance Assigned to Milestone
Fuel for OpenStack
Confirmed
High
Fuel Sustaining
Nominated for Ocata by Oleksiy Molchanov
Mitaka
Confirmed
High
Fuel Sustaining
Newton
Confirmed
High
Fuel Sustaining

Bug Description

The MTU of the link between br-ex and br-floating is set to 65000. However, due to the firewalling on br-ex, the fragmented packets from external networks are reassembled and fragmented again in br-ex. The incorrect MTU size leads to oversized refragmented packets, dropped later.

If the MTU of the link would be the same as the external network, the packets would be reassembled and refragmented correctly. Since the MTU is set to 65000, the fragmentation after reassembly for incoming packets do not happen correctly and the packets are oversized when exiting br-ex. They get discarded later on before reaching any router or VM.

Expected output :
4: ens18.20@ens18: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue master br-ex state UP mode DEFAULT group default qlen 1000
    link/ether 16:ae:18:f4:71:56 brd ff:ff:ff:ff:ff:ff
7: br-ex: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP mode DEFAULT group default qlen 1000
    link/ether 16:ae:18:f4:71:56 brd ff:ff:ff:ff:ff:ff
10: p_ff798dba-0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue master br-ex state UNKNOWN mode DEFAULT group default qlen 1
    link/ether 36:22:d0:cd:bf:d4 brd ff:ff:ff:ff:ff:ff

Actual output :
4: ens18.20@ens18: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue master br-ex state UP mode DEFAULT group default qlen 1000
    link/ether 16:ae:18:f4:71:56 brd ff:ff:ff:ff:ff:ff
7: br-ex: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP mode DEFAULT group default qlen 1000
    link/ether 16:ae:18:f4:71:56 brd ff:ff:ff:ff:ff:ff
10: p_ff798dba-0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 65000 qdisc noqueue master br-ex state UNKNOWN mode DEFAULT group default qlen 1
    link/ether 36:22:d0:cd:bf:d4 brd ff:ff:ff:ff:ff:ff

brctl show :
br-ex 8000.16ae18f47156 no ens18.20
                                                        p_ff798dba-0

ovs-vsctl show :
Bridge br-floating
        Controller "tcp:127.0.0.1:6633"
            is_connected: true
        fail_mode: secure
        Port "p_ff798dba-0"
            Interface "p_ff798dba-0"
                type: internal
        Port br-floating
            Interface br-floating
                type: internal
        Port phy-br-floating
            Interface phy-br-floating
                type: patch
                options: {peer=int-br-floating}

configuration file of p_ff798dba-0:
# *********************************************************************
# This file is being managed by Puppet. Changes to interfaces
# that are not being managed by Puppet will persist;
# however changes to interfaces that are being managed by Puppet will
# be overwritten.
# *********************************************************************
auto p_ff798dba-0
allow-br-floating p_ff798dba-0
iface p_ff798dba-0 inet manual
ovs_type OVSIntPort
mtu 65000
ovs_bridge br-floating

The MTU should be the same as for the external interface, 1500 by default, but configured anyways through the Fuel UI as the interface MTU.

The environment is the following:
Fuel 10, Ubuntu 16.04, Ubuntu packages (not UCA), up-to-date
1 controller, 2 computes. 1 physical interface per node, vlans for all the networks. VXLAN tunneling.
every node where the external network is deployed presents the issue.

version 10.0.0 build ID 865

Tags: l23network
Changed in fuel:
milestone: none → 11.0
importance: Undecided → High
assignee: nobody → Fuel Sustaining (fuel-sustaining-team)
status: New → Confirmed
tags: added: l23network
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.