yahoo-eng-team team mailing list archive

[Michael Sherman <1988069@xxxxxxxxxxxxxxxxxx>]

Public bug reported:

High level description: 
When a user creates a tenant network with a very small MTU (in our case 70), neutron-dhcp-agent stops updating the dnsmasq configuration, causing DHCP issues for all networks.

Pre-conditions:
Neutron is using the openvswitch, baremetal, and networking-generic-switch mechanism drivers.
A physical network named `physnet1` is configured, with MTU=9000

Step-by-step reproduction steps:
As an admin user, run:

# Create "normal" network and subnet
openstack network create --provider-network-type vlan --provider-physical-network physnet1 --mtu 1500 test-net-1500
openstack subnet create --subnet-range 10.100.10.0/24 --dhcp --network test-net-1500 test-subnet-1500

# Create "small MTU" network and subnet
openstack network create --provider-network-type vlan --provider-physical-network physnet1 --mtu 70 test-net-70
openstack subnet create --subnet-range 10.100.11.0/24 --dhcp --network test-net-70 test-subnet-70

# attempt to launch an instance on the "normal" network
openstack server create --image Ubuntu --flavor Baremetal --network test-net-1500

Expected output: what did you hope to see?
We expected to see neutron-dhcp-agent update the dnsmasq configuration, which would then serve requests from the instances.

* Actual output: did the system silently fail (in this case log traces are useful)?
Openstack commands complete successfully, but instance never receives a response to its DHCP requests. Neutron-dhcp-agent logs show:
https://paste.opendev.org/show/b4r0XCu5KpguM72bnh0u/

Version:
  ** OpenStack version "stable/xena", hash bc1dd6939d197d15799aaf252049f76442866c21
  ** Linux distro, kernel. Ubuntu 20.04
  ** Containers built with Kolla, and deployed via Kolla-Ansible

* Environment: 
Single node deployment, all services (core, networking, database, etc.) on one node.
All compute-nodes are baremetal via Ironic.

* Perceived severity: is this a blocker for you?
High, as non-admin users can trigger an DHCP outage affecting all users.

** Affects: neutron
     Importance: Undecided
         Status: New


** Tags: l3-ipam-dhcp

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to neutron.
https://bugs.launchpad.net/bugs/1988069

Title:
  neutron-dhcp-agent fails when small tenant network mtu is set

Status in neutron:
  New

Bug description:
  High level description: 
  When a user creates a tenant network with a very small MTU (in our case 70), neutron-dhcp-agent stops updating the dnsmasq configuration, causing DHCP issues for all networks.

  Pre-conditions:
  Neutron is using the openvswitch, baremetal, and networking-generic-switch mechanism drivers.
  A physical network named `physnet1` is configured, with MTU=9000

  Step-by-step reproduction steps:
  As an admin user, run:

  # Create "normal" network and subnet
  openstack network create --provider-network-type vlan --provider-physical-network physnet1 --mtu 1500 test-net-1500
  openstack subnet create --subnet-range 10.100.10.0/24 --dhcp --network test-net-1500 test-subnet-1500

  # Create "small MTU" network and subnet
  openstack network create --provider-network-type vlan --provider-physical-network physnet1 --mtu 70 test-net-70
  openstack subnet create --subnet-range 10.100.11.0/24 --dhcp --network test-net-70 test-subnet-70

  # attempt to launch an instance on the "normal" network
  openstack server create --image Ubuntu --flavor Baremetal --network test-net-1500

  Expected output: what did you hope to see?
  We expected to see neutron-dhcp-agent update the dnsmasq configuration, which would then serve requests from the instances.

  * Actual output: did the system silently fail (in this case log traces are useful)?
  Openstack commands complete successfully, but instance never receives a response to its DHCP requests. Neutron-dhcp-agent logs show:
  https://paste.opendev.org/show/b4r0XCu5KpguM72bnh0u/

  Version:
    ** OpenStack version "stable/xena", hash bc1dd6939d197d15799aaf252049f76442866c21
    ** Linux distro, kernel. Ubuntu 20.04
    ** Containers built with Kolla, and deployed via Kolla-Ansible

  * Environment: 
  Single node deployment, all services (core, networking, database, etc.) on one node.
  All compute-nodes are baremetal via Ironic.

  * Perceived severity: is this a blocker for you?
  High, as non-admin users can trigger an DHCP outage affecting all users.

To manage notifications about this bug go to:
https://bugs.launchpad.net/neutron/+bug/1988069/+subscriptions