yahoo-eng-team team mailing list archive

[Edward Hope-Morley <edward.hope-morley@xxxxxxxxxxxxx>]

** Also affects: cloud-archive
   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/1694337

Title:
  Port information (binding:host_id) not updated for
  network:router_gateway after qRouter failover

Status in Ubuntu Cloud Archive:
  New
Status in Ubuntu Cloud Archive mitaka series:
  New
Status in Ubuntu Cloud Archive newton series:
  New
Status in Ubuntu Cloud Archive ocata series:
  New
Status in neutron:
  Fix Released
Status in neutron package in Ubuntu:
  Fix Released
Status in neutron source package in Xenial:
  In Progress
Status in neutron source package in Yakkety:
  Won't Fix
Status in neutron source package in Zesty:
  In Progress

Bug description:
  [Impact]

  When using l3 ha and a router agent fails over, the interface holding
  the network:router_gateway interface does not get its property
  binding:host_id updated to reflect where the keepalived moved the
  router.

  [Steps to reproduce]

  0) Deploy a cloud with l3ha enabled
    - If familiar with juju, it's possible to use this bundle http://paste.ubuntu.com/24707730/ , but the deployment tool is not relevant

  1) Once it's deployed, configure it and create a router see https://docs.openstack.org/mitaka/networking-guide/deploy-lb-ha-vrrp.html )
    - This is the script used during the troubleshooting
  -----8<------
  #!/bin/bash -x

  source novarc  # admin

  neutron net-create ext-net --router:external True
  --provider:physical_network physnet1 --provider:network_type flat

  neutron subnet-create ext-net 10.5.0.0/16 --name ext-subnet
  --allocation-pool start=10.5.254.100,end=10.5.254.199 --disable-dhcp
  --gateway 10.5.0.1 --dns-nameserver 10.5.0.3

  keystone tenant-create --name demo 2>/dev/null
  keystone user-role-add --user admin --tenant demo --role Admin 2>/dev/null

  export TENANT_ID_DEMO=$(keystone tenant-list | grep demo | awk -F'|'
  '{print $2}' | tr -d ' ' 2>/dev/null )

  neutron net-create demo-net --tenant-id ${TENANT_ID_DEMO}
  --provider:network_type vxlan

  env OS_TENANT_NAME=demo neutron subnet-create demo-net 192.168.1.0/24 --name demo-subnet --gateway 192.168.1.1
  env OS_TENANT_NAME=demo neutron router-create demo-router
  env OS_TENANT_NAME=demo neutron router-interface-add demo-router demo-subnet
  env OS_TENANT_NAME=demo neutron router-gateway-set demo-router ext-net

  # verification
  neutron net-list
  neutron l3-agent-list-hosting-router demo-router
  neutron router-port-list demo-router
  ----- 8< -------

  2) Kill the associated master keepalived process for the router
      ps aux | grep keepalived | grep $ROUTER_ID
      kill $PID

  3) Wait until "neutron l3-agent-list-hosting-router demo-router" shows the other host as active
  4) Check the binding:host_id property for the interfaces of the router
      for ID in `neutron port-list --device-id $ROUTER_ID | tail -n +4 | head -n -1| awk -F' ' '{print $2}' `; do neutron port-show $ID ; done

  Expected results:

  The interface where the device_owner is network:router_gateway has its
  property binding:host_id set to where the keepalived process is master

  Actual result:

  The binding:host_id is never updated, it stays set with the value
  obtainer during the creation of the port.

  [Regression Potential]
  - This patch changes the UPDATE query to the port bindings in the database, a possible regression will express as failures in the query or binding:host_id property outdated.

  [Other Info]

  The patches for zesty and yakkety are a direct backport from
  stable/ocata and stable/newton respectively. The patch for xenial is
  NOT merged in stable/xenial because it's already EOL.

To manage notifications about this bug go to:
https://bugs.launchpad.net/cloud-archive/+bug/1694337/+subscriptions