yahoo-eng-team team mailing list archive

[Simon Effenberg <launchpad@xxxxxxxxxxxxxx>]

Public bug reported:

We tried to create a 2-network-node active/passive HA setup using
Corosync/Pacemaker and two identical network nodes with only one having
all services running.

To ensure that agents looks like coming only from one "virtual" network
node we specified the "host" parameter in /etc/neutron/neutron.conf

So having:

'network1' and 'network2' machines both configured with "host =
network".

If you try to do a failover almost everything works out of the box
except the dhcp agent because this one creates a new port with a new IP
address. But all VMs are using the old IP address as nameserver. You
could run somehow a active/active setup on network1 and network2 and add
both dhcp/dns agents to the /etc/resolv.conf but this is also giving you
timeouts (at least in Linux with round-robin option enabled).

If the dnsmasq would re-use the same port everything would work.

I attached a patch to fix this "issue". But I'm not sure if I miss any
situation which could lead to new issues.

** Affects: neutron
     Importance: Undecided
         Status: New


** Tags: dhcp dns neutron-dhcp-agent

** Patch added: "linux dhcp agent patch"
   https://bugs.launchpad.net/bugs/1372586/+attachment/4211667/+files/dhcp.patch

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to neutron.
https://bugs.launchpad.net/bugs/1372586

Title:
  dhcp-agent creates unexpected ports (ignores 'host' parameter) so HA
  is not working

Status in OpenStack Neutron (virtual network service):
  New

Bug description:
  We tried to create a 2-network-node active/passive HA setup using
  Corosync/Pacemaker and two identical network nodes with only one
  having all services running.

  To ensure that agents looks like coming only from one "virtual"
  network node we specified the "host" parameter in
  /etc/neutron/neutron.conf

  So having:

  'network1' and 'network2' machines both configured with "host =
  network".

  If you try to do a failover almost everything works out of the box
  except the dhcp agent because this one creates a new port with a new
  IP address. But all VMs are using the old IP address as nameserver.
  You could run somehow a active/active setup on network1 and network2
  and add both dhcp/dns agents to the /etc/resolv.conf but this is also
  giving you timeouts (at least in Linux with round-robin option
  enabled).

  If the dnsmasq would re-use the same port everything would work.

  I attached a patch to fix this "issue". But I'm not sure if I miss any
  situation which could lead to new issues.

To manage notifications about this bug go to:
https://bugs.launchpad.net/neutron/+bug/1372586/+subscriptions