arping can be spawned in background to speed-up floating IP assignment

[Brian Haley]

The arping command is spawned in a shell from the l3-agent with a default of sending 3 requests (send_arp_for_ha = 3). This can increase the time the agent spends waiting by 2 seconds for each IP address, which adds up on a very busy network node, and is worse then the agent is restarted and associates many IP addresses on a router at the same time.

Spawning the arping call in a greenthread will help speed this up, and is perfectly fine since if the arping fails for some reason it will still log an error, just asynchronously.

In addition to floating IP assignment, this also applies to any other address assigned to the external interface of a router.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/49063

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/49063
Committed: http://github.com/openstack/neutron/commit/3bcc64cf108e6a3999ab803d3c7625eb4f66843e
Submitter: Jenkins
Branch: master

commit 3bcc64cf108e6a3999ab803d3c7625eb4f66843e
Author: Brian Haley <email address hidden>
Date: Fri Sep 27 10:57:40 2013 -0400

    Spawn arping in thread to speed-up floating IP

    Change _send_gratuitous_arp_packet() to spawn a thread to call
    arping after a floating IP is assigned. This way it doesn't
    stall _process_routers() from returning quickly due to calling
    pool.waitall().

    Fixes Bug: 1233391

    Change-Id: Id1f5eb75c222ba6a0935a294e3973292f50d0559