DHCP agent: RPC too slow because of get all networks info in one time

[changzhi]

========== Problem Description ========
Currently, DHCP agent will get all active networks info (including subnets and its ports) from neutron server.
But, on a large scale, for example, if there have 2000+ networks, 2000+ subnets and thousands of ports. RPC
will be very very slow because of the message is too big so that time out for waiting for reply message.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/257100

[Carl Baldwin (carl-baldwin)]

This sounds like a few other bugs that I have seen. I'm going to look for duplicates.

[Carl Baldwin (carl-baldwin)]

Similar bugs:

https://bugs.launchpad.net/neutron/+bug/1516260 (L3 agent)
https://bugs.launchpad.net/neutron/+bug/1430999 (OVS agent)

Kevin Benton has began thinking about this problem. His spec is here: https://review.openstack.org/#/c/225995/

[changzhi (changzhi)]

hi Carl, please review my patch: https://review.openstack.org/#/c/257100

[OpenStack Infra (hudson-openstack)]

Change abandoned by Zhi Chang (<email address hidden>) on branch: master
Review: https://review.openstack.org/257100