From: "Steven Dake (stdake)" <[email protected]<mailto:[email protected]>>
Reply-To: "OpenStack Development Mailing List (not for usage questions)" 
<[email protected]<mailto:[email protected]>>
Date: Tuesday, April 28, 2015 at 7:52 AM
To: "OpenStack Development Mailing List (not for usage questions)" 
<[email protected]<mailto:[email protected]>>
Subject: Re: [openstack-dev] [kolla] implement openvswitch container



From: FangFenghua <[email protected]<mailto:[email protected]>>
Reply-To: "OpenStack Development Mailing List (not for usage questions)" 
<[email protected]<mailto:[email protected]>>
Date: Tuesday, April 28, 2015 at 7:02 AM
To: 
"[email protected]<mailto:[email protected]>" 
<[email protected]<mailto:[email protected]>>
Subject: [openstack-dev] [kolla] implement openvswitch container

I want to enbale openvswitch container.  I tink i can do that like:

1 add a container that run ovs process

OVS is broken down into several daemons, so I suggest using supervisord to have 
these daemons run in a single container before breaking them apart. This is 
because these daemons communicate through sockets by default. The daemons can 
be configured to use tcp, but I don’t know the details for changing that config 
off hand.

2 add a container that run neutron-openvswitch-agent
3  share the db.sock in compose.yaml
I think this is where you want to use tcp instead of sockets
4 add configure script and check script for the 2 containers

As a first step, you may want to simply replicate the current neutron-agents 
container and swap linux bridge for ovs and get ovs/l3/meta/etc working in a 
single container first, then split the agents apart into individual containers.

that's all i need to do, right?

That should do it

You may need to configure the ovs process in the start.sh script and 
neutron-openvswitch-agent, which will be the most difficult part of the work.

Note our agents atm are a “fat container” but if you can get ovs in a separate 
container, that would be ideal. We are planning to redux the fat container we 
have to single-purpose containers.

Regards
-steve
__________________________________________________________________________
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: [email protected]?subject:unsubscribe
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev

Reply via email to