[ovs-dev] [PATCH v1] Enable manager_option configuation for NB and SB

Amitabha Biswas azbiswas at gmail.com
Thu Jul 28 20:14:37 UTC 2016


Hi Ben,

No there is no advantage for having a separate database, we can incorporate a Manager Table into the Northbound and Southbound database. This was an initial hack (with minimal code and schema change) to verify the efficacy of disabling the probe timer on our cloud setups and get some feedback from the community. I think our cloud deployment team has come to the conclusion that we do need the ability to modify the probe timer values in the ovsdb-servers that manage the NB and SB databases.

So v2 of this patch will contain Manager Table in the NB and SB database. Does that sounds reasonable?

Thanks
Amitabha


> On Jul 27, 2016, at 11:45 AM, Ben Pfaff <blp at ovn.org> wrote:
> 
> On Mon, Jul 18, 2016 at 10:21:09PM -0700, Amitabha Biswas wrote:
>> NB and SB dbs are handled by separate ovsdb-server processes. The
>> ovsdb-server processes manage dbs based on the schemas for OVN NorthBound
>> and SouthBound. These schemas do not have any Manager Options similar
>> to the Open_vSwitch schema. As a result, for e.g., the frequency of
>> inactivity probes sent to clients from the ovsdb-server cannot be
>> modified.
>> 
>> This patch addresses the above problem by creating independent
>> "conf.db"s for NB and SB. The ovsdb-server process which handles
>> the NB will handle a NB_conf.db as well, ditto for the ovsdb-server
>> that handles the SB.
>> 
>> A similar result can be obtained by adding a Manager Table to the
>> NorthBound and SouthBound DBs as well. In such case there would
>> not be a need for a separate "config" database. I'm willing to
>> consider this solution as well pending feedback.
> 
> Is there an advantage to having a separate database for this purpose?
> It strikes me as an odd design.
> 
> The vtep schema also in the OVS tree (vtep/vtep.ovsschema) has its own
> Manager table.




More information about the dev mailing list