On Fri, Jan 8, 2016 at 11:25 AM, Muthukumaran K <[email protected] > wrote:
> Hi Kamal, Anil, Abhijit, > > > > We were testing the reconciliation with back-to-back switch reboots across > nodes – ie. switch disconnects from Node A and reconnects to Node B in > rapid successions. > > OVS does NOT use OF-HA capability and hence switch connects to only one > controller instance at a time. > > > > Observation : > > Disconnecting from Node A and connecting to Node B – Reconciliation FAILED > > But , when we introduced a delay of 1 minute to trigger reconciliation > after switch reconnected to Node B, reconciliation went through (of course, > we would continue to stress this path further to push 50-60K flows via > reconciliation) > > > > I assume that at Entity-Ownership level, still some wait-time is involved > before declaring the controller-node as entity-owner. > > > > But, should actual sequence not be as follows – so that once > DataChangeListeners of inventory gets notified of switch connected they can > immediately start acting without any delay (as done in our experiment > above)? > > Elect Entity-owner for switch -----à Perform Routed RPC registration ---à > Write switch into Inventory Operational DS shard > That's my understanding as well, that node data is written to operational data store once rpc registration is done ? is that not a case? Looking at the clustering logs it looks like even statistics are started before the rpc registration (but i need to look at the code to confirm that). > > > Regards > > Muthu > > > > _______________________________________________ > openflowplugin-dev mailing list > [email protected] > https://lists.opendaylight.org/mailman/listinfo/openflowplugin-dev > > -- Thanks Anil
_______________________________________________ openflowplugin-dev mailing list [email protected] https://lists.opendaylight.org/mailman/listinfo/openflowplugin-dev
