On Mon, Dec 18, 2017 at 04:51:39PM -0600, Mark Michelson wrote: > This modifies ovn-controller to measure the amount of time it takes to > detect a change in the southbound database, generate the resulting flow > table, and write the flow table down to vswitchd. > > The statistics can be queried using: > > ovs-appctl -t ovn-controller performance/show ovn-controller-loop > > Signed-off-by: Mark Michelson <[email protected]>
In ctrl_register_ovs_idl(), I think that the new code here should be in a performance_*() function, since it's the performance code that knows what columns are needed. Since these columns are "write-only" from ovn-controller's perspective, ovsdb_idl_omit_alert() should be called on each of them. Maybe there should be a helper function to call both ovsdb_idl_add_column() and ovsdb_idl_omit_alert() on a column. This patch adds an unneeded blank line in main(). I am a bit uncomfortable adding this amount of stuff to the ovs-vswitchd database exclusively for ovn-controller. I wonder whether ovn-controller should have its own hypervisor-local database, or whether this data should instead be turned over to some external time series database. Whatever we cook up is going to be inferior to any dedicated time series database. _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
