V3 fixes handling of deleted records passed by _FOR_EACH_TRACKED and closes a missed continue to avoid skipping a bypassed record from a later run.
I've tested this with a setup of 8000 openstack routers, 16000 openstack networks and 32000 openstack ports. At full load, lflow_run now takes about 1.5 seconds each cycle, while patch_run takes between 2 and 2.5 seconds each cycle. So, patch_run is now the long pole in the tent. _______________________________________________ dev mailing list [email protected] http://openvswitch.org/mailman/listinfo/dev
