When a virtual interface is brought down it makes a call to rt_cache_flush() with a -1 parameter which causes an up to <min_delay> delay before flushing all the routes. So, the server keeps responding for a little while after shuting down the interface. With a "real" interface, rt_cache_flush() gets called with a 0 parameter meaning that the routes get flushed immediately. I dont want the server to keep responding for a little while after shutting down the interface.

My questions are as follows:

1) Why rt_cache_flush() is called with a -1 parameter when virtual interface is brought down? Is this a bug or expected behavior?

3) Why the /proc/sys/net/ipv4/route/min_delay setting is by default set to 2?

3) When would we not want route flushes and route changes to take immediate effect?

Thanks,
Asha