Hi,

I have Solaris 10 update 5. I have a user-level process doing RPCs to a
remote host.

Sometimes if the TCP queue is full, my RPC call blocks until the request
gets in  the TCP queue.
If the host is down while the TCP queue is full, then the RPC gets blocked
till TCP tcp_ip_abort_interval expires (which is 8 minutes).

This is exactly why non-blocking IO was introduced:
http://docs.sun.com/app/docs/doc/816-3576/6m9rl9rgp?l=ko&a=view
I have used non-blocking IO as described in this link. However, even with
non-blocking I/O I see the same behavior - RPC blocks till the data gets in
the TCP queue.

Has anyone seen this before? Am I missing something here? Thanks.
_______________________________________________
Solaris-Users mailing list
[email protected]
http://www.filibeto.org/mailman/listinfo/solaris-users

Reply via email to