RE: Weirdness with 'ons' process....

Mark,

This won't solve your problem, but connections in TIME_WAIT state aren't open anymore ;)

The socket info is kept just for being able to receive any "lost" and "detoured" packets from the network for that connection.

There's Linux kernel parameter tcp_fin_timeout which you can tune lower - that closed sockets connection data would be cleared out sooner.

http://publib.boulder.ibm.com/wasce/V1.0.1.1/en/Tasks/Tuning/Linux.html

Reducing this for LAN networks shouldn't be much of a problem, as generally the packets there don't get detoured and lost that often.

I have never changed it myself though, don't have an idea about its side effects..

Tanel.

> -----Original Message-----
> From: oracle-l-bounce_at_freelists.org
> [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Bobak, Mark
> Sent: Wednesday, November 01, 2006 21:47
> To: Alex Gorbachev
> Cc: oracle-l
> Subject: RE: Weirdness with 'ons' process....
>
> Hi Alex,
>
> No, I mean there are 41,446 distinct open connections in
> TIME_WAIT state and 63,961 distinct open connections on the
> other node. Those numbers were a reference to the total
> aggregate number of connections on port
> 6101 on each node, not a reference to the source port number.
> Running 'netstat -an|grep -c ":6101 "' takes several minutes.
>
> -Mark

--
http://www.freelists.org/webpage/oracle-l