Very Computer

I would like to know if there is a way to kill the sqlturbo processes
which are left idle whenever a PC gets rebooted (or loses it's network
connection).  I am told that there is a kernel parameter which was able
to be set under Sun OS 4.1.3.

How did I get the disconnected sqlturbo processes to die and allow new
connections through the network?

Quote:

>I would like to know if there is a way to kill the sqlturbo processes
>which are left idle whenever a PC gets rebooted (or loses it's network
>connection).  I am told that there is a kernel parameter which was able
>to be set under Sun OS 4.1.3.

>How did I get the disconnected sqlturbo processes to die and allow new
>connections through the network?

Document ID: 1007

SYNOPSIS: How to set socket keep alive timeout length?

DETAIL DESCRIPTION:

Under 4.1.x the socket KEEP_ALIVE time out length was configured in the
kernel.  One could change the timeout value by editing the 'tcp_keepidle'
value in /sys/netinet/in_proto.c

How can the timeout value be changed in Solaris 2.x and in what units?  
Trying to add a line like "set tcp_keepidle=240" in /etc/system just
causes a warning at boot time that there is no tcp_keepidle param.

SOLUTION SUMMARY:

You can change the TCP keepalive interval by using the "ndd" command.  
You can run the following command to view the current setting of the TCP
keepalive timeout value:

        % ndd /dev/tcp tcp_keepalive_interval
        7200000

The value is in units of milliseconds.  Use the "ndd" command with the
"-set" flag to change the keepalive timeout value, e.g.:

        % ndd -set /dev/tcp tcp_keepalive_interval 1200000

Note that there is a bug in Solaris 2.2 that prevents TCP keepalives from
working correctly.  The bugid is 1084913.  This bug is fixed in Solaris
2.3.

Guess that about sums it up!  Hope that this helps.