configuring tcp timeout in Solaris 10

[Radje]

An application (which I can't change / configure) is setting up a
connection towards another machine (certain ip-address and port). When
this machine is not reachable, it takes almost to 4 minutes before a
timeout is given.

My question is if the timeout is configurable in Solaris 10?
I printed tcp parameters with: ndd /dev/tcp \?
However, I can't clearly see which parameter it could be.

Anyone who could help me out?

Regards,
Marcel

[ufo]

> My question is if the timeout is configurable in Solaris 10?
> I printed tcp parameters with: ndd /dev/tcp \?
> However, I can't clearly see which parameter it could be.

Timeout:
/usr/sbin/ndd -set /dev/tcp tcp_ip_abort_cinterval <milliseconds>

Useful as well: TCP port reuse wait time:
/usr/sbin/ndd -set /dev/tcp tcp_time_wait_interval <milliseconds>

Uwe

[Andrew Gabriel]

In article <c2fa5608-516f-4ebe...@s19g2000prg.googlegroups.com>,

tcp_ip_abort_cinterval is the major part (3 minutes) of that timeout.
The extra is retransmit timers, which you shouldn't fiddle with.

It's a long time since I played in that code, but there are some other
fancy things which might be coming into play, such as TCP telling IP
after a period of no response to forget any next hop it has cached
away in the IRE and to work out the route again (in case it needs to
change).

--
Andrew Gabriel
[email address is not usable -- followup in the newsgroup]