My original message:
> Under Solaris 2.2, I am having a lot of problems with sockets that
> did not exist under SunOS. Notably, if a process crashes or a network
> connection goes down, I get sockets left in LAST_ACK, FIN_WAIT_2, and
> other states.
>
> Sun's patch 101027 which addresses SO_KEEPALIVE problems has helped
> somewhat, but I'm wondering if there are other patches available or if I
> can force sockets in this state to timeout.
>
> I have experimented with reducing some of the timeout periods and it has
> helped, i.e. "ndd -set /dev/tcp tcp_close_wait_interval 60000" but I can't
> find any "tweakable" parameter which will allow me to force sockets in
> FIN_WAIT_2 to timeout.
Sun technical support has confirmed that they are working on Bug #1135394
which addresses detached sockets in the FIN_WAIT_2 state. My understanding
of the problem is that these sockets never timeout, so if they are bound
to an address you need, you have to reboot to clear them. The tcp
keepalive timer should time these sockets out, so once this fix is in
place, ndd can be used to reduce the timeout value of the keepalive timer.
-- Mike Bauer -- Valmet Automation (Canada) Ltd. --
-- -- Sage Systems Division --
-- (403) 253-8848 x228 -- #1, 6325 - 12th St. S.E., --
-- mikeba@sa-cgy.valmet.com -- Calgary, Alberta, Canada, T2H 2K1 --
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:08:04 CDT