Anne Moore wrote:
Hi Marshall Well I've already determined that this will fix the issues. The problem is indeed with our firewall and it cannot be changed due to our security policy. Thus, I created a script that continually pings every 30 seconds and that keeps the logons alive.
This is part of the problem with 'sekuritee people' that don't actually understand the protocols.
TCP Keepalives are supposed to work to allow servers to figure out that persistent connections that have not sent data are still there - the RFC states that this should not default to anything less than 2 hours (its possible, but not advised)
http://www.uic.rsu.ru/doc/inet/tcp_stevens/tcp_keep.htm for a good, easy to read writeup
http://www.faqs.org/rfcs/rfc1122.html is the host requirements RFC, section 4.2.3.6 deals with keep alives.
There are a number of reasons for this default (explained nicely in the first link) and most sekuritee people cause no end of headaches for systems/network people when they start fiddling with this value in the name of 'sekuritee !'
It is completely normal for a TCP session to be idle, and it is also completely normal for it to wake up hours later and send data, this is simply how stuff works in the IP world, and what it appears is happening is that your ssh sessions are (as would be expected) idle for a few minutes and due to some sekuritee 'professional' deciding that this could NEVER happen, your user sessions are being disconnected. The correct fix is to lart the sekuritee moron and change the default keep alive value. If they want to enforce logoff on idle sessions then install or enable this on the servers. Changing these values on a firewall can have some VERY undesirable and difficult to fault-find consequences. (I had one instance where someone had set the value to 30 mins, oracle was timing out connections and things would sporadically work, not work, then semi work - took the best part of a day to fault find.)
The primary purpose of keep alives is to enable the host to not exhaust its resources by having 65500 dead yet open telnet/ssh/tcp sessions and being able to close these after a defined period., the firewall not working in sync with the host just compounds this problem, and depending on the number of users/types of processes, can actually cause the problem that keep alives are supposed to prevent.
-- Steve () ascii ribbon campaign - against html e-mail /\ www.asciiribbon.org - against proprietary attachments -- redhat-list mailing list unsubscribe mailto:redhat-list-request@xxxxxxxxxx?subject=unsubscribe https://www.redhat.com/mailman/listinfo/redhat-list
