We have a TSM 6.2.2.0 server configured as a library manager. One of
the library manager clients is a TSM 5.5.4.0 server. Both run under
mainframe Linux. TCP connections between the two TSM servers pass
through a firewall. Some TCP connections involved in library
management are idle for long periods. Sessions for tape mounting
may wait a long time for a tape drive to become available. Sessions
for ejecting tape volumes (triggered by 'move drmedia' commands with
'tostate=vault') may wait a long time for somebody to unload the
library I/O station. Sessions that are inactive tend to fail with
ANR3174E messages on the library manager. We suspect that this happens
because the firewall drops TCP sessions that are idle for too long.
We are hoping to prevent the session drops by having keepalive packets
sent at sufficiently short intervals. As I understand the process, the
program that opens a socket decides whether keepalive packets will be
sent, and operating system parameters determine how often keepalive
packets will be sent if a program requests them for a particular
connection. Is this correct? Does TSM request keepalive packets for
TCP connections involved in library management? Do we need to get
the tcp_keepalive_time below the firewall time limit on the host
system for the library manager, the host system for the library
manager client, or both?