We have a 5.1.7.2 TSM server running under OS/390 2.9. We have seen
a rather peculiar problem several times in the last few weeks. The
problem was discovered when I checked on the status of TSM early in
the morning. In each case there were a number of clients with more
sessions than usual. Some of the extra sessions were in 'Idle wait'
status, some in 'Receive wait' and some 'Running'. The ones in
'Running' status reported 0 second wait times but did not show any
change in the number of bytes transferred. The others showed wait
times up to several hours. The recovery log utilization was unusually
high. The 'show logpin' command would report that the oldest
transaction was connected with one of the sessions in 'Receive wait'
or 'Running' status. When the offending session was cancelled the
log utilization would drop somewhat, and one of the other extra
sessions would be the owner of then oldest transactions. I generally
had to cancel about half a dozen sessions before the log utilization
dropped to a normal value and ownership of the oldest transaction
started shifting rapidly among different sessions. All of the clients
with extra sessions have been Windows systems. Most of them have been
Windows 2000 but a few have been Windows NT. All have been running
5.1.5.0 client code. We this this phenomenon on four days so far.
Two of the occurences have been on Saturdays and two on Tuesday.
Each of the Tuesday occurances has been three days after one of the
Saturday occurances. On each of the Saturdays when the problem
occured the OS/390 console log reported large numbers of OSPF
adjacency failures. We did not see any sign of TCP/IP problems
associated with the Tuesday occurances. Does anyone recognize this
as a known problem, or have any suggestions for trouble-shooting it?