I had promised to update those attending our session via the list, which
is why I posted the note.

Kurt was kind enough to post the APAR description of the problem.  The symptoms
are very bizarre.  It begins with virtual machines being unable to logoff
the system (going into logoff/force pending mode).  Then you notice that
new machines cannot log on, they just sit at the logon message.  Then
other things can start to happen, such as a machine sending a message to
another machine would hang.  In one case, I was inserting some records into
a DB2 database, and that triggered the problem.  (We use dataspaces heavily.)

This has been a hard problem to pin down since the symptoms are so strange
and once it starts, you lose the ability to control the system or
do diagnostics.  (Try shutting down a large system manually because your
PROP machine is hosed!)

We were the only site in the ESP to experience this problem, but I have
heard that at least one other site has had it since (I don't know who).  We
experienced it within 4-5 hours of upgrading in production on our z/900.
The system on our z/990 has not seen the problem at all, and has been running
z/VM 5.2.0 since October.  That system is only running Linux guests and
is not heavily loaded.

Our z/900 system, which has seen the problem, is very heavily loaded with a
lot of z/OS guests, large SFS servers and an active DB2 machine.  The SFS
servers and DB2 use dataspaces heavily.  So, if your workload is similar to
this, you should probably be concerned.  If you are just running some guests,
you might be fine.  The other ESP sites like this did not see the problem.

It took awhile for us to be able to get enough diagnostic information for
IBM to work with.  They are treating this as a high priority.  It is the
first time since the old VM/XA SP days (late '80s) that Marist has had to
back out of a CP upgrade.  However, given the size and complexity of the
changes in VM 5.2.0, I think it is understandable.  That is why Marist
participates in these ESP tests....we find the weird and wacky problems.
Comes from being built on an ancient Indian burial ground.....

Martha

Reply via email to