Never change the TOD clock that the sysplex is using,
once it is set. Doing so can lead to a sysplex-wide IPL, or having
working systems within the sysplex at risk of being partitioned out
of the sysplex (that is, placed in a non-restartable wait state).
Whether you choose to set the TOD clock to local time or to GMT time,
local time changes can be accomplished using one of the techniques
discussed in
Adjusting local time in a sysplex. You should be aware of the possible
consequences to XCF and the multisystem applications using XCF services
when you set the TOD clock backward or forward:
- Setting the TOD Clock Backward
As it monitors system
status, MVS™ records, in the sysplex
couple data set, the time stamps obtained from the TOD clock. If you
set the TOD clock back, then time stamps obtained from the clock may
duplicate time stamps already recorded in the sysplex couple data
set. For example, if the TOD clock is reset to 2:00 AM at 3:00 AM,
then all timings between 2:00 and 3:00 will occur again. When a system
attempts to IPL into the sysplex, MVS compares
time stamps from the TOD clock with time stamps recorded in the sysplex
couple data set. If time stamps in the couple data set are later than
those specified by the incoming system, MVS determines
this is an error and does not allow the system to join the sysplex.
The system is not allowed to IPL into the sysplex until the time stamp
from the TOD clock is greater than time stamps recorded in the sysplex
couple data set. In the example described, this condition would persist
for one hour.
Furthermore, it is common practice for many MVS applications to record time stamps
on external media. For example, time stamps might be written to a
data set containing a database log. These logs could be used by several
processes, such as database recovery, which depend critically on progressively
increasing time stamps to preserve the ordering of events. Setting
the TOD clock backward could make the log useless for database recovery.
- Setting the TOD Clock Forward
As part of its status
monitoring, MVS periodically
checks the status of each system in the sysplex. To determine system
failure, MVS uses the failure
detection interval value specified in the COUPLExx parmlib member.
If you set the TOD clock ahead, then the difference between a time
stamp previously recorded in the couple data set and a time stamp
obtained from the reset TOD clock could exceed the specified failure
detection interval. In this case, MVS would initiate a “status update missing”
condition, which in turn would cause the other system(s) to be removed
from the sysplex (that is, placed in a non-restartable wait state).
This would leave only the system with the future TOD clock in the
system or sysplex.