The Question is:
I have several Java 1.4.1 processes using threads which are periodically put to
sleep. I have found that when the system time is changed by a negative value
e.g. -1 hour, the normal sleep duration of 10 ms becomes 1hour 10ms. It is as
if at the time the s
leep method is invoked the JVM uses a time stamp of when to wakeup, and setting
the clock back has the effect of increasing the sleep duration.
Is this expected behaviour on OpenVMS V7.3-2?Are there system logicals that can
be changed to influence or remove this behaviour?
The Answer is :
The OpenVMS Wizard would use the existing Timezone Differential Factor
(TDF) TDF mechanisms, as described in the OpenVMS Frequently Asked
At its core, this question was extensively covered during the Y2K
era, and this situation is basically the same as the run-time time
problems seen (and documented) during Y2K testing.
As was the case with Y2K, directly resetting the running system time
directly can and has shown various timer-related problems with various
software packages, layered products, and local applications. Even a
one-hour change made directly using the running system time can cause
unexpected application stalls of an hour, or can cause timers scheduled
for an hour in the future to fire early, depending on the direction of
the change. Further, such changes can also cause applications that
utilize the system time for various purposes to fail with duplicate
errors, or to otherwise mis-sort or mislabel the data.
Details on managing the TDF and on time in general on an OpenVMS
system are available in the OpenVMS FAQ.