changing the time on an hp-ux system

[EJ Stremler]

Just want to verify this, some of the testers want to test the DST patch where they want to change the time an hour ahead and then back to the right time. Would I just use the date command to accomplsish this? I would think the only things that might get affected is cron or at.. i also think you can change the time without a reboot.. Please let me know i am correct..thanks, Ed

[OldSchool]

there is a perl script floating around here to test this.

if you insist on changing the clock and are runnning the databases, make sure they are down during the test period.

I *DONT'T* recommend this approach, esp as the patch is just the tztab file (which has been out for a year by the way)

[Pete Randall]

There have also been several threads claiming that the patch does not work when changing back. This generally seems to be because the time was set to an "ambiguous" time (between 1:00 and 2:00AM) - you need to set it earlier than 1:00. Your best bet is to search out A. Clay Stephenson's perl script that will test out the time changes for you. Just put "dst" into the search box at the top of the page and click on the arrows.


Pete

[Steven E. Protter]

Shalom,

Oracle and other products tend to crash when system time is set back.

At the least their internal time database gets messed up as does the timestamps of files on the system.

If you must do tests like this, you may wish to back up everything and Ignite the system so you can restore the system after it becomes confused and unhappy.

SEP

[Pete Randall]

Here's a link with Clay's script:

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=1007176


Pete

[A. Clay Stephenson]

I'll answer your question this way. Why do you think I wrote the Perl script? Bear in mind that a UNIX box might have connections from all over the planet (and beyond) with separate and distinct TZ settings for each of them. How would you test these? The test is trivially simple with dst.pl and can be done with as many TZ's as you like. Changing the time (expecially backwards) can be extremely sensitive and can cause many unexpected problems. The most obvious are database timefield but make dependencies can also be destroyed. Because the dst.pl Perl scripts doesn't know anything about tztab, it is a rigorous test of the time transitions. The only thing that won't be tested are home-grown "cowboy" time routines that are used in place of the standard libc routines --- and if you have "cowboys" then both the developers and the functions should be replaced with their more standard UNIX counterparts.