HPUX - Reboot for New Timezone?

[Ralph Burnette]

PHCO_34668 installes a new tztab file and does not require a reboot.

However, is a reboot required for the new tztab file to take effect? Why or why not?

I have seen and heard conflicting statements on this.

[James R. Ferguson · Accepted solution]

Hi Ralph:

In order for the new rules to be honored, existing processes must be restarted (most notably, the 'cron' daemon). Thus, after applying the patch, I'd plan a reboot sometime *before* the Spring 2007 time transition.

Regards!

...JRF...

[Steven E. Protter]

Shalom,

Though I know of cases where HP-UX servers were run 5 years non-stop thats a bad idea.

Run security_patch_check, build some a security update depot and plan a reboot.

SEP

[Bill Hassell]

tztab is consulted with every request for the time. That's why you can run Clay's dst.pl script to test the daylight saving switch. Run it before you add the patch, then right afterward and you'll see the change for -y 2007. In fact, you can add your own timezone entries in tztab at any time. Just be sure you track your changes so they can be ported to a new version (ie, patch). Now if you haven't planned a reboot before next January, this is a good time to schedule a reboot sometime in the near future. The actual changes will not affect US timezones until next year.

[Ralph Burnette]

Bill,

I understand from your posting that a reboot is NOT necessary but good system management would dictate that a system be rebooted on a regular basis.

The reason that the question of rebooting is so important is that I have 38 servers all of which require change control processing and negotiated outages. Plus add year-end processing, etc. to the mix and some real scheduling issues arise.

Hence my concern about rebooting the servers.

BTW, do you know where I can get Clay's dst.pl script?

Thanks!

[Calandrello]

Friend boot in this in case that it would be interesting to only validate if would not have no problem of alteration of horario in reboot.

[James R. Ferguson]

Hi (again) Ralph:

As I noted in my original post, you *should* plan a reboot of your servers sometime after applying the patch and *before* the Spring 2007 transition point.

Clay's script can be found here:

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=1012307&admit=-682735245+1161176297085+28353475

Regards!

...JRF...

[A. Clay Stephenson]

Actually, I don't think Bill is saying that periodic reboots are a good thing; I certainly don't. The only time my boxes get rebooted is when periodic patches are applied. It's a good idea to schedule a reboot before the actual tztab changes take effect only because long-running processes will not re-consult /usr/lib/tztab after doing an initial localtime() or similar functions. Oriinally, each localtime() or strftime() did re-read tztab but that resulted in a performance penalty (at the expense of missing possible changes to tztab) but later patches cache that lookup so that once read, a given process will not re-read a tztab entry. If you know about all the long running processes (the cron daemon being the most obvious) and can kill and restart them then there is no need to reboot BUT it's difficult to know all of these processes especially across multiple boxes so the least evil and most reliable approach is to schedule a reboot sometime before the actual time change revisions take effect.

[Rita C Workman]

Hi Ralph,

You have the answer(s) above, but couldn't resist responding when I saw who it was.

Quit complaing about rebooting 38 servers...you know you love the thrill of negotiating the downtime, organizing the schedule and documenting the tasks to be done to accomplish it all !

I'm getting mine done this weekend....reboots and all.

Regards Sir !
Rita

[Bill Hassell]

Clay clarified my comment...periodic reboots are never necessary with HP-UX *except* for your regular patching schedule (you have a patch schedule, correct?). The only time regular (non-patching) reboots are necesary are when patches are never applied or bad applications are allowed to run, or both. Since there is plenty of time between now and next spring to schedule your patch job, that will take care of the reboot. And with a lot of production servers, you should have one of more test servers to checkout patches before adding them.

[Ralph Burnette]

To everyone -

Thank you very much for the clarifications and advice. I wanted to 1) be sure and 2) be able to explain it to others.

Special note to Rita: I wasn't complainging, I was bragging!!! :-)

[Ralph Burnette]

We will be scheduling reboots!

Thanks!