Servers using PIPE DELAY need to be restarted, an IPL CMS is required.
One can not only suffer from wrong wakeup times, but -depending on the REXX
code- suffer from a CPU loop: my server wanted to wakeup at 23:59 (and stop
the pipeline then too).
Here some comments from my ACCOUNT EXEC
******* BEWARE ********************************************************
* At least up to z/VM 4.2, the DELAY stage cannot cope with dynamic *
* timezone changes, e.g. a DELAY to wakeup at 23:58:00 would cause a *
* wakeup at 22:58:00 when the timezone changed from Summer to Winter. *
* Result is a loop -every day- during the hour of TZ change. *
* The only solution is to re-IPL CMS (NUCXDROP PIPE doesn't help) *
* We try to solve that problem too. *
***********************************************************************
I programmed around this problem. When waking up: check the time it is
and compare with the expected time
In the PIPE:
,/* Do stop at midnight to close files and cleanup ------------------*/
'?Literal 23:59:00|Delay', /* To stop around midnight */
'|VAR MIDNIGHT',
'|SPEC 1-* 1 /: passing midnight../ N|cons',
'|STOP: FaninAny',
'|PipeStop',
....
when symbol('MIDNIGHT')='VAR' then call Daily_Clean
....
/*-----------------------------------------------------------------*/
Daily_Clean: /* Close current file, format the recs and open a new one*/
/*-----------------------------------------------------------------*/
parse value date('S') time() with today now
if left(Midnight,2)<>left(now,2) then do
/* Delay's time does not match CP's time */
Say 'There has been a dynamic timezone change... Must IPL'
'CP IPL CMS PARM AUTOCR'
end
say today now 'Daily Cleanup started'
.....
2011/11/8 Paul Gilmartin <[email protected]>
> I have an event that I intended to occur daily at 12:30:
>
> 'literal 36:30 | dup * | ... delay | ...'
>
> Last week it was happening at 12:30. This week, it is
> happening at 11:30. It's easy to suspect the Daylight
> Time change. The event timestamps its output with the
> Rexx time() function, which shows 11:30, so it appears
> that the system local time offset is correct.
>
> I have not restarted the pipeline for several weeks.
> Is it possible that Pipelines fetches the clock offset
> once, when the pipeline starts, and never again?
>
> Or might this depend on how ops set the local time?
>
> Thanks,
> gil
>
--
Kris Buelens,
IBM Belgium, VM customer support
