On Sun, 2009-02-08 at 21:34 -0600, Paul Gilmartin wrote: > On Sun, 8 Feb 2009 22:21:12 -0500, David Andrews wrote: > > > >TOD clocks drift. Those of us without sysplex timers *have* to rely > >upon SET CLOCK, which in turn sets CVTLDTO to an offset that doesn't > >necessarily fall on a conventional TZ boundary. > > > Doesn't this leave the critical GMT incorrect?
At LPAR activation we set the TOD to GMT. Initially this and CLOCKxx (and individual TZ settings) is good enough, but TOD drift eventually knocks us off real GMT. We periodically SET CLOCK to locally correct for this. (I understand that this sets CVTLDTO.) > Mightn't this cause the Unix time to go backwards for some > values of the adjustment. Yeppers, it sure could, which is why we do this in quiet periods -- one of the advantages that we little fish have over the megacorporate 24/7 behemoths. While I'm wishing, I wish the sysplex timers weren't so expensive. > What if instead of the complexity you propose one simply fudged the > content of CVTLSO, not CVTLDTO, 'splainey? I'm not familiar with CVTLSO, and the CVT DSECT only says: "LEAP SECOND OFFSET IN TOD FORMAT" -- David Andrews A. Duda and Sons, Inc. [email protected] ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
