On Saturday, 07/21/2012 at 03:39 EDT, Rob van der Heij <[email protected]> 
wrote:
> Ever since my presentation in 2005, I find customers who were told by
> their friendly IBMer not to get STP/ETR on the machine because "it's
> not supported by z/VM"  The STP/ETR feature is a priced feature that
> is best negotiated into the deal early rather than acquired separately
> later. When the LPAR TOD is kept in synch with the world, by pure
> magic CP, CMS as well as Linux guests will see accurate time. There is
> no need for ntpd in Linux (in fact, doing that will make things
> worse).
>
> The only disruptive time shift would be when you POR with the feature
> on. Since the LPAR also picks up the actual time at activation, there
> should not be a jump. And STP/ETR will nicely speed up TOD increments
> to make it match real time. Except when the leap second meets a
> software bug and takes the system down :-)

While one can get this information from the published IBM Redbooks on STP 
and the Principles of Operations, I will summarize the best I can. 
Warning: My terminology is sloppy in the interests of understandability.
- If you have STP enabled and configured, TOD clock steering will occur. 
It doesn't matter if the OS "supports" STP or not.
- On modern machines, STP steers the TOD at a rate of approximately 1 
second in 7 hours  (ETR takes 11 hours).
- That means it will take about 7 hours for QUERY TIME to match the 
external time source after a leap second has been added.
- Good news, that means QUERY TIME will be correct when you come in a 8am 
(leap seconds are added at 11:59:60).
- If your Linux guests can't live within that tolerance, then you need to 
consider ntp for Linux.
- When the difference between the LPAR TOD and the time source is +/- 50 ยต
s (model dependent), an STP-sync-check interrupt is raised.
- When I/O timestamps are being used, CP uses the STP-sync-check interrupt 
to re-calculate the correct time adjustments to apply to the TOD used in 
the I/O.
- The oscillators are extremely stable, so it doesn't take much pull on 
the wheel for STP to steer out the drift, as it were.

Alan Altmark

Senior Managing z/VM and Linux Consultant
IBM System Lab Services and Training 
ibm.com/systems/services/labservices 
office: 607.429.3323
mobile; 607.321.7556
[email protected]
IBM Endicott

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
----------------------------------------------------------------------
For more information on Linux on System z, visit
http://wiki.linuxvm.org/

Reply via email to