Re: Is it possible to prevent a structure into a particular CF (Coupling Facility)?

Hi Jan,
I'll try to answer your questions but I can't cover all aspects. Just give
some general ideas.
The CF Hint functionality is for duplexed structures and is designed to
ensure that the
instance at SITE2 is used in the case of a SITE1 disaster.
This functionality is available in GDPS/HM V3.3.

The question on where you place CF structures and whether they are duplexed
is very
dependent on your environment and business requirements. There are some
recommendations which would be true:
Have XCF signalling structures at both site1 and site2
Have ISGLOCK at site1 but have the KSYS LPAR weighted higher so that
it can be rebuilt at site2. (This area is complex and there are many

The KSYS does have to be part of the production parallel sysplex and the
KSYS disk
must be at site 2 for a HyperSwap implementation.

If you are already doing CF duplexing at the production site you would not
want to move the second copy to site2 for performance reasons. If you have
a single
site workload (all prod systems at site1) and there is a site1 failure you
will have to
IPL your systems at site 2 anyway and so duplexing across sites will not
buy much.

In fact, one way to implement GDPS/HM would be to have the KSYS LPAR at
site1 and
the KSYS disk at site2. All CF structures at site1 with the site2 CFs
defined in the CFRM
policy for use in the event of a site1 failure. This negates the need for
cross-site CF
and Timer links. The disadvantage is that KSYS would have to be IPL'd at
site2 first
before starting the recovery in the case of a site1 failure and adds about
20 minutes
to the RTO. There would be no difference for site1 disk failure. KSYS would
as its disk is at site2 and perform the HyperSwap.

We can take further discussion off-list.

George Kozakos

For IBM-MAIN subscribe / signoff / archive access instructions,
send email to listserv@xxxxxxxxxxx with the message: GET IBM-MAIN INFO
Search the archives at