Jesse 1 Robinson wrote:
As noted earlier in this thread, sysres datasets should be cataloged with
volser ****** and unit 0000. That tells the system to look on the currently
IPLed sysres regardless of volume name.
BTW I don't subscribe to the floating MCAT strategy. Our master catalog lives
away from sysres and therefore does not change on sysres swap. We've done this
forever and don't see any serious problems. And you don't need multiple MCATs
during upgrades as long you always call LINKLIB 'SYS1.LINKLIB'.
<snip>
I prefer Skip's approach, for what it's worth. In my view, the master
catalog can be fully divorced from the software, as is the case for
other operational data sets, and should be. Minor work is needed once
in a while for things you have to add when we add data sets to z/OS and
minor cleanup is needed once in a while for data sets we delete from
z/OS and you're past the point where you will back out to the old
release, but if you keep the same data set names and relative volume
locations this cleanup is generally less work and holds fewer
opportunities for error than creating new master catalogs or cloning
existing master catalogs. The more operational data sets you can leave
as-is during migrations, the faster your migrations are likely to be.
I'll note that shared master catalogs have an upside and a downside.
The upside is that the frequency of failure is lower because one volume
will fail less often than any one of *n* volumes. (This is not quite
1/*n* but it's close enough to that for government work.) The downside
is that the impact of failures or egregious human error is greater,
affecting *n* systems rather than one. Some thought should be given to
sharing boundaries. For instance, you might avoid sharing between
systems that back each other up if you can locate them in different
storage subsystems. Sharing them is not hard given workable naming
conventions for system-specific data sets, such as but of course not
limited to page data sets.
Also, a reasonable alternative to ****** in indirect catalog entries is
using the system-defined symbol &SYSR1. From a readability standpoint I
prefer the latter although they are functionally equivalent. Also, for
anyone still using more than one target volume, you can base symbols for
volumes other than IPL volumes from &SYSR1 using a naming convention to
derive the rest; indirect cataloging works even for multiple volumes
using such derived system symbols.
--
John Eells
IBM Poughkeepsie
[email protected]
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
