Ed Jaffe wrote:
<snip>
I never liked defining secondary of zero. I realize it's IBM's recommendation, but it all but ensures a compress will be performed when the library fills up. I've had systems crash either during or after such a compress -- sometimes with disastrous consequences!
<snip>

Whether it's "IBM's" recommendation depends on who you talk to (smile). I believe Peter and I have disagreed about this for about 25 years.

There are two possible problems when using secondary extents for link list data saets, exceeding the link list extent limit the topic of this thread, everyone's favorite ABEND106-F RC40. If your link list is not especially long, hitting the 255 limit will probably not happen; you can count extents before IPL after putting on PTFs and if necessary compress or reallocate. In the meantime, you will have many fewer x37 abends while putting on PTFs. This should be an informed choice, in my view, rather than a blanket recommendation. But see below.*

If you don't update running systems, the second problem will never occur.

Likewise, whether it's IBM's recommendation to never update a copy of system software that is in use depends on who you talk to and whether DYNACT is specified for a particular PTF. Here there be tygers, and my recommendation is never to do that unless you clearly and thoroughly scope out the probable result first and are prepared to IPL if you miss something (which is easy to do) and the change goes pear-shaped.

*I expect most people to move to thin provisioned volumes over time because the business case is compelling. On TP volumes, there is no reason not to use very large primary extents, which can obviate any advantage to secondary space allocation for system software volumes. "Just let the disk controller manage the actual real estate" is my (new) recommendation for system software volumes.

--
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

Reply via email to