On Thu, 16 Oct 2008 09:15:07, Chase, John <[EMAIL PROTECTED]> wrote: >Are there any "compelling" reasons to place a shared PARMLIB on a volume >that holds a sysplex couple dataset?
Only if you haven't the DASD to put it anywhere else. >Are there any "compelling" reasons NOT to place it there? As already mentioned, reserves & I/O contention can send the sysplex casters- up. All arguments to the contrary, we unfortunately do have a small, shared PARMLIB on one of our sysplex couple volumes. We tend to be less than orthodox here, anyway, but allow me to explain... We are what IBM affectionately calls a "BronzePlex", split across two sites. The only shared DASD we have between our sysplex images are the couple volumes. Our Storage folks claim the gave us all the spare custom volumes they had on globally shared control units (LSS') to carve out the six couple volumes for this BronzePlex. (We have recently "right-sized" our DASD configuration.) Only members that must be common to the sysplex are in the shared PARMLIB - IEASYM, IEASYS (PlexCFG, GRS, and Couple and RNL poitners), COUPLE, and GRSRNL. "Must" is a strong word, but better to maintain one copy of these members, particularly Couple and GRSRNL, than forget to update a carbon copy after making a change. All other members are taken from system-specific PARMLIB's. We do not back up the PARMLIB nor the couple volumes. We spread our primaries, alternates, and spares across the six volumes, three in each building, to minimize contention and maximize availability. We tried to pick a couple volume for PARMLIB that would be less contentious (mostly alternates and spares, and if a primary, a low-use one like SFM or WLM). Though I am sure others here can point out examples if they do in fact exist, I have yet to see a z/OS component or ISV product that puts a RESERVE on a PARMLIB volume (other than DSS/FDR/etc). Note that we CONVERT SPFEDIT. Because we only keep those four members in PARMLIB, I/O is brief, and limited to IPL and SET GRSRNL=xx. I realize the risks, and I realize some folks' heads are probably spinning, so I absolutely do not recommend anyone put PARMLIB on a couple volume. We do it here because we painted ourselves into a corner. Our IODF volumes are shared, but only within a site. I'm trying to convince line management to take the next step into shared DASD (moving toward what IBM terms a "GoldPlex"), but it's a tough sell right now (just read the papers). Once we get there, PARMLIB will be moved off. BTW, "BronzePlex" and "GoldPlex" (as well as "PlatinumPlex") are terms taken from IBM's Redpaper "z/OS Systems Programmers Guide to: Sysplex Aggregation". Regards, Art Gutowski Ford Motor Company ITInfrastructure ---------------------------------------------------------------------- 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
