What's magic? Some missionaries were confronted by a group of hostile locals. Things were looking grim. One quick thinker pulled out a lighter, held it high, and flicked it on. Animosity melted into awestruck admiration...It was the first time they had ever seen a Zippo light on the first try.
When I joined SCE in the mid-90s, tape was already being shared between data centers. There was production in both sites. Tape was bus-and-tag, yet every drive in both centers was usable by every LPAR in both centers. In order to manage the drives, MIA was employed to control online/offline on demand. MIA required control data sets shared by all LPARs in both centers. The tape technology was STK, whose HSC software also required shared control files. Channel extension was used for both tape and control data sets. Technology progressed. First to ESCON, then to FICON. Channel extension moved to DWDM. The number of tape drives increased from a handful of physical units in the 80s/90s to a *huge* number of virtual ones. Still we kept MIA because--why not. Tape--now Oracle--still need shared control data sets. A tiny handful of disk volumes held all and only these shared control data sets and the ICF catalog that described them. Today we are moving to (now) Dell DLm tape. All virtual, all managed by RMM and for the first time OAM. We have so many virtual drives that we can segregate by sysplex. No more need for MIA. OAM is a different animal that does need shared data sets. Again, I would not advocate sharing catalogs outside a sysplex, but it's not a show stopper either. . . J.O.Skip Robinson Southern California Edison Company Electric Dragon Team Paddler SHARE MVS Program Co-Manager 323-715-0595 Mobile 626-543-6132 Office ⇐=== NEW robin...@sce.com -----Original Message----- From: IBM Mainframe Discussion List <IBM-MAIN@LISTSERV.UA.EDU> On Behalf Of R.S. Sent: Wednesday, May 27, 2020 4:10 AM To: IBM-MAIN@LISTSERV.UA.EDU Subject: (External):Re: Opinions/experience on sharing catalogs outside plex CAUTION EXTERNAL EMAIL Not magic, but also not STK/Sun/Oracle. STK tape world is significantly different from IBM tapes. There are also other players (EMC DLM, etc.) which follow way similar to IBM, but all of them are defined as MTL (Manual Tape Library). Few differences between STK and IBM: IBM provide all necessary software components in z/OS base. No need to add/buy/install anything else. STK require its own software suite, many components are in the suite. Some of them depend on configuration. Communication host-ATL. Control path for IBM is embedded in data path (FICON). For STK ATL the control path is over IP. For older STK ATL there was coax connection required (big pain nowadays!) which meant it was like coax-attached terminal. However communication to VSM (STK virtual tape) is over FICON. Virtual tapes. Data are stored on some disk array, however VTS emulate whole library, so theoretically you can move your VTS and attach it to installed from sratch host and you can access data. In STK world data are still on disk (SVAA), but virtual volume definitions are kept on the host. So, you need working STK software *and* proper control data set to access data. Because of differences DFSMS recognize IBM ATL and that is important for ACS routines, storage classes, etc. Everything is in DFSMS. For STK most definitions are in SMC/HSC software components configuration. From system point of view usually you define some esoterics in IODF. -- Radoslaw Skorupka Lodz, Poland W dniu 27.05.2020 o 03:32, Ken Smith pisze: > *totally different technology...* > > Magic? > > On Tue, May 26, 2020 at 5:11 PM Jesse 1 Robinson > <jesse1.robin...@sce.com> > wrote: > >> I'm not advocating this practice. Just saying that we did it for 25 >> years without a failure. We're now moving to totally different >> technology where such action is not even in the picture. >> >> . >> . >> J.O.Skip Robinson >> Southern California Edison Company >> Electric Dragon Team Paddler >> SHARE MVS Program Co-Manager >> 323-715-0595 Mobile >> 626-543-6132 Office ⇐=== NEW >> robin...@sce.com >> >> -----Original Message----- >> From: IBM Mainframe Discussion List <IBM-MAIN@LISTSERV.UA.EDU> On >> Behalf Of R.S. >> Sent: Monday, May 25, 2020 5:53 AM >> To: IBM-MAIN@LISTSERV.UA.EDU >> Subject: (External):Re: Opinions/experience on sharing catalogs >> outside plex >> >> CAUTION EXTERNAL EMAIL >> >> Skip, >> In the past I implemented STK tapes, the largest tape system in >> Poland at the time. Interesting job. :-) It seems, you shared control >> datasets between datacenters. Assuming your second datacenter is for >> disaster recovery, you had single point of failure. Catalog is not >> important there, but availability of datasets is. How could you work >> with tapes in case the datasets are lost due to catastrophe of primary DC? >> IMHO the only way was to have remote copy of control datasets. Last, >> but not least, as far as I remember the datasets were very specific - >> they have >> (had?) hardcoded both volume labels and device numbers. While remote >> copy replicate volume label, the dev num is IODF dependend. There was >> a method for that, I forgot details. Other method could be a trick >> with duplicate device numbers. >> >> -- >> Radoslaw Skorupka >> Lodz, Poland >> >> >> >> >> >> >> W dniu 24.05.2020 o 21:29, Jesse 1 Robinson pisze: >>> Until recently, we shared a catalog not only across sysplexes but >>> between data centers. All because of tape. We had STK virtual tape >>> (in both data centers) supported by MIA (Multi Image Allocation). >>> These products require control data sets shared among all exploiting >>> systems. We could have managed with uncataloged data sets, but that >>> was deemed riskier in the long run than a shared catalog. The only >>> entries in the catalog were for tape management data sets. We never >>> had a catastrophe. 😉 >>> >>> . >>> . >>> J.O.Skip Robinson >>> Southern California Edison Company >>> Electric Dragon Team Paddler >>> SHARE MVS Program Co-Manager >>> 323-715-0595 Mobile >>> 626-543-6132 Office ⇐=== NEW >>> robin...@sce.com >>> >>> -----Original Message----- >>> From: IBM Mainframe Discussion List <IBM-MAIN@LISTSERV.UA.EDU> On >> Behalf Of R.S. >>> Sent: Monday, April 20, 2020 7:42 AM >>> To: IBM-MAIN@LISTSERV.UA.EDU >>> Subject: (External):Re: Opinions/experience on sharing catalogs >>> outside plex >>> >>> CAUTION EXTERNAL EMAIL >>> >>> Well, it is not good idea, but sometimes even such idea is better >>> than >> nothing. >>> What's important the risk covers THIS catalog only, not whole world. >>> >>> And yes, this catalog and shared datasets should be shared without >> "sysplex features" like changes is serialization. RESERVE should be >> used here or CA-MIM should be used, but the last one is add-on tool. >>> BTW: BCS can be defined with SHR(3 4) or SHR(3 3). For this case it >>> has >> to be SHR(3 4). AFAIK it is alterable. >>> Again: small activity is your friend here. Small number of datasets >> cataloged in the BCS is good here. Potential problems with the BCS >> will not affect other BCSes. >>> I use it for years (with limited activity). Mostly PS files and some >> VSAM. No problems observed. >>> Caution: PDSE *will* break despite of way how catalog is shared. No >>> help >> from CA-MIM, AFAIK. Observed many times educated many guys who used >> PDSE for sharing. >>> -- >>> Radoslaw Skorupka >>> Lodz, Poland >>> >>> >>> >>> >>> >>> >>> W dniu 20.04.2020 o 14:45, Allan Staller pisze: >>>> Yes, it can be done. I reiterate, IMO, this is most likely not a >>>> good >> idea. >>>> In order to accomplish this safely, you also need to regress GRS >>>> to >> pre-GRS functionality. >>>> Everything affecting this catalog must be handled >>>> w/Reserve/Release, and not normal processing VSAM Sharoptions for >>>> the catalog need to be >> changed. IIRC when I "undid" this the catalog hand Shareoptions (2,3) >> (or was it 4,3?). >>>> This option tells z/OS that the user is responsible for Catalog >> seriailization. >>>> SYSDSN, SYSIGGV2, SYSVTOC, SYSZVSAM (?) and the SPF* queues need to >>>> be >> excluded from GRS processing. >>>> In my case that led to various deadly embraces that usually led to >> manual intervention. >>>> My $0.02 USD on this is: Why point the shotgun at your foot? >>>> >>>> -----Original Message----- >>>> From: IBM Mainframe Discussion List <IBM-MAIN@LISTSERV.UA.EDU> On >> Behalf Of R.S. >>>> Sent: Monday, April 20, 2020 3:45 AM >>>> To: IBM-MAIN@LISTSERV.UA.EDU >>>> Subject: Re: Opinions/experience on sharing catalogs outside plex >>>> >>>> [CAUTION: This Email is from outside the Organization. Do not click >>>> links or open attachments unless you trust the sender.] >>>> >>>> W dniu 09.04.2020 o 02:11, Rob Schramm pisze: >>>>> I am considering sharing some usercats outside of a sysplex. What >>>>> I can find is that sysiggv2 must be kept as a reserve to do so. >>>>> >>>>> Looking for others that have had to do this. >>>>> >>>>> One question I had was, what happens on a ispf 3.4 when the data >>>>> set is part of the catalog but exists in another system? Ief238d? >>>> My €0.02 >>>> >>>> 1. You can share catalogs between sysplexes. Note: we mean BCS, >>>> which >> is usually called catalog. >>>> 2. The less activity on the BCS the better. >>>> 3. The above means: >>>> 3.1. Avoid keeping non-shared datasets in the BCS. Use another BCS >>>> for >> that. >>>> 3.2. It is not bad idea to have multiple "small" shared BCSes. >>>> 4. You cannot use any sophisticated catalog sharing features like >>>> RLS >> or ECS. >>>> Regarding you last question: I understand it as you have entry in >>>> the >> BCS, but the dataset reside on volume which is not share, that mean >> it is unavailable for one system. It's nothing exotic. It's like >> orphan catalog entry, which sometimes may happen even without BCS >> sharing (usually as result of human error). >>>> However that also means the sharing is not done correctly. The best >> scenario is when all datasets cataloged in shared BCS reside on >> volumes which are also shared. Preferably the BCS is also on the >> volume from that group. >>>> Keep it simple. >>>> >>>> -- >>>> Radoslaw Skorupka >>>> Lodz, Poland >> ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
