Minutes of the GSE UK IMS Working Group Meeting - South Bank, 3^rd December 1998

Chairman: Frank Fleming (Frank.Fleming@barclays.co.uk)

Secretary: Peter Armstrong (peter_armstrong@exchange.bmc.com) - 01276-419835

Administration

• Introduction
  • Welcome to all and a run through the agenda. Last time we asked for suggested topics and people have asked for IMS maintenance strategies, which we will cover at the next meeting.

• Minutes
  • The minutes were unfortunately not available on the Web (normally to be found at http://web.ukonline.co.uk/dougie.lawson/gse) due to Dougie's PC having a severe case of mal-de-hard-drive. Everybody had received - many thanks to John Boyle for standing in as secretary.

• Matters arising
  • None.

• New Members / attendees
  • None.

• Apologies
  • IBM ISC Hursley, Cap Gemini, American Express, Guardian Insurance, Nationwide Building Society

• Correspondence
  • Flyers for the GSE Management Summit on OS/390 - 25/26 January in Berlin.

• General
  • Please update the signing sheet with email addresses etc.

  • Held annual Steering Committee meeting yesterday at which all members of the Committee confirmed that they could carry on. However, the DBCTL sub-group needs a new leader as Ian Tyson now "works" for IBM and is not so involved in DBCTL as he used to be. Volunteers please talk to Ian. The DBCTL sub-group wants to carry on for another year at least.

  • We want to distribute the minutes via Web and email in the future. The default will be email, with a copy on the web at Dougie’s website. People who cannot access them or receive them electronically will be able to request a boring old hardcopy from Andrew.

  • Meetings next year will be:

    March 4th- Birmingham - DBCTL

    April 9th- South Bank - IMS

    July 1st- Manchester - Joint with DBCTL

    October 28th- South Bank - IMS

IBM What’s New – Alan "I hate computers" Cooper (IBM UK)

• CA37 - the Fast Path workshop - has been updated to Version 6 and is now CM41U. Version 6 education is now complete. The APPC class has been upgraded. Are there any courses on DL/1 Programming? No classroom, but lots of CBT. Pete and/or Alan will teach anywhere - just ask (Peter Armstrong had a request from Belgium). Comment from Frank - look at www.uk.ibm.com/education.

• Working on republishing the IMS Primer as a red book.

• IMS and DB2 Technical Conferences are in Paris from 12-16 April and San Diego week of August 16^th.

• New product come out - IMS Parallel Change Accumulate - 5697-E32. Don't know how it works yet or how it interfaces with DBRC.

• Last Thursday was 400 days before Y2K and there were 3 major outages in Europe - lost DASD and cache. Do your procedures cater for this?

• Partitioning Version 2 with new product number and partitioned indexes.

• Performance Analyser - combination of PARS and ASAP.

• New tools announced this year - see later.

• V6 enhancements:
  1. 1. Can say how much expansion you want to allow in the COMPRTN. (V5 and V6).
     2. New callable interface for OTMA - currently C and C++, others later
     3. VTAM Generic Resources - define who manages - IMS or VTAM; and improved ESTAE
     4. IMSID on BMPs. New group parameter - this is designed for restart, not cleanup. There is an APAR suggesting that the group name should be the most commonly used IMSID. Consensus was that this is not a good idea. General discussion on lack of workload balancing for batch.
     5. What is used for security checking on CHNG or AUTH call in non-message driven BMP changes with PTF level.
• Documentation:
  1. 1. New red book coming out on shared queues courtesy of Pete Sadler
     2. V6 Guide - you must have the 01 level at least
     3. IMS Quarterly - available via www.ibm.com/ims

A copy of the foils was distributed.

Data Sharing Experiences - Diane Seed (Barclays Bank)

• The project was started about 3 years ago. Diane reviewed why they went to sharing - filling up largest available machine and wanted better availability, an then went on to discuss "The Big One", which was Barclays' biggest change in 15 years. The project had heavy involvement from IBM and there were 20 people on the project team.
  • • IRLM needs SHAREOPTIONS(3 3), even if not sharing or you are at SHRLEVEL(1) in RECON.
    • You get CPU Overhead as soon as you go to sharing (which in Barclays' case was for their batch throughput, but was also there for online)
    • Conversion of MSDBs to VSO DEDBs was not nearly as transparent as expected. Big problems with locking.
    • Lots of new function APARs on IRLM 2.1 at present - see later
    • BMC EP level of products needed for support of Shared SDEPs

• Started IRLM in global mode with SHRLEVEL(1) and CF structures. Then went to SHRLEVEL(3) - moves to CF and uses IRLM for everything. Have CF very big, but causes no performance problems.

• Lots of problems, because Barclays was one of the first to do it.

diane.seed@barclays.co.uk

A copy of the foils was distributed.

Shared Queues - Pete Sadler (IBM UK)

• Pete has been out to a customer site in Australia to play with shared queues and write a red book on the subject (SG24-5257), which should be ready in the summer.

• You don’t have to use VTAM generic resources with SQ but it helps. You must allocate EMH structure if you have the FOPCTRL macro, even if you are only using DEDBs.

• The Coupling Facility is an MVS thing, not an IMS thing, so you need to talk to the MVS people. Dougie also pointed out that you need lists of XCF groups e.g. 20. General discussion on which libraries should be shared e.g. FMT.

• In V6 for all conversations, SPA is stored on the queue for inbound and outbound even if not using shared queues. Therefore allow for conversations when sizing structures, or if not using shared queues size your MQ datasets correctly.

• SERIAL – is within one IMS. If you want true SERIAL, run one region on in the SQ group. MSC uses SQ if relevant. N.B. MSC across a link = WADS I/O. COLD Q – get messages off via tool, structure delete or write own code.

• SC26-9157 is new name for CQS Guide and Reference. (No longer a licensed pub.)

Pete handed out copies of the foils and then amused us all by blowing down a piece of tree trunk which produced some very rude noises (well, we think that’s where the noises came from!?)

Problems forum, hints and tips, contacts for lunch

• Requirements – none.

• Name: Chris Huffington – Friends Provident – 01306-653426
  Problem: Is anyone using Work Load Manager? Are you getting transaction volumes reported correctly?
  Contacts/Comments: Dougie – bug in WLM code.

• Name: Chris Huffington – Friends Provident – 01306-653426
  Problem: Is anyone using the Spool API? If so, have you hit any problems?
  Contacts/Comments: Dougie – JES2 ignores record length and makes 32760. Look at hints and tips. Toyota?

• Name: Nicola Diamond – Bank of America – 0181-760-6488
  Problem: We recently experienced an MVS storage shortage which resulted in an abend of the IMS control region. The DLI and DBRC regions remained active but were assumed to have abended. The subsequent emergency restart of IMS started a second DBRC and DLI region resulting in allocation errors against all databases (already in use by the previous DLI) amongst other errors. It was some time before the errors were diagnosed, and the duplicate DLI and DBRC regions detected, resulting in significant downtime. IBM were contacted to explain why a start/restart of an IMS (control region) would not detect an existing DLI/DBRC. The product is apparently working as designed, and we have now put automation in place to alert to any reoccurrence. IMS GSE users should be aware of this exposure if they are not already. Have any other IMS GSE users experienced the same situation? Would IMS GSE users support an Enhancement Request / fix to encourage the closure of this apparent loophole?
  Contacts/Comments: Frank has seen this as well. Raise as a requirement.

• Name: Neil Price – TNT Express Worldwide – 01827-710702
  Problem: 1.(PMR currently open with I BM). One of our CICS regions had to be FORCEd out after some serious problems. IMS didn’t clean up the CICS connection so wouldn’t allow CICS to reconnect following restart. IMS restart was required to clean up the old connection. Possibly the problem is addressed by UQ13668 and UQ15364.
  2.DB2 failed and restarted, but IMS failed to reconnect. This was due to a BMP using DB2, which had not issued a DB2 call or a checkpoint since the DB2 failure. (This can also happen with IFP or MPP WFI regions). The BMP had to be terminated with /STO REG xxx ABDUMP.
  Contacts/Comments:

• Dougie pointed out a whole bunch of IRLM new function – search on "569516401 NF" and there are 34 hits including PN92333, PQ01040, PQ12126.

• Smartmvs is moving to service.software.ibm.com/390holddata.html – gives you PEs and HIPERs.

LUNCH

DL/2 – Stan Hoey (Circle)

• Circle are based in Brighton and are best known for CICS education. DL/2 is designed to migrate IMS to DB2 without changing code. (Copies of the slides to be sent to Dougie). You migrate database by database. Set of utilities provided to migrate data, relink load modules etc.

• Supports SIs, LRs etc. Currently working on / testing a ten-level BOMP LR at one customer. Obviously does not replace analysis and testing phases, but it severely reduces/eliminates the program rewrite.

• Does DL/2 enforce NOTWIN? Yes.

• Restrictions – see foils.

Stan@circle-dl2.com
www.circle-group.com

IMS Tools – Alan Cooper (IBM)

• Alan Cooper ran through a review of all the IMS Tools available from IBM.
  • • IMS DB Control Suite – integration point for all tools. Front-end to anybody’s tools
    • Recovery Saver – Timestamp recovery to any point-in-time by cutting logs and playing with a copy of the RECONs
    • DBRC Security Tool – RACF protection for the RECONs and command level security.
    • Year 2000 Exit Tool – gets called twice. First between call and IMS and second between IMS and application. V5 and V6 – have to reorg to implement for a DB.
    • Ad-hoc reporting tool – quick, simple access ./ fix / report etc. on IMS DBs and flat files
    • IMS Index Builder – broadly comparable to BMC’s Secondary Index Utility. No DBRC interface.
    • Partitioning – Partitions or DSGroups. Interim solution – in the future there will be a full solution in IMS with no restrictions. Are NEON and IBM versions the same? Should be the same today.
    • Performance Analyser – Replacement for PARS and ASAP – does not need GPAR.
    • DBT – 12 features, choose the bits you want. Alan went through all the bits and showed some benchmark figures from Tucson Medical Centre which indicate that DBT is "broadly similar in performance" to BMC (see later for Peter’s comments).
    • Compression Extended – makes hardware compression easier to manage.
    • DBICF – ISPF front0-pend to DBRC plus added functions, but has overlap with Control Suite.
    • DEDB Fast Recover – Designed for use after /ERE failure.
    • ADF ToolPak – To move ADF to Y2K. Needs ADF V2.2.
    • BTS – Supports Y2K, MQ, LU6.2, APPC etc.
    • Message Requeuer V3 – Includes SQC support.
    • Workload Router – Probably dying / dead because of Shared Queues.

• Alan then ran his Christmas Quiz. Easier than last year and splendidly silly – winner Frank Fleming.

• Peter then made some quick comments on the Tucson benchmark figures as he had met the lady who produced them whilst at the IMS Tech Conference in Washington. They ran reorg for a partitioned DB and ran the DBTools utilities in parallel by turning DBRC off (DBRC will not allow multiple updating utilities to run in parallel because Authorization is at the DB level not at the partition level). The BMC utilities (which do allow parallelism as they do authorization once at the DB level) were actually run serially, as the customer did not know they could be run in parallel. Peter then showed the results run "properly", which show BMC significantly faster.

A batch problem with GDGs – Philip Nelson (Scottish Widows)

• Phil described a recent problem. They lost power in one machine room. Database is in room 1, and log (disk) is in room 2. Batch job (which uses GDGs for log) abended 8 times, so they kept doing batch backout and restart. But when they went to close the log they found it was empty. The disk manufacturer (not IBM) says it is an IMS or OS/390 problem and IBM says it is a disk problem.

• So, they then ran Timestamp Recovery prior to doi8ng a rerun. However, the discovered that the same log appeared multiple times – the BBO Logs had been overwritten. A message was found in the batch output – IGD7001I GDG ROLL IN ERROR.

• Analysis shows that there was a GDG with 255 entries using DASD, and some of these had been dumped to cartridge (to save space) but they leave the catalog entries on DASD. If you now reach the limit and the volume is offline, then you get the overwrite.

• The solution was to run CA to wend of previous good log, recover to here and rerun batch.

• This was a familiar problem to several people. Moral – watch for these messages!!

• Question on what to do with OPEN PRILOGs – Peter - close using CHANGE.PRILOG, do not DELETE as then you get orphan ALLOCs.

AOB

• Peter then talked about the new version of his DBRC in Practice book, which has been updated to Version 6, Y2K, n-way sharing etc. etc. and he gave copies to everybody. Contact BMC if you want a copy.

• Next meetings: 4^th March for DBCTL and 9^th April for full group.