Minutes of the GSE UK IMS Working Group Meeting - Manchester, 17^th September 1998

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

Secretary: John Boyle standing in for Peter Armstrong (peter_armstrong@exchange.bmc.com) - 01276-419835 (many thanks John!!)

Administration

• Thanks:
  • Thanks to Ian Tyson for organising the meeting and handling registration.

• Minutes
  • All attendees had received the minutes, which are now also on our Web site. A few attendees had not received the agenda. See Dougie’s web site for details of accounts and other GSE related items like the minutes and the agenda - http://web.ukonline.co.uk/dougie.lawson/gse.

• Matters arising
  • None

• New Members / attendees
  • None

• Correspondence
  • None

• Apologies
  • Neil Price,
  • Peter Armstrong,
  • Dougie Lawson

IBM What’s New – Alan Cooper (IBM UK)

• A separate V6 shared queues class is in preparation. Pete Sadler has been in Australia working with a customer.

• Technical Conferences – Washington, 26 to 29 October (IMS only). European one will be IMS and DB2 in Paris 12 to 16 April 1999.

• STL need customers to test ODBA-MVS address space which can contact DBCTL to get at Dl/1 and DB2. (Need to be on V6?). Contact Alan if interested.

• Y2K issues covered later.

• DB Control Suite for OS/390 available – it is a DBA workbench, where you can plug in your ISV products to a DBRC front-end á la DBICF. More info at next GUIDE.

• DBRC Security Tool available – controls access to RECONs by users and programs.

• IMS Connectors – although these are free on the WWW, they are part of IMS – standard support shipped every 6 months (too often to include in normal IMS shipment).

• IMS Object Connector (Object Programming) enables you to write batch or MPP with no understanding of DL/1. Map "Object" to DL/1 call. Also free – part of IMS.

• HIDAM root pointers – recommended not to use TF and TB any more. Specify NOTWIN.

• New Red Book – SC24-2211-01 Making your IMS Ready for the Year 2000 does not mention IMS V6 – it is all about getting to Version 5.

A copy of the foils was distributed.

WWW.IMS.COM- Ed Jones (Neon)

• SHADOWWEBSERVER and SHADOWDIRECT

• Ed presented on two offerings from Neon aimed at simplifying the process of web-enabling your existing IMS online transactions, and developing new web applications which execute on the mainframe.

• SHADOWWEBSERVER runs as a started task within MVS and translates incoming HTTP requests into pre-defined functions or ‘sections’. A section can return a file (e.g. GIF, DATA$, Binary, Java Applet) run a REXX EXEC (e.g. to access DB2) run a program.

• The product also provides web enablement for existing MFS based online transactions by providing the capability to read in existing MFS source, and generate the equivalent HTML. Existing transactions can then be driven from a web browser, rather than a 3270 terminal or terminal emulation.

A copy of the foils was distributed.

Abend U0999 – Eric Kent (IBM UK)

• Eric discussed a number of ways in which IMS customers can help the Support Centre to help them.
  • Correct dump options
    • For IMS CTL RGN, specify FMTO=D.

  • Check SYS1.PARMLIB members which control MVS dump options:
    • SYSABEND - IEAABD00
    • SYSUDUMP - IEADMP00
    • SYSMDUMP - IEADMR00
    • For SVC dumps – no PARMLIB member, use CHNGDMP command.
      • OPTIONS(CSA, PSA,RGN,SQA,GRSQ,SUM,TRT). Make sure dump datasets are big enough.
  • MVS trace
    • Set on and give it at least 64K table size.

  • IMS Trace
    • You can run the dispatcher and scheduler traces with minimal overhead

  • What to do when problems occur
    • If IMS hangs, dump all the address spaces, using MVS Dump command. For data base problems, e.g. 85X abends, keep IMS logs. These contain X’67FF’ records which contain the dump for the abend.

Problems forum, hints and tips, contacts for lunch

• Requirements – none.

• Name: Andrew Jack – Standard Life Assurance Co – 0131-245-7783
  Problem:We have a problem whereby IMS restart time varies from about 5 minutes to 20 minutes. We seem to have hit a problem described in APAR PN10735 (from 1992) which describes a situation where IMS restart can take an hour to complete due to IMS checking the message queues for errors. This APAR has been closed ‘SUG’; the suggestion being that IMS should report now and again on its restart progress so that you don’t think that it has hung. This has been raise with IBM as we cannot see any justification for IMS having to check message queues for integrity during a WARM restart (following a clean closedown). Is this a problem of interest to other sites? If so, you may wish to add your weight to the request to IBM to review this process.
  Contacts/Comments:FF commented that APARs closed SUG were retained for inclusion in future releases, but unlike GUIDE raised requirements, they have no weighting associated with them. AJ said they had up to 108,500 messages on the queues.

• Name: Andrew Jack – Standard Life Assurance Co – 0131-245-7783
  Problem:Passing CICS transactions to IMS. Standard Life have user code to pass CICS transaction to IMS, but now need to pass IMS transaction to CICS – possibly using ISC.
  Contacts/Comments:Frank Fleming, Alan Lovett.

• Name:Paul Clarke – Guardian Insurance – 01253-662672
  Problem:Martin Christensen, one of our IMS DBAs, would like to contact anyone using DFSMSdss SNAPSHOT.
  Contacts/Comments:Ian Tyson suggested they should talk to Peter, who had presented on IMS concurrent copy options recently at GUIDE.

• Name: Ian Tyson - IBM
  Problem:IMS hang up for 4-5 minutes every 2 hours. Problem traced to BMP, which was run to issue /DIS AREA, to check IOVF utilisation of DEDB AREAs for 30-40 AREAs. IMS ‘goes to sleep’ while these commands are being processed. IT has discovered that there is an alternative format to the /DIS AREA command, which avoids the problem by retrieving the information from IMS without actually processing the commands.
  Contacts/Comments:

• Name: Alan Lovett – EDS – 01253-688311
  Problem:We had reported to us problems of exceptional response times (minutes!). Our initial investigation led us to expect a network problem, as whenever we looked at the system all was well. In the end we found that the periods of poor response time followed log switches. When we compared the timestamps of the DFS3257I log switch message and the RECON log start we found a difference of up to 2 minutes, coinciding with peak demand. Omegamon analysis showed that the RECONs were being locked out by our change accumulation post processing. We run CA after every log switch. We archive to disk RLDS, which we delete after Change Accumulation. To keep DBRC sweet, we issue CHANGE.LOG commands to reset the PRILOG entry to point at the PRISLDS. Our processing used GENJCL.USER processing to do this. To determine the PRILOG start time(s) for the log(s) that were being deleted we used %SELECT ALLOC(PRILOG,ALL) in our skeletal; JCL. When we timed our processing (against a RECON copy of course) we found that this select took 6 and a half minutes. After recovering our composure we replaced the GENJCL.USER with LIST.LOG (+REXX against the report) and cut down the enqueue to a few seconds.

  The morals of this tale are:

  1. Try to ensure that archive and any subsequent processing completes before the next log switch
  2. Avoid %SELECT ALLOC(PRILOG,ALL). It is VERY expensive. LIST whatever + REXX processing is much safer.
  3. When analysing random incidences of poor performance, check DFS3257I message against RECON PRILOG times. If there is a gap, you have a problem.
  • Contacts/Comments:

  LUNCH

  Data Movement – John Boyle (BMC)

  • John presented on BMC’s new Change Data Move product, which is now available, providing the ability to capture changes to IMS, DB2 and VSAM data and propagate these changes to DB2.

  • This product is part of BMC’s Enterprise Data Movement strategy which provides comprehensive facilities for movement of data among all the platforms and storage techniques deployed within the enterprise.

  • The product consists of four main components:
    • A user interface (choice of ISPF or Windows GUI) to map source structure to target

    • A set of environment specific change capture routines, to capture changes in IMS TM, DL/I, DBCTL, DB2, CICS and VSAM Batch.

    • A transformation routine to translate the captured changes to s standard relational format

    • A set of change apply routines (only DB2 and Sybase available in current release) to apply the changes to the target system

  A copy of the foils was distributed

  2001 and all that – Discussion Topic

  • Alan Cooper announced availability (via PTF UQ17756) of a new IMS user exit to assist in manipulating stored dates in IMS data (DFSDBUX1). He also announced the IMS Y2K Exit Tool, available from IBM, which utilises this exit point (see foils).

  • Frank Fleming reported some problems with IMSPARS and IMSASAP. The ‘fix’ for these is to move to the new IMS Performance Analyser product.

  • The consensus was that people are much mire concerned about applications than they are about their IMS software and supporting products. Particular concern was for applications which don’t abend, but do not process data correctly.

  AOB

  • Ian Tyson asked for suggestions for agenda topics for the next meeting:
    • Alan Lovett suggested ‘How do you do your software maintenance?’ e.g. put each PUT Tape on when available or spend hours looking at PTF descriptions and selectively applying the right ones? Frank Fleming volunteered to contribute. Other volunteers welcome – contact Ian Tyson.

    • Alan Cooper suggested we ask Pete Sadler to talk about shared queues in IMS Version 6.

    • Other contributions on Version 6 experiences welcome.