On Thu, 29 Jan 2009 11:13:57 -0600, Arthur Gutowski wrote: >On Wed, 28 Jan 2009 17:51:05 -0600, Tom Marchant wrote: > >>On Wed, 28 Jan 2009 12:44:50 -0600, Arthur Gutowski wrote: >> >>> >>>The book mentions the JES3 consideration for C/I, but unless I'm blind, it >>>doesn't describe how to "refresh" the JES3 copy of the symbol table without >>>IPL, namely *S <main>,CONNECT >> >>I'm confused, Art. I used SYMUPDTE many times when I was there and never >>had any problems with JES3. Was there something that I was missing? Or >was >>I just lucky to have been wearing my bulletproof shoes that day? > >Ha! You probably were. > >Let me attempt, in my less-than-JES3-fluency try to address both Mark's and >your questions. My original foray into this post is tangential, but I picked up >on the reference to the IBM redbook, and thought someone might find useful >some additional info on the JES3 caution useful. > >It's not that a program doesn't go through ASASYMBM service on a JES3 >system, it's that JES3 C/I doesn't go through ASASYMBM for STC (including >started JOBs) like JES2 does. ALIAS / SYMBOLICRELATE is unaffected >because CATALOG uses ASASYMBM. Thus the tangent. > >On JES2, as with JES3, a batch job may go through C/I on one MAS member, >but initiate on another - which is why you can't use system symbols in jobs. >On JES2, STCs always C/I and execute on the same member where the START >command were issued - which is why you can use system symbols in STCs. > >On JES3, that ain't necessarily so. I'm not sure if you *can* run multiple C/I's >in a single JES3 MAS, but I know we don't. While we may be able to run C/I >on a LOCAL, we always run it on the GLOBAL. This means an STC, unless >started on the GLOBAL, will C/I on a different member than it executes. > >For JES3 to support symbols in STCs, the GLOBAL has to know the contents of >all of it's LOCALs' symbol tables. To avoid overhead of obtaining the table, or >parts of it, every time a START command is issued, JES3 obtains a copy of the >symbol table for each member, including itself (to keep code path the same for >each type of MAIN), only during CONNECT processing. CONNECT happens >during initialization, but can be re-driven any time with a *S <main>,CONNECT. > >We never got burned during your tenure here because we hadn't yet tried to >implement or back out a product involving a dynamic symbol change (to avoid >IPL) that referenced said symbol in STC JCL. If you're curious, send me a >note offline, and I'll cite examples and go into more (is that possible?) detail.
That's true. The JCL used a data set alias that was defined with SYMBOLICRELATE. The redbook cautions that SYMUPDTE should only be used to update symbols used for SYMBOLICRELATE. -- Tom Marchant ---------------------------------------------------------------------- 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
