On Fri, 24 Jun 2016 10:27:16 -0500, Walt Farrell wrote: >On Fri, 24 Jun 2016 15:17:50 +0000, Sankaranarayanan, Vignesh wrote: > >>This is to cope with MQ FTE's "feature" of not locking datasets for the >>duration of a transfer. >> >>Using ANT scripts to initiate a transfer of a list of datasets that resolve >>to a wildcard, we've found that in the time gap between the >>initial resolve of the wildcard and the transfer of datasets found in the >>wildcard, those datasets are processed by some other entity, >> That hazard exists for any process that generates the list and processes it in separate operations.
>>leaving the MQ FTE step to fail saying dataset deleted (already processed). > If the MQ process were to fail more gracefully, issuing an error message and proceeding to the next entry you'd be better off. Data sets created in the window might yet pose a problem. >>There's the outcome=defer option in the ANT script but I don't believe it has >>led to the job to stay active until the MQ FTE transfer >>completes fully. > If your objects were members of a PDSE rather than data sets (but that would be a major design change) DESERV might be able to make a member list and establish lasting connections in an atomic operation. (I haven't tried it.) Or you could ENQ EXC on the entire library, locking out all other users. If your objects were UNIX files (but that would be a major design change) you might play games with directory links. Was MQ designed as a multi-user facility? It seems to lack locking, COMMIT, and ROLLBACK. >I am concerned that your allocation as OLD will prevent the subsequent use of >them for transfer in a later job-step. It may work, but it may not. That is, >if you have a data set allocated OLD by your exec, and then a later step in >your job tries to do a dynamic allocation either SHR or OLD I think that later >allocation may fail. > A batch job (drawbacks already discussed) could ENQ all in a monolithic operation, waiting for others to become available. >Plus, you have my initial concern that your use of ALLOC in the exec may >itself fail if someone else has the data set allocated already. > I do much of my Rexx work under z/OS UNIX with _BPX_SHAREAS=NO so I needn't worry about freeing resources before address space termination. -- gil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
