Sending OS/390 datasets 1

[Grimmonkey]

Is AMHSNFL appropriate to send large variable length record datasets through MQ?

Is there an easy method of performing the above PUT from within a batch Rexx?

 

[kevinf2349]

Be careful sending large files as messages as the sequencing can get out of order. If the data sequence doesn't matter then fine, but if it does it would be far better to use reference messages.

 

[Grimmonkey]

What problems could arise with sequencing?

The data will be PUT sequentially onto a dedicated queue and the GETs will check for sequence numbers within the data to ensure messages are retrieved FIFO.

 

[kevinf2349]

I guess it really depends on if you are the only person putting to that queue right and/or how the recieving application is written.

If the sole purpose of what you want in to file transfer an entire file to another platform....take a look at reference messages.

If you are just wanting to load a Queue up, then there is an IBM utility (CSQUTIl) that may be of use to you.

 

[pmeekin]

Don't forget that Reference messages are not supported on the Mainframe.

Although not necessarily the best solution, a common method of doing this is for the putting program to read the first n bytes of the file, append a header detailing e.g. segment number 1, total bytes in file etc. and then put it to the target queue. Then each subsequent chunk of the file has a header with the segment number until it sends the last bit with a header showing that it's the end of the file.

The receiving program then ensures that it has all the necessary segments in the correct order. You could use Correlid to hold the segment number if you wanted or message groups although the latter are also not supported on z/OS.