Mainframe File Transfer WTF



  • Yesterday, I was given an overview of the Mainframe transfers files to a Unix server every half hour.  The catch is that the files are processed in sequential order, or there can be messy consequences. 

    Here are the most important details about said sequencing:

    1) The Mainframe will occasionally send three files in close succession and in reverse order.  As a result, if the Unix server is expecting file number 10, but gets 15, it won't do anything until file number 10 is sent (files 11 through 14 are sent before file 10 is sent).  The true fun comes in the fact that for every file sequence above the one received, a **CRITICAL** email is sent out warning of this and it takes 2 or 3 of these before anybody will react.

    2) The sequence number is stored in a file on the Unix server.  The Mainframe does not access this file, but rather it will maintain the value on its own.  Nobody could explain why the original developer, nearly a decade ago, did not opt to use a different method.

    Also, I should point out that to communicate back to the Mainframe, a realtime mechanism is used.  This mechanism could replace the file loads, but unfortunately, that costs time and money to re-engineer an established, "working" system.



  • @DoctorFriday said:

    Also, I should point out that to communicate back to the Mainframe, a realtime mechanism is used.  This mechanism could replace the file loads, but unfortunately, that costs time and money to re-engineer an established, "working" system.

    Control-M, maybe? Or the other Connect-something? Basically I remember some weird mix of both of these being used for Mainframe<->UNIX data transfers. Any of these would use files moving between the Mainframe and the UNIX box. True for B2B uses, and also true for internal business stuff. I think the reason for using files in these transfers has something to do with the files themselves leaving an auditable trail, as it would be easier to just use the realtime mechanisms.



  • @DoctorFriday said:

    Also, I should point out that to communicate back to the Mainframe, a realtime mechanism is used.  This mechanism could replace the file loads, but unfortunately, that costs time and money to re-engineer an established, "working" system.

    I think I had exactly the same meeting today too.  One of the accounting team says, "Why do we need to change the current, working system?"  Well maybe because 2.5 months after the slo-mo train wreck that was "go live" we still don't know how many millions of dollars aren't accounted for.  The meeting broke up pretty quietly after I said that.


Log in to reply