Conference azur::mcc

6493.0. "the AM routine is not invoked" by JRDV04::MATUZAWA () Mon May 12 1997 07:46

Re .5

>    In your failure log all the arguments seem to be set up correctly and
>    the last line which is printed before the call to ACMS is normal:
>    
>    HANDLE STATE, before call: FIRST
>    
>    The first lines when the call returned is also normal as expected.
>    The CVR returned %SYSTEM-E-DEADLOCK is what is returned from the 
>    called process. The error has to come from the system or RMS.
>    

  By inserting printfs into the access module routine that ACMS provides, I
  found that the control doesn't reach the routine when a deadlock occurs,
  where it should be called by MCC in the normal case.

>    If the error was detected by DECMCC then the error would start with
>    MCC not system.

  Is there any chance to invoke system calls or RMS calls, e.g.
  lib$find_image_symbol, just before invoking the access module routine?

>     I noticed that the second file was created very shortly after the
>    first. Is it possible that the ACMS pocesses and threads have not
>    terminated when the next one is started. Would a delay between each
>    time you submit the MCC and ACMS commands help?

  I'v not answered your previous question: 20 jobs are submitted with
  almost no delay. Each of submitted processes invokes a MCC command (to
  start one server), the command I believe invokes the access module
  routine ACMS provides, and the routine does DCE RPC to the ACMS
  controller process(the controller is a multi-threaded process, but this
  fact doesn't seem relevant).  So, the answer is YES in the sense that the
  next job can start before previous ones don't terminate.

  A delay may solve the problem, but this has already been done at the user
  site because of another reason.  They are greedy enough to want more
  reduction during initialization.

regards,
matuzawa DEC japan