Conference vaxaxp::vmsnotes

   I'd look at recreating the queue manager database file, with fewer
   entries -- one cannot have multiple queue managers running in a
   (supported) VMScluster configuration.  

   One can get a queue file with this many entry slots only if one has
   this many simultaneous jobs queued -- this may have been a one-time
   fluke, or it may be something typical at the site.  If the former,
   recreating the file with fewer slots will help.  If not, CA7 needs
   some work...

   What the customer is trying to do -- multiple queue files -- is
   just asking for trouble.  Further, once that entry wrap hits, all
   bets are off...

re .*
    
>one cannot have multiple queue managers running in a (supported) VMScluster
>configuration.  
    
    Is this documented somewhere? I realise that what is supported is what
    must be documented (not what is unsupported) but the START command
    online help nicely tells you how to start a different queue manager and
    it certainly doesn't mention that you should stop the existing one
    first. [I have run clusters with two queue managers and not had any
    problems apart from user confusion but that was a while ago.]
    
    It would seem in fact that the existence of the multiple queue managers
    functionality is the cause of the customer's problem i.e. that because
    of this functionality any one queue manager will take a range of 10000
    entry numbers e.g. 0-9999, 10000-19999, 20000-29999 and hence when this
    range is exhausted a queue manager will have to go much higher for more
    numbers.
    
    Is there any way to control this? That is, if a customer *doesn't* want
    to use multiple queue managers then the sole default queue manager can
    just go beyond 9999 sequentially.

>>one cannot have multiple queue managers running in a (supported) VMScluster
>>configuration.  

  For OpenVMS V6.2 and higher you can have multiple queue managers (up to 5)
  running against a single, cluster wide queue data base (QMAN$MASTER.DAT),
  but each has its own .QMAN$QUEUES and .QMAN$JOURNAL file. See Sec 12.8 in
  "OpenVMS System Manager's Manual: Essentials" for details.

  The entry numbering scheme is as Derek outlined in .2, the upshot is you
  will get high numbers if you ever go over 10000 entries in one queue 
  manager's domain. See notes in CSC32::BATCH_PRINT_SERVICES for a precise
  description of the numbering scheme.

						John Gillings, Sydney CSC

SO..... If you have one queue manager with entry numbers in the range of 1000001
- 1009999 sat pending in a queue( just so they can't be used in any other
queue). What will happen when entry numbers 1-9999 are then used up. Does the
queue manager start useing 2000001.... and so on? or does it fall over?

:SO..... If you have one queue manager with entry numbers in the range of 1000001
:- 1009999 sat pending in a queue( just so they can't be used in any other
:queue). What will happen when entry numbers 1-9999 are then used up. Does the
:queue manager start useing 2000001.... and so on? or does it fall over?

   Queue managers share a database, and you can have -- pardon my earlier
   misphrasing -- one queue *database*.  The configuration you are trying
   to use likely requires multiple queue databases, and is thus unsupported.

   In other words, get the broken code fixed, or reduce the size of the
   queue database file.  (Again, the latter works only if there are less
   than `n' entries normally outstanding in all queues.)

  re .4: (Matthew)

    You asked exactly the same question in Dec 1995 in the BATCH_PRINT_SERVICES
  conference (note 201). Is the same customer still suffering under broken
  code? The answer posted back then by Dave Sweeny in note 150.1 has not
  changed:

from BATCH_PRINT_SERVICES 150.0 
>An entry number can be interpreted as follows:
>
>        X-QQ-NNNN where NNNN = 0 to 9999, normal entry number range
>                          QQ = 00 to 04, queue manager number (qman id - 1)
>                           X = 0 to 9, to extend 0 to 9999 range

282.4>>   Does the queue manager start useing 2000001

  Yes. Which means each queue manager can handle a maximum of 100,000
  simultaneous entries.

  "That should be enough for anyone" (to quote Mr Gates :-)

    further to .5
    
    The queue managers share the *master* file (name=QMAN$MASTER,default=
    SYS$COMMON:[SYSEXE].DAT). The master file contains, among other things,
    for each queue manager, the name and the directory location for its
    files and the execution node list.
    
    Each queue manager has two files located as follows
    
    spec=sysmgr_specified_dev_dir+name_of_manager+".QMAN$QUEUES"
    spec=sysmgr_specified_dev_dir+name_of_manager+".QMAN$JOURNAL"
    
    The name of the queue manager created by default is SYS$QUEUE_MANAGER.
    
    The sysmgr_specified_dev_dir defaults to SYS$COMMON:[SYSEXE] (at queue
    manager creation time).