[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference iosg::all-in-1_v30

Title:*OLD* ALL-IN-1 (tm) Support Conference
Notice:Closed - See Note 4331.l to move to IOSG::ALL-IN-1
Moderator:IOSG::PYE
Created:Thu Jan 30 1992
Last Modified:Tue Jan 23 1996
Last Successful Update:Fri Jun 06 1997
Number of topics:4343
Total number of notes:18308

2949.0. "TMA Breaks in a V3.0 Cluster Environment! " by MIMS::BEKELE_D (My Opinions are MINE, MINE, all MINE!) Wed Jun 30 1993 20:19

    
    Has anyone tried "TMA" (Transfer Mail Area) in a cluster environment?
    
    The code (OA$SM_TRANSFER_MAIL_AREA) assumes that it is in an ALL-IN-1
    sub-process when it runs the command procedure that defines the cluster
    logicals (@OA$LIB:OA$SM_SUBMIT_CLUSTER_WIDE OA$SM_DEFINE_SHARED_LOGICALS). 
    
    Looking at the code (do not have a V3.0 cluster to test it) it appears 
    that logicals are successfully re-defined only on the node where TMA gets 
    submitted.  Furthermore, having errored out on the other nodes, the
    code skips cleaning up (deleting sDAF(s) located on the old location)
    resulting in users writing to sDAF(s) on both the new and the old locations!
    
    Regards,
    Dan
    
T.RTitleUserPersonal
Name
DateLines
2949.1More info please!IOSG::THOMPSONThu Jul 01 1993 11:5212
   Dan,
	I don't understand your problem - what exactly is it that 
    is broken? I believe that this code is the same as in V2.4 
    - is the problem reproducable on a 2.4 system?

    Did you run OA$SM_DEFINE_SHARED_LOGICALS on a standalone system?
    If so remember that this requires the node-queue master file to
    be populated - it won't be on a stand-alone system.

Karl    

2949.2Gory details, solution and further issueMIMS::BEKELE_DMy Opinions are MINE, MINE, all MINE!Thu Jul 01 1993 20:2762
    Hi Karl,
    
    > ...what exactly is it that is broken?
    
    .ALL-IN-1 is in a cluster,
    
    .You submit TMA and you go home (sorry Karl, you have been 
       identified as my 'confident' customer for unknown reasons :^)),
    
    .The job copies the sDAF to the new location you specified,
    
    .The logical for the new shared area gets successfully redefined
       on the node the job was created on,
    
    .Next OA$SM_SUBMIT_CLUSTER_WIDE.COM tries to run but fails due
       to the fact that it is not in an ALL-IN-1 subprocess (which
       is how it was run under V2.4.  V3.0 uses A1SUBMIT.CLD) nor 
       has ALLIN1/NOINIT been done before calling OA$SM_SUBMIT_CLUSTER_WIDE.  
       OA$SM_SUBMIT_CLUSTER_WIDE, however, tries to WRITE to OAMAILBOX 
       (same code as the V2.4 code),
    
    .OA$SM_TRANSFER_MAIL_AREA, as far as I can tell, having encountered 
       a "warning", should jump to error handling routine and exist 
       without restarting ALL-IN-1 (customer insists otherwise),  
    
    .The sDAF on the source device does not get deleted (further proof
       that code jumped to error routine),
    
    .ALL-IN-1 at this point gets restarted (even his own logfile says 
       "Error occured whilst updating the logicals with new values - 
       this needs to be done before restarting ALL-IN-1"),

    .Users login on all nodes.  Those who logged on the node that the 
       TMA job got created on get to write to the sDAF on the new
       location.  Those who logged on other nodes get to write to the 
       same sDAF but on the old location (and you are at home, relaxed
       and drinking some cool-aid (ha!)
    
    To solve the problem of redefining the logicals cluster-wide I 
    have suggested to the customer to modify OA$SM_TRANSFER_MAIL.COM 
    as follows:
    
    Change 
     
    	$@OA$LIB:OA$SM_SUBMIT_CLUSTER_WIDE OA$SM_DEFINE_SHARED_LOGICALS
    
    to
    
     $ALLIN1/USER=MANAGER/NOINIT/OVERRIDE
     $DECK
      OA$INI_INIT
      COMMAND OA$SM_SUBMIT_CLUSTER_WIDE OA$SM_DEFINE_SHARED_LOGICALS  
      EXIT
     $EOD
    
    The problem of ALL-IN-1 getting restarted cluster-wide eventhough 
    there were problems encountered (as I stated before I don't have
    ALL-IN-1 in a cluster to disprove what he is saying) on some nodes
    in the cluster is what I would like others to comment on.
    
    Many thanks,
    Dan
2949.3workaround looks good for now..IOSG::TYLDESLEYWed Jul 07 1993 11:3810
    hello Dan.
    
    I think that your assessment of the situation in
    OA$SM_SUBMIT_CLUSTER_WIDE is essentially correct; the oamailbox has
    not been set up. We have a THR- number, and I hope this might get 
    fixed in a PFR.
    
    Not sure what your 'further issue' was?
    Thanks for your help.
    DaveT
2949.4mod OA$SM_TRANSFER_MAIL_AREA.COMIOSG::VOWLESRaphael VOWLES DTN 830 6910 REO2 G/M2Wed Jul 07 1993 11:5532
re:  <<< Note 2949.0 by MIMS::BEKELE_D

Here is a formalised solution. As Dan has requested: Has anyone had this 
problem??

Please reply to let us know the state of play.
tta rv  Wednesday 7-JUL-1993 10:54 
==========================================================================
==========================================================================



    Using Customization Management modify the command file element:
    OA$SM_TRANSFER_MAIL_AREA.COM as follows:

==========================================================================
    	 Change the one code line that reads:        (on or about line 94)
==========================================================================

$   @OA$LIB:OA$SM_SUBMIT_CLUSTER_WIDE OA$SM_DEFINE_SHARED_LOGICALS

==========================================================================
    	 To:                                                (6 code lines)
==========================================================================

$   ALLIN1/USER=MANAGER/REENTER/OVERRIDE/NOINIT
$DECK
    OA$INI_INIT
    COMMAND OA$LIB:OA$SM_SUBMIT_CLUSTER_WIDE OA$SM_DEFINE_SHARED_LOGICALS
    EXIT
$EOD