Conference iosg::all-in-1_v30

365.0. "Users receive msg with subject, no text" by AUNTB::GRAYD () Mon Mar 30 1992 20:23

    I've looked through the old conferences and found nothing, so I need
    some assistance.  The customer scenario is as follows:  It is a large
    site (over 14,000 ALL-IN-1 users spread over 6 nodes/clusters) running
    ALL-IN-1 V2.3.  
    
    In order for the EW job to complete within a weekend on some of the larger
    machines, they have modified it so that it splits the users up into
    subgroups, alphabetically, based on a predefined number specified when
    the routine is scheduled.  So, instead of running one big job, they run
    20 small ones, most of them going at the same time.  They also have
    cleanmail incorporated.  Customer asked me to add some code to EW that
    would send a message to users with the total number of documents over a
    certain limit (defined by a logical) informing them that they have 'x'
    documents of a certain type (documents or mail messages).
    
     So, I added the following psuedo-code to the JANCLEAN script that
    executes immediately after the CLEANMAIL message is sent:
    
        Initialize counters
    	For CAB$ do increment counters based on document type/age
    	If #total_docs (total # in user's DOCDB) is greater than
    		LOG$OA_OLD_DOC_LIMIT then
    	MAIL PUSH
    	MAIL CREATE/OPEN/NOSEND
    	MAIL TO "ME"
    	MAIL SUBJECT "Document Report"
    	MAIL TEXT " "
    	   ...
    	MAIL CLOSE_MESSAGE
    	MAIL SEND
    	CABINET DELETE_OR_REFILE ,OA$WASTEBASKET
    	MAIL POP
    
    Naturally, this ran like a champ on the small 50 user test system I
    had.  The first time I ran it on the live systems, almost no one received 
    the message because I ran into a CHANNELCNT limitation and had to 
    increase it.  Now, I'm finding that all users are getting the message, but
    some are getting the complete message and some (the majority) are
    getting a message that contains just the subject, but no text at all. 
    The only consistency that I've found is that once a user
    in a subgroup fails to receive the message with text, everyone below
    that user in that subgroup fails as well.  Example:
    
     On one two-node cluster, each subgroup was submitted as a batch job on
    alternating nodes.  The first job was on node A, the second on node B
    and so on.
    
      Subgroup 1 - first 4 users had text, the rest didn't
      Subgroup 2 - all users had text
      Subgroup 3 - no users had text
      Subgroup 4 - all users had text
      Subgroup 5 - first user had text, rest didn't
      Subgroup 6 - no users had text
      Subgroup 7 - first user had text, rest didn't
      Subgroup 8 - first 3 users had text, rest didn't
    
      I must be running out of something, but I'm not sure what.  Any
    ideas?  I appreciate any help.
    
    Thanks,
    
    Danny

    I've been trying to isolate the problem further.  The CSC suggests one
    of two problems, the old NEWDIR not releasing global buffers problem or
    a PGFLQUO problem.  Because users are continuing to receive the
    CLEANMAIL message, I'm ruling out the NEWDIR bug for right now.  The
    question I need to know is how do the FOR CAB$ and MAIL functions
    create pages.  Are they READ or WRITE and how are they charged against
    PGFLQUO.  In order to do this, I may needt to bump up PGFLQUO for the
    ALLIN1 account, but I also have to figure out how much.  I may also
    need to bump VIRTUALPAGECNT on the system.  Again, I need to know how
    much.  If anyone can explain how this works, please give me some
    assistance.  
    
    Thanks,
    
    DG

    I'm afraid that I can't give you any details about how ALL-IN-1 uses
    PGFLQUO other than to say that ALL-IN-1 tends to enjoy maximizing its
    use of resources.

    I would suggest that you might want to observe the operation in an
    attempt to observe the resource usage.  One way could be to use
    MONITOR/RECORD PROC.

    Another approach is to double PGFLQUO and see if that has any changes. 
    The setting is supposed to be at 20,000 for normal operations.

    Good luck,
    don