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

Conference clusta::acms

Title:ACMS comments and questions
Notice:This is not an official software support channel. Kits 5.*
Moderator:CLUSTA::HALLAN
Created:Mon Feb 17 1986
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:4179
Total number of notes:15091

4101.0. "SYSTEM-F-VECFUL and MIN_CPIS?" by CSC32::C_BENNETT () Wed Feb 12 1997 12:20

    Hello All,
    
    I have a customer who executes a ACMS/ENTER and upon this command 
    gets a SYSTEM-F-VECFULL error condition.    The application consists
    of a DECforms form which gets cached from the application node.  The
    DECforms form does not have the corresponding ESCAPE ROUTINE linked
    into it - this has been manually located on the submitter node and
    is pointed to by the FORMS$IMAGE logical name.  The ESCAPE ROUTINE
    performs functions in privledged mode.  
    
    To make a long story short the dispatch message vector is full and
    this is the cause of the SYSTEM-F-VECFULL error condition.  Additional
    investigation of CPs table thru SDA shows that duplicated entries
    in the message dispatch vector are present.  
    
    My question deals with the multithreaded nature of the CP in this
    context - specifically is there a one for one correspondence between
    the MIN_CPIS parameter at the CP level and the amount of space/entries
    made into the dispatch message vector?    
    
    Would including the ESCAPE ROUTINE in the shareable form reduce the
    space requirements in the dispatch message vector? 
T.RTitleUserPersonal
Name
DateLines
4101.1OHMARY::HALLBill Hall - ACMS Engineering - ZKO2-2Wed Feb 12 1997 14:2244
    
    
    	MIN_CPIS has nothing to do with the dispatch vector table.  The
    	thread package in CP (called THD) has a preset number of thread
    	table entries.  This is a compile time value, it used to be 100,
    	I think it's now a little higher, but all it does is change
    	the size of the thread table.
    
    	Each user that enters ACMS through ACMS/ENTER has a CPI (which 
    	stands for Command Processor Interpreter) thread on which they
    	execute.
    
    	The MAX_TTS_CP and MIN_CPIS parameters are TSC parameters.  When
    	a user does an ACMS/ENTER, The TSC looks for a place to put that
    	user.  If a CP is already at the number specified by MAX_TTS_CP,
    	then it will lokk elsewhere or may startup another CP.
    
    	MIN_CPIS comes into play as users log into ACMS.  For example,
    	if MAX_TTS_CP is set to 20, and PERM_CPS is set to 1, there
    	are 20 CPI threads available.  As users log in, this counter
    	is decremented (not really but think of it that way).  When
    	the number of free CPI threads equals MIN_CPIS, the TSC will
    	start another CP.
    
    	The error message your customer is seeing, VECFULL says:
    
     Facility:     SYSTEM, System Services
    
      Explanation:  This severe error message is associated with a status code
                    returned from a system service request. The privileged vector
                    limit of 42 is exceeded.
    
      User Action:  Remove unnecessary vectors and relink.
    
    
    	It sort of sounds like they are activating too many images or
    activating them incorrectly.  If the images are installed images, you
    should only see 1 copy of them in CP's virtual address space.  If they
    are activating them by filename and specifying a version #, then this
    will bypass the known file list lookup and will probably result in individual
    activations.
    
    Bill
    
4101.2IFDL::RICEThu Feb 13 1997 06:586
We've seen problems like this in the past where they are placing DECforms
extracted forms into images.  The common problem seen is too many of these
images, i.e. they make an image for each form.  By grouping many forms into one
image they can work around the, rather small, vector table limit on VAX.

-Tim
4101.3CSC32::C_BENNETTThu Feb 13 1997 08:5512
    This is on an Alpha and works fine when the form doesn't get cached
    between a OpenVMS 6.2 to OpenVMS 6.1 system.    The ESCAPE ROUTINE 
    performs priv code and is not apart of the form so it is manually
    copied to the OpenVMS 6.1 system.    The ESCAPE ROUTINE was linked
    under OpenVMS 6.1.    
    
    The odd issue is that further examination of the dispatch message
    vector reveals duplicated entries - this originally made me think
    1 thread - 1 duplication?  I suppose if the customer pushes this
    I'll just SPR the issue and have you guys look into it.
    
    			Thanks alot