Conference virke::mrmemo

    Hello Dave!

.0> %DDS-F-OPSYSFATAL, Fatal operating system interface error
.0> -SYSTEM-F-IVLOCKID, invalid lock id
     .
     .
.0> The problem only occurs when messages are received from MEMO and 
.0> DDS validation for MR recipients is enabled.

    We have experienced this problem sometimes during test and development 
    of MRMEMO but never been able to track down the reason. It is probably 
    some resource problem that makes DDS fail.

    We have never run into this problem on our big development machine (8800
    with 272MB memory and high quotas everywhere). It has only happened on
    our small (�VAX II) test machine, which makes the resource problem 
    theory stronger.

    I saw a similiar error:

        %DDS-F-OPSYSFATAL, Fatal operating system interface error
        -SYSTEM-F-NOSYSLCK, operation requires SYSLCK privilege

    in a NETSERVER.LOG for DDS$LSTN a couple of weeks ago. At that time, the 
    system in question had run out of GBLPAGES and since the DDS data files
    are global buffered, this could have been the reason (yet another
    resource problem.

    So, check that the system has some free GBLPAGES left and try to add
    some more explicit quotas in the MRMEMO start command (in 
    MRMEMO$DIR:MRMEMOLOGIN.COM). For example /IO_BUF=40 and /IO_DIR=40.


.0> The customer also noted that MRMEMO was page-faulting excessively.

.0> I advised them to start the detached process with specific values for
.0> the working-set parameters.

    Yes, that's good. A large /EXTENT doesn't cost much and will help a 
    lot if PQL_DWSEXTENT is low.


.0> What are the support channels for MRMEMO ?

    In Sweden and Germany, MRMEMO is a product and support is handled the
    "usual" way. For the rest of the world, I think support is handled by
    the ASSETS library in Turin, Italy. I'm not sure though if they are
    supposed to take care only of Europe.

    This Notes file can also be used for support questions. We are a little 
    more positive to that than many other Notes files (at least as long as 
    the support load is moderate :-). Why send around lots of paper forms
    through "formal" channels when we have Notes?

    Anders

    Hello Anders!!


       The resource theory has the ring of truth about it.  The customer
    is putting together a demo EDI system on a microVAX 3100, just to
    prove DEC/EDI, MR/MEMO, etc all really works, and I think the 3100
    is pretty heavily extended.

       I'll report here when things become clearer.


               Dave...

Hello Anders,

i can now give some more information about a customer in germany like the
problem from Dave.

the customer has 
MR/MEMO V2.0
MAILbus V3.1
with MRG
with DDS	<=== Masternode
with ER		<=== Network Management Node
DDSSYNC V1.0
SNA Software
	Terminal Emulation
	Printer  Emulation
	SNAAPI
DTF
a �VAX with 10 MB and 1 RD54 and he tells me he becomes an upgrade
of 6 MB and a sec. DISK. (16.12.1990)
--  the �VAX is the only gateway to the IBM stuff, some ALL-IN-1 nodes
    are behind
    
    
- problem
	sporadical the MRmemo process generates error messsages and grasps 100%
	CPU time. the opcom message sounds:
	     %DDS-F-OPSYSFATAL,FATAL OPERATING SYSTEM INTERFACE ERROR
     	     -SYSTEM-F-IVLOCKID,INVALID LOCKID

	     %MRMEMO-I-SHUTTING,SERVER SHUTTING DOWN
     	     %MRMEMO-W-DDSINITTIME, DDS INITIALISATION TIMED OUT


   	in the MRMEMO1.LOG we can read:

    	%DDS-F-OPSYSFATAL,FATAL OPERATING SYSTEM INTERFACE ERROR
     	-SYSTEM-F-IVLOCKID,INVALID LOCKID
    
   	%MRMEMO-W-TRACE, TRACEBACKFORCED FROM THE SENDER HANDLER
    	-MRMEMO-I-FSM,STARTING,CONNECTING TO MR AND MEMO GATEWAY AFTER CLOCK TICK
    	-MRMEMO-I-RING,RING:'][4A][4A][4A][4A][4A.'.SUBSTATES:0000000

    %TRACE-W-TRACEBACK,SYMBOLIC STACK DUMP FOLLOWS
    MODULE NAME		ROUTINE NAME	LINE	REL PC		ABS PC

    SRVLOG		SRV$LOG_HANDLER	3641	000001BB	000BE593
    ----ABOVE CONDITION HANDLER CALLED WITH EXCEPTION 01458804:
     %DDS-F-OPSYSFATAL,FATAL OPERATING SYSTEM INTERFACE ERROR
     -SYSTEM-F-IVLOCKID,INVALID LOCKID
    ----END OF EXCEPTION MESSAGE
						0019B74B	0019B74B
						001A737E	001A737E
						001071E2	001971E2
   SRVMRC		SRV$MRC_INI 4289	0000012C	000C1928
   SRVACT		SRV$ACT_C   3938	000001BF	000C59AD
   SRVDSP		SRV$DSP_FSM 3109	00000338	000BE2C0
   SRVMMO		SRV$MMO_ONE_LIFE
				    4013	000001EF	000BBF77								
   SRVMMO		SRV$MMO_MAIN3845	00000017	000BBD7B
						00111C7A	00111C7A
   KOTERM		KOTERM	     804	00000039	0010F66E
						00111C55	00111C55
   KODOC		KODOC	    1768	00000097	0010C8B0
						00111C55	00111C55
						001677B2	00111C55
   ADA$ELAB_DDS		ADA$ELAB_DDS		0000000E	0009FE0E
						00111C55	00111C55


        %MRMEMO-I-SHUTTING,SERVER SHUTTING DOWN
     %MRMEMO-W-DDSINITTIME, DDS INITIALISATION TIMED OUT

 	... and more of the above .....

i gave him the hints of .1 and he will now observe the system, may be the 
memory-upgrade helps.

regards
	Michael
    

    
> i gave him the hints of .1 and he will now observe the system, may be the 
> memory-upgrade helps.

    Possibly. We haven't really tracked down this error in detail yet, all
    we know, as Anders says, is that it's some kind of resource problem.
    We'd appreciate if you could take the time and post your customer's
    experience here, to help us provide some general configuration guidelines.
    
    cheers,
    
    	Stefan

	Hi guys... I'm back from changing dipers!

	We had the problem of invalid lock ID at several sites. I got the
	problem to go away by setting up the ENQLM for MRMEMO, seems DDS
	uses a lot of locks and seems to take out looks at some point without
	checking that it is actually granted the lock requested. When the lock
	eventually is used, the mentioned error is discovered, my guess is
	that the program passes a lock id like 000000 or something like that
	from a parameter that did not get the lock id requested...

	After putting up the ENQLM by a factor of 10 the problem automagically
	disappeared! This only seemed a problem at sites that had V1.x installed
	later vesrsion came with higher ENQLM for the MRMEMO account but 
	upgrades allways kept the old (small) value.

	Take it away boys...

	H�kan

    
    Hi H�kan,
    
    Welcome back!
    
    	Stefan on behalf of the MRMEMO Engineering team

Hi,

-2
>After putting up the ENQLM by a factor of 10 the problem automagically
>disappeared! This only seemed a problem at sites that had V1.x installed
>later vesrsion came with higher ENQLM for the MRMEMO account but 
>upgrades allways kept the old (small) value.


I encountered the Lock-Id Problem with DDS Validation enabled for MR-sender 
and MEMO-recipient with MRMEMO V2.1!

Thank you H�kan, the hint with ENQLM came just in time for me, I will visit 
PHILIPS tomorrow and try the solution.

BTW, Philips Austria will probably buy the MEMO-Gateway!

Kind regards

Helene

With an ENQLM of 800 on the MRMEMO account the Problem did not disappear.
My customer is now trying with ENQLM set to 2000. 
Is this approximately the value, you were talking about?

Regards

Helene

    
    Helene,
    
    Changing the account parameters won't help. There should be an explicit
    qualifier entry for ENQUEUE_LIMIT in MRMEMO$DIR:MRMEMOLOGIN.COM. The
    value used for MRMEMO V2.1 is 400. This should be enough as far as we
    know. Does your customer's MRMEMOLOGIN.COM contain the following (note
    especially the /enqueue_limit entry)?
    
    -------
    $       run/detach sys$system:loginout-         ! start the server from
                    /process="''prc'"-              !  the MRMEMO account
                    /ast_limit=36-
                    /file_limit=32-
                    /page_file=10000-               ! insert non-default limits/
                    /buffer_limit=36000-            ! bytlm
                    /enqueue_limit=400-             ! enqlm
                    /prio=5-                        !    quotas here
                    /input=Mrmemo.Com-              ! setup file
                    /output='logfil'-               ! log file
                    /error='logfil'
    -------
    
    Stefan
    

Hi Stefan,

thank you, the values in the customers MRMEMOLOGIN.COM were the same as in
your entry. 

We tested the following values for /enqueue_limit in the command procedure
(I was on the phone, the customer himself on the system):
800: nearly no change
2000: 4 times as many mails were transferred, before the error occured 
(approx. 25 mails from VMSmail to Memo)

So it seems to be a resource problem, but is it useful to increase enqlm still
more?

The context is the following:

The customer is making load tests and is sending mails with a command procedure
from both sides (MR to Memo and Memo to MR).

DDS validation: MR to MEMO sender enabled
                MEMO to MR recipient enabled

Address translation: enabled, type: Memo


The error messages from the logfile:

%MRMEMO-F-STRANGE, non-expected event for current stage

%MRMEMO-W-TRACE, traceback forced from Server Handler
-MRMEMO-I-FSM, sending to MEMO after data received from MEMO
-MRMEMO-I-RING, ring: '2A 27 1B 13 36 18 2A 2B 23 26.', substates: 00002070
stack dump follows
.
.
.
%DDS-F-OPSYSFATAL, Fatal operating system interface error
-SYSTEM-F-IVLOCKID, invalid lock id
%ADA-I-TASTERUNH, Task with ID %TASK 7 of type results_handler has terminated 
due to unhandled exception
Time: .... message from server MRMEMO1:
%ADA-F-EXCCOPLOS, Exception was copied at a "raise;" or "accept", but some details
were lost
%DDS-F-OPSYSFATAL, Fatal operating system interface error
-SYSTEM-F-IVLOCKID, invalid lock id
.
.
-MRMEMO-I-RING, ring: '1A 1A 1A 1A 1A 1A 1A 1A 1A 19.', substates: 00000430
stack dump follows

The MRMEMO-Server process hangs in different waitstates, e.g. RWMPW or even
MUTEX wait, which cannot be seen very often on a VMS System.

Do you need more information? Are there any System Parameters that could affect
DDS so badly?

Thank you for any hint!

Kind regards

Helene

(I'll be back in the office on Thursday, the 17th)

    
    I'm afraid we're still fumbling in the dark on this one. We have
    established that raising ENQLM will help. So far so good. Memory seems
    to be a problem too. Have you tried to raise the Working space and
    extent as suggested a few notes back? 
    
    Try these values:
    
$	run/detach sys$system:loginout-		! start the server from
		/process="''prc'"-		!  the MRMEMO account
		/ast_limit=36-
		/file_limit=32-
		/page_file=10000-		! insert non-default limits/
		/buffer_limit=36000-		! bytlm
		/enqueue_limit=400-		! enqlm
		/EXTENT=2048		    !***! max physical memory
    		/MAXIMUM_WORKING_SET=2048   !***! max working set
		/prio=5-			!    quotas here
		/input=Mrmemo.Com-		! setup file
		/output='logfil'-		! log file
		/error='logfil'

Hi,

I've been testing the System at Philips and increased the workingset parameters
in MRMEMOLOGIN.COM with no success, but increasing the astlm seems to have done 
it. I changed it from 36 to 80 and from that time the error did not occur
again. My customer will do extensive load testing this week and I will post 
the results here. 

Kind regards

Helene

    Hi Stefan,
    
    re .4 : the customer uses since December the memory-upgrade and he have
            no more problems with MRMEMO
    
    cheers
    	Michael

Philips did extensive load testing in both directions (sending mail with command
procedures from both sides):the problem did not occur any more with ASTLM 
set to 80. 

rgds.

Helene