Conference humane::scheduler

1162.0. "ACCVIO while using the Decwindows interface - 2.1b-9" by BACHUS::BANKEN () Fri Sep 27 1996 09:52

Dear friends,

One of our largest customer is experiencing the following problem with 
Decscheduler 2.1b-9 on OpenVMS AXP 6.2-1H2.

- Sched/inter=decw fails with an accvio 

Here are the details :

EPLZ02 > SCHED/INTER=DECW
SCHED-I-NODEFAULTS No DECW$USER_DEFAULTS:SCHEDULER$XUI.DAT file found
%SYSTEM-F-ACCVIO, access violation, reason mask=04, virtual address=00369318,
PC
=00375A30, PSL=03C00000

  Improperly handled condition, image exit forced.

        Signal arguments              Stack contents

        Number = 00000005                39313100
        Name   = 0000000C                00000033
                 00000004                00004220
                 00369318                00000000
                 00375A30                00000004
                 03C00000                00000492
                                         00000491
                                         002B4908
                                         00000000
                                         203C0000

        Register dump

        R0 = 00369100  R1 = 00000000  R2 = 00369100  R3 = 0001CC40
        R4 = 000093B8  R5 = 00353BF0  R6 = 00007F6C  R7 = 00000002
        R8 = 7FDF945C  R9 = 00000001  R10= 00807A28  R11= 7FDF9778
        AP = 7FDF92E0  FP = 7FDF92A0  SP = 7FDF931C  PC = 00375A30
        PSL= 03C00000
                                     


This happens when they select a user with over the 1000 jobs !.
They don't have the problem with other users having about 30 jobs.

Here is the recommanded quotas (note 492.2) I ask them to use :
 
    
    Maxjobs:         0  Fillm:       200  Bytlm:        49000
    Maxacctjobs:     0  Shrfillm:      0  Pbytlm:           0
    Maxdetach:       0  BIOlm:        50  JTquota:       2048
    Prclm:          60  DIOlm:       600  WSdef:         4096
    Prio:            4  ASTlm:       200  WSquo:        48000
    Queprio:         0  TQElm:       128  WSextent:    100000
    CPU:        (none)  Enqlm:      2000  Pgflquo:     200000
    
Regarding the huge number of jobs they want to load are the above
quotas suffisant for an AXP ?

Any other idea ?

Thanks in advance,

Alain.

    Note 25 has some discussion about large number of jobs.
    
    Steve

Hi Steve,

Many thanks for your reply.

Indeed, I found the problem back in note 25. I got the customer's 
scheduler db and I reproduced the problem under vax-vms 6.1 and
scheduler 2.1b-9.
As mentionned in the note 25 the problem occurs around the 1100 jobs.
It seems that a qar has been opened (QAR #232). Do you know or does
someone know where is this qar db ?

Anyway I'll raise an IPMT.

Just to know how alone we are, is someone of NSI listening ?

Thanks,

Alain.