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

Conference cookie::sls

Title:Storage Library System
Moderator:COOKIE::REUTER
Created:Sun Oct 13 1991
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:2270
Total number of notes:7850

2243.0. "Alpha VMS crashes on Rddeallocate (RDF 4.1a)" by BACHUS::RENTY () Thu Apr 17 1997 08:29

I already mentioned this problem in note 2233.10, but satrt a new note here as
I have more information and as it is was a different problem form the base note
2233.0 :

>    On the customers site, we are also investigating a lot of VMS crashes
>    on their Alpha system (about 3/day).   Since we disabled SLS, the
>    system did no more crash.   We are currently checking their sysdump files
>    (about 10, of whom 3 with POOLCHECK enabled)>  I will ask the customer
>    to enable SLS again, but without compression.
>    Our main suspect is RDF, causing some nonpaged pool corruption.

We did some more tests, and we found that the system is crashing when the
backup logfile is executing RDDEALLOCATE.COM.   The backup itself succeeded
without problems.
The system is an Alpha 4000 /VMS V6.2-1H3/SLS V2.8A/RDF V4.1a
Enabling or disabling RDF compression doesn't change the behaviour.
We started 2 simple backups via RDF, and the system crashed twice with 
"BADDALRQSZ, Bad memory deallocation request size or address", which confirms
our opinion about nonpaged pool corruption.   This evening we will receive the 
dump files on tape.   As the problem becomes urgent, we will escalate these
problems tomorrow, unless someone can point me to an answer on this question :

Are there any known problems of crashes caused by SLS/RDF (SLS V2.8a on Alpha 
OpenVMS 6.2-1H3) ???


Bart
T.RTitleUserPersonal
Name		DateLines
2243.1CX3PST::BSS::SAULThu Apr 17 1997 10:594
I also have a dump locally and have left message with Marty  that
its available if someone would like to see it.

Ted
2243.2How could you reproduce this crash ?BACHUS::RENTYFri Apr 18 1997 03:5113
    re: 2233.11
    
 >   My first attempt to reproduced this here caused my Alpha client
 >   to crash.  It was in the RDF driver.

    How did you reproduce it.  I am continuously doing backups from my Alpha 
    system to an RDF served tape on a VAX, and I couldn't crash my system...

    The main difference with my customer's system is the VMS version (7.1
    instead of 6.2-1H3), and SLS V2.9 instead of V2.8a, although these
    contain both the same RDV version V4.1a.

    Bart
2243.3CX3PST::BSS::SAULFri Apr 18 1997 13:577
>How could you reproduce this crash ? 

I wasn't trying...it just went down.  I'm have 7.1 Alpha on the RDF client with
2.9 and Alpha 6.2 on a 2.8 RDF server.  Sounds like its intermittent, kind of
like the old problem.

Ted
2243.4For RDCDRIVER crashes, do this...COOKIE::MCCLELLANDMarty, SLS/MDMS EngineeringTue Apr 22 1997 08:3129
Bart,

  We received your IPMT case (CFS.50602/BRO100972) describing the 
  INVEXCEPTN bugcheck.  I'm in the process of reporting your case
  and three others (including Ted's) to TTI.  

For all readers..

  If you experience any RDCDRIVER system crashes, please do the
  following:

    1. If an Alpha system, verify the system was rebooted after
       upgrading to V2.8A (or V2.9-FT1).  This can be accomplished
       by ensuring the link date for RDCDRIVER.EXE is earlier than
       the system's last boot time as shown by SDA commands
            
             $ANALYZE/CRASH sysdump-file-name
             SDA> exam/time exe$gq_boottime

    2. If it had not been booted after the upgrade, wait for a recurrance.

    3. If it had been booted afer the upgrade, save the crash dump
       using the SDA command COPY and submit an IPMT case including
       the output of SDA commands SHOW CRASH and SHOW STACK (at a
       minimum).

thanks,
Marty
2243.5IPMT case BRO100972BACHUS::RENTYThu Apr 24 1997 05:159
    For those who are interested in this case, this problem was escalated
    through IPMT : case BRO100972.
    
    As a temporary workaround, engineering proposed to downgrade SLS V2.8A
    to V2.8 + eco 2.    After this downgrade, the customer had indeed no more
    crashes.
    
    Bart