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

Conference smurf::ase

Title:ase
Moderator:SMURF::GROSSO
Created:Thu Jul 29 1993
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:2114
Total number of notes:7347

2065.0. "installation thro' remode node ?" by HGOV08::SUBBU () Fri May 16 1997 13:27

    
    Hi
    
    we are facing some inconsistancy issues while installing OPS V 7.3.2.3
    on dual AS 2100 5/300. the OS is DU V4.0b. The common scsi disks are on
    HSZ40.
    
    Can we install, create and import data from remote node (ie. the
    machine does not own the DRD service)?
    
    I will be greatful if someone can give some insight ?
    I hv put drd-data-compare=3 and seeing the machine panicing when we
    access thro' MC. Any quick clue as to find what piece of h/w is the
    culprit ?
    
    I will be greatful for a quickest reply.
    
    
    Thanks in advance,
    Subbu
    
    PS: note 1991.* was really helpful to me. Thanks a lot for that too.
T.RTitleUserPersonal
Name
DateLines
2065.1drd-data-compare set everywhere?NNTPD::"pelle@zk3.dec.com"Pelle WahlstromFri May 16 1997 14:4215
You can access any DRD service from any node in exactly the same way. The
difference is that if you do it form a client node, data is going to be
transported over MC before/after the physical disk is accessed from the 
DRD server node. There may be a performance penalty doing that.

When using drd-data-compare be ABSOLUTELY SURE that all nodes in the cluster
have drd-data-compare set to the same value. To do this, set it in
sysconfigtab
and reboot ALL the nodes. If you don''t do this the machines may panic because
one machine lookg for checksums that doesn't exist.

If you have data inconsistency problems, are you sure that new-wired-method
has
been set properly (See note 1991 last message).
[Posted by WWW Notes gateway]
2065.2still one node panics...HGOM22::JENNYCONGFri May 16 1997 16:1023
    
    hi 
    
    thank u vey much for your quick reply.
    
    i am sure on the both machines i have set drd-data-compare =3
    Also we hv set new-wire-method = 0 on both the machines. (i hope
    it is new-wire-method NOT new-wired-method).
    
    The latest observation was that when ever we access the database over
    memory channel, one of the machines always getting paniced. In this
    case how can I conclude which is the hardware piece giving me trouble?
    
    My bet is on MC card ? May i have your opinions please !!!!
    
    This is a very serious exercise bcos it results not to get a lot of
    forthcomming orders.
    
    May I solicite your kind help ?
    
    Thanks once again.
    Subbu
    
2065.3similar problem with 3.2fKYOSS1::GREENFri May 16 1997 17:1214
    	The last time I was involved in something similar, the fix was
    two fold(rackmount 2100a):
    	1. PCI backplane change.
    	2. Several patches to OS (3.2f)
    
    	Also, there is a pre-elease patch (simport.o) for data corruption
    problems on KZPSAs.
    
    	Along with drd-data-compare=3 I set (under dbx) dlm_checksum_msg=1
    and when we started OPS and started sending data across MC one machine
    would crash immediately with a "dlm corrupt message" panic.
    	I will mail a detailed report of the outage.
    		dick
     
2065.4simport patch applied already...HGOM22::JENNYCONGFri May 16 1997 17:5118
    
    hi
    
    The machines are AS 2100 5/300 standard systems with digital unix v
    4.0b and trucluster s/w v 1.4.
    
    we hv already applied sim_port patch for KZPSA already.
    
    pls mail your inputs to the following addresses:
    
    subbu@hgovc.mts.dec.com
    jennycong@hgom22.mts.dec.com
    
    I highly appreciate your help.. any more precise suggestions or
    alternates are welcome. 
    
    thanks in advance,
    Subbu
2065.5still panics...HGOV08::SUBBUThu May 22 1997 04:3117
    
    Hi
    
    We have changed the MC-cards, cable and PCI backplane too.
    
    Problems still persists. i.e. drd-data-compare=3 makes one 
    system to panic when we are accessing the data from remote
    node using oracle database.
    
    Could some one please let know your thoughts??
    
    If you need any crash dumps pls let me know.
    
    Thanks in advance.
    Subbu
    
    Note 2078.0 talks the same issue.
2065.6erronious hardware..HGOV08::SUBBUTue May 27 1997 05:169
    
    Thank You all for helping me in understanding the issue.
    
    Finally, the problem dis-appeared on changing MC cards
    and PCI backplane. It looks both are problematic here.
    
    Thank you once again,
    Subbu