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

Conference turris::digital_unix

Title:DIGITAL UNIX(FORMERLY KNOWN AS DEC OSF/1)
Notice:Welcome to the Digital UNIX Conference
Moderator:SMURF::DENHAM
Created:Thu Mar 16 1995
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:10068
Total number of notes:35879

8953.0. "DECsafe & Oracle Replication: compatible?" by NETRIX::"[email protected]" (jjosol) Tue Feb 25 1997 09:42

Hi,

I need to validate my position that Oracle Replication (or any other
replication
from other RDBMS vendor) is not compatible with a DECsafe solution. Either
you implement Replication or you do it via ASE on the LAN.

Customer wants to do on the LAN what he is doing over WAN on an ASE
environment. Two sets of storage are shared by two servers. A-server uses/
owns A-storage. At the same time, A-server replicates data to the B-server
on its B-storage. Looks fine until a failover scenario is invoked. The Oracle
guys insists that prior to failover BOTH servers will mount their storage 
on an identical mount point name. Presto! The failover disables the service
in B-server with its B-storage and B-server takes the A-storage. Replication
falls apart as the B-storage fails to mount on a used mount point.

Is there a better way for this customer to do what he envisions with the
availability of his data over ASE? This big telco is so paranoid about losing
his data that he wants ASE to fit in the solution which is an unsupported
cconfiguration. Any thoughts?

/joeljosol
[Posted by WWW Notes gateway]
T.RTitleUserPersonal
Name		DateLines
8953.1http://www-unix.zk3.dec.com/www/dt.htmlHGOVC::JOELBERMANThu Feb 27 1997 01:1129
    Joel,
    A bit of apples and oranges.
    
    ASE is to make sure that the CPU and network resources are not single
    points of failure.
    
    Replication, in your instance, is to ensure that the DB Engine and site
    is not a single point of failure.
    
    For a single site it is probably sufficient to use disk mirroring and
    ASE.
    
    For multiple sites, replication can provide a good means of (almost)
    synchronizing the data base. And if one site dies the other can take
    over.  
    
    For your case it sounds like TruCluster with OPS is the best.  That way
    both boxes can access the data base, and if either side goes down, the
    other side can take over.
    
    A common scenario is to have two ASE's, one at each site.  Then the
    sites replicate between themselves.  If you lose a box, the ASE pair
    takes over, if you lose a site, the other site takes over.
    
    A good place to start with this is at
    http://www-unix.zk3.dec.com/www/dt.html
    
    /Another Joel
8953.2MANM01::JOELJOSOLFri Feb 28 1997 20:3634
    Joel,
    
    But the customer is paranoid after the AdvFS corrupted the Oracle files
    when a KZPSA went bad. Telco customer wrote off as bad debts the
    billing files that got lost and were no longer reconstructable. You
    should have been in the customer meeting after that happen and feel
    the heat of the implications of the incident.  Imagine that a hw
    failure creeping up to AdvFS level!
    
    So, he is no longer confident with disk mirroring over HSZs and KZPSAs.
    What with cpio, tar, and vdump not working with his GBs of Oracle files
    to backup.
    
    What Oracle did is to ship archive logs across the other system. Only
    one ASE service is online - the one that holds the production data.
    One other storage holds a copy and receives the archive log from the
    other system. This is not a service. It's just ordinary mounted AdvFS.
    They haven't opted for full Replication it turns out on the ASE site.
    They are in fact doing this for the other sites.
    
    The conflict is in that the ASE service on the active system and the
    ordinary AdvFS mounted fs from the idle server both uses a common
    mount point. So, when a failover happens, the ASE fails to do its
    job precisely because the idle server is using the mount point.
    I need help on constructing a script to manage this transition
    in different failover scenarios - dismount the idle server's mount
    point, start service on that freed up mount point; when the other
    server comes back, mount the non-ASE service AdvFS on it - basically
    exchange roles when a failover happens.
    
    /joeljosol