[Veritas-vx] VVR and ShadowImage?

[Jim Senicka]

We support FlashSnap and VVR at primary site.
We make no support claims what so ever to the config you
are attempting.
Yes, if writes are in SRL and do not get snapped off at
secondary, this is what you are seeing


-----Original Message-----
From: veritas-vx [mailto:veritas-vx at mancinelli.mechanicalism.net]
Sent: Tuesday, February 17, 2004 1:13 PM
To: veritas-vx at mailman.eng.auburn.edu
Cc: Jim Senicka
Subject: RE: [Veritas-vx] VVR and ShadowImage?


Jim Senicka wrote:
> Assuming the SVOLs are at the primary site? I believe you would
> need to have VVR installed on each (if you are breaking off a mirror
> of a VVR config, you need VVR to do volume recovery on the off host
> systems).
> Have you tried this instead with FlashSnap?
 
Jim,
 
Yes, the PVOLs and the SVOLs are all within the same array at the
Primary site.  Because we are not licensed for VVR on the off host
systems, we do not shadow the SRLs and we have a script that Support
assisted with that blows away any and all VVR objects before we can
mount and fsck the volumes.
 
Looking at the "How Data Flows in VVR Asynchronous Replication" figure
(reproduced at http://foo.mechanicalism.net/~fortezza/vvr_async.png),
it does make sense that if you only shadow the data volumes and you go
to cut a shadow, there could be writes that have been committed to the
SRL and are waiting to be written to the data volumes.  If Oracle has
made certain assumptions on what has actually been comitted to disk,
cutting a shadow could cause ORA-600 errors.

Do you know if Veritas supports ShadowImage and VVR concurrently at the
Primary site?  If not, FlashSnap might be worth a try...
 
 
Best Regards,
 
--
Jason Fortezzo
fortezza at mechanicalism.net