Re: OPS and raw filesystems

From: Johnny Chan <j4ychan_at_PROBLEM_WITH_INEWS_GATEWAY_FILE>
Date: 1997/03/19
Message-ID: <5gpa2p$nt7@gw.PacBell.COM>#1/1

> Garry Robbins wrote:
> >
> > They are a requirement for SUN PDB clusters running OPS.
> >
> > Not only the data files, but control files and redo logs
> > must also be raw devices.
> >

> Are you sure that redo logs have to be raw devices ?

> Datafiles and control files are shared between instances, but redo logs
> are not. There should be no reason why redo logs have to be on raw
> devices.

> It may be better for performance but thats a different issue.

yes, the redo's have to be on raw devices because during instance recovery the surviving instance in an OPS cluster will need to access the dead instance's redo to rollforward to a read-consistent image of the db. And currently, the only way the files can be accessed by multiple nodes in a UNIX cluster env is via raw devices, not filesystems.

Johnny Chan
Independent Oracle Specialist Received on Wed Mar 19 1997 - 00:00:00 CST