OT: Re: Can we use database with high availability in LVS cluster?

To: Jeremy Hansen <jeremy@xxxxxxxxxxxxx>
Subject: OT: Re: Can we use database with high availability in LVS cluster?
Cc: Horms <horms@xxxxxxxxxxxx>, Madhav <msp@xxxxxxxxxxxx>, lvs-users@xxxxxxxxxxxxxxxxxxxxxx, linux-ha@xxxxxx
From: Stephen Zander <gibreel@xxxxxxxxx>
Date: 31 Mar 2000 17:15:37 -0800
>>>>> "Jeremy" == Jeremy Hansen <jeremy@xxxxxxxxxxxxx> writes:
    Jeremy> In my experience with Linux Oracle, it supports all the
    Jeremy> standard Oracle functions, replication, hot standby, etc.
    Jeremy> To me a gfs situation with Oracle would be an ideal
    Jeremy> situation.  Put Oracle data files on the shared array and
    Jeremy> just start oracle on the standby machine if the primary
    Jeremy> fails, although there amy be something with Oracle that
    Jeremy> wouldn't allow this.

Nope, you just have to endure the overhead of failure recovery before
the secondary system is available.  You would, however, need the
primamry box to go down or at least dismount the f/s or you have
issues with disk cache flushing on the non-active node.  AFAIK, Oracle
doesn't support CODA.  It certainly doesn't support NFS and anything
else is probably more easily done with raw devices and direct physical
access to the media (no f/s abstraction). I've not yet come across the
Linux equivalent of Veritas under Solaris, though, so I'm not sure how
you'd go about managing access.  SCSI sub-systems can also get very
unhappy about multiple host initiators on a single bus.

    Jeremy> I've seen Oracle run in hot standby mode with Linux, but
    Jeremy> in this mode it requires you to eventually rebuild and
    Jeremy> repopulate the standby server, or you can go back and
    Jeremy> forth, the standby because the primary and the original
    Jeremy> primary then becomes the standby.

You can also use replication and snapshots.  It all depends on the
data aceess and availability you need to support.

    Jeremy> I also think there something with Oracle 8i that allows
    Jeremy> the standby to be mounted in read only mode which makes
    Jeremy> things easier when restoring the primary.

Yep, you can run a system in the moral equivalent of permanent
recovery mode.  You will still have overhead to deal for failure
recovery.  This mechanism is more for disaster recovery though.
Shared disk is faster but useless if there's a catastrophic failure at
the site (power outage/earthquake/act-of-god).


"If I claimed I was emporer just cause some moistened bint lobbed a
scimitar at me they'd put me away"

<Prev in Thread] Current Thread [Next in Thread>