[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Fault tolerant NAS
- To: misc_(_at_)_openbsd_(_dot_)_org
- Subject: Re: Fault tolerant NAS
- From: Chuck Yerkes <chuck+obsd_(_at_)_2004_(_dot_)_snew_(_dot_)_com>
- Date: Wed, 21 Apr 2004 01:49:27 -0400
Quoting Matt Provost (mprovost_(_at_)_termcap_(_dot_)_net):
> On Apr 20 09:09 AM, Balazs Nagy wrote:
> > Hi there,
> >
> > Has anyone used OpenBSD to create a fault tolerant NAS?
> >
> > It seems that carp could handle the front end, but how
> > to replicate between two SCSI or Fiber systems? I'm
> > thinking of a hardware architecture like here:
> > http://www.winsys.com/products/flnas_confi_nspfserv.php
> >
> > Cheers,
> > Bal?zs
> >
> > http://www.thenewpush.com
>
> I assume you're talking about NFS in this example. Although carp can
> move the IPs around, I think you'll find that an NFS server isn't quite
> as stateless as you'd like. Any failover NFS implementation that I've
> seen requires some syncing of filehandles between the servers, at least.
> You'd have to have something like pfsync for NFS before the clients
> would fail over without noticing. I think if you'd try it now you'd just
> get a lot of stale file handle messages after a failover.
Actually (at least for NFSv2) it *is* stateless.
And if you can toss a RAID array (or disk) from one server to
another (a non-trivial matter to arrange well), and you present
the same MAC/IP address to the network, clients won't notice a
failover.
Visit your host, monkey.org