drbd (was nmbd)

Alan Robertson alanr at bell-labs.com
Mon Nov 8 07:48:57 MST 1999


"Stephen C. Tweedie" wrote:
> 
> Hi,
> 
> On Mon, 8 Nov 1999 08:54:02 +0100, Philipp Reisner <kde at ist.org> said:
> 
> > I am not going into solving this issue, because this is somewhat beyond
> > the socpe of drbd.
> 
> You *must* at least prevent it from happening (perhaps by making the
> raid array readonly if there is only one disk present).  Otherwise, you
> will get data corruption or data loss when the network is restored.
> 
> > I will rather concentrate on issues making drbd a viable choice for
> > real world applications.
> 
> If you ignore cluster partition then it will never be a viable choice.
> You don't have to solve the problem, but you do at least need to prevent
> it from happening in some way.
> 
> For applications such as web or ftp servers, going readonly on a cluster
> fault is perhaps quite acceptable.

He can always delegate it to the surrounding/controlling software (like
heartbeat), and call it "somebody else's problem" ;-).  This of course, doesn't
eliminate the need for *someone* to deal with it.

	-- Alan Robertson
	   alanr at bell-labs.com



More information about the Linux-HA mailing list