AW: [LinuxFailSafe] Now ok for 1.04. But some question about sync and serial .....

[Martin Bene]

Hi Giulius,

> I would like to build a 2 node without shared storage disks=20
> cluster for HA my ftp server.
> All data file must be replicated between 1st and 2nd node.

You can use drbd as a "poor mans" substitute for shared storage. have a =
look
at http://www.linbit.com/en/ for more information and download of drbd. =
in
short, it provides block device replication over the network.

> Reading the AG, i found that failsafe need a serial line between 2 or=20
> more NODE in a CLUSTER.
> Is it true? If yes why? The only pourpose of a serial line is to shut=20
> down and doing remote management system?

No it's not true. You do, however, need SOME way to shut down / reset a =
node,
esp. if you're using shared storage (or drbd) to avoid data corruption.

The usual way of implementing these cluster reset services in linux is =
by
using the stonith library from the linux heartbeat project
(http://www.linux-ha.org/download/), which supports quite a few =
different
devices/methods.

For testing the software and experimenting with different setups, =
stonith can
use ssh to reset the other node - this is obviously not an option for a
production evironment, where something like the WPS power switches or =
APC
masterswitch or ... should be used.

Bye, Martin