Sorry for asking again for an easy way for redundant
xen images. After reading some threads and numerous howtos i'm a bit confused.
I have two xen hosts, each of them has one disk and should
be the host for several domU's. Live migration should be supported. If one host
fails, it should be possible to boot the affected virtual machines on the other
host, having the last written data available.
It seems to be the best way to mount the mirrored/replicated
disks in both dom0's, create virtual groups using lvm and export logical
devices to the domU's, one logical device for each virtual machine.
Maybe it's a good idea then to use clvm to make
changes of the virtual groups possible at runtime. Is there any easy to install
and configure alternative to clvm?
What is an easy way of mirror and mount both disks in
dom0?
As far as i could see, the best ways are:
A) DRBD / GNBD / Some kind of shared FS
OR
B) Export all drives using ISCSI / Create software
raid of the same disks in both dom0's
TO A):
Can DRBD handle and replicate simultaneous writes to
both drives in primay/primary mode?
Does the use of any shared FS always needs a
server/client structure? Defining one of the hosts as the filesystem server
creates a single point of failure... Is this the reason for using heartbeat (to
switch the server in case of failure)? Doesn’t sound like an easy
setup...
How are writes to images handled at a client of the
shared fs? Are they sent to the server and then mirrored back by DRDB? Sound's
a bit like waste of ressources, doesn't it?
Can You recommend a good howto for such a setup?
TO B)
Is it possible to have independent software raids of the
same both disks in both dom0's? Raid on host A doesn't recognizes writes to the
disks from host B, doe's it? Maybe i misunderstood the setup...
Can a software raid be configured to read only from
the local disk and write to all disks (to improve performance)?
Can You recommend a good howto for such a setup?
Can You recommend any other setup?
Thank You,
Marc
