This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
Home Products Support Community News


[Xen-users] xen cluster with heartbeat and resource fencing (SFEX)

To: xen-users@xxxxxxxxxxxxxxxxxxx
Subject: [Xen-users] xen cluster with heartbeat and resource fencing (SFEX)
From: Massimiliano De Ruosi <max.deruosi@xxxxxxxx>
Date: Wed, 01 Jul 2009 15:49:57 +0200
Delivery-date: Wed, 01 Jul 2009 06:50:37 -0700
Envelope-to: www-data@xxxxxxxxxxxxxxxxxxx
List-help: <mailto:xen-users-request@lists.xensource.com?subject=help>
List-id: Xen user discussion <xen-users.lists.xensource.com>
List-post: <mailto:xen-users@lists.xensource.com>
List-subscribe: <http://lists.xensource.com/mailman/listinfo/xen-users>, <mailto:xen-users-request@lists.xensource.com?subject=subscribe>
List-unsubscribe: <http://lists.xensource.com/mailman/listinfo/xen-users>, <mailto:xen-users-request@lists.xensource.com?subject=unsubscribe>
Sender: xen-users-bounces@xxxxxxxxxxxxxxxxxxx
User-agent: Internet Messaging Program (IMP) H3 (4.1.3)
Here is my (very experimental) setup:

- an heartbeat cluster with 2 nodes, with the standard RA provided for Xen.
- Xen kernel and hypervisor on every node, installed on local disks
- a SAN shared disk (with ocfs2) for domU's disk images (sparse files).
- Resource fencing with SFEX.

All with standard Debian Lenny packages.

My choice for SFEX (compared to STONITH) for fencing is due to this fact: the only shared data are the domU disk images, an the only way they can be corrupted in a split-brain scenario is when 2 domUs are started on both nodes at the same time. THis can be elegantly avoided using sfex, with the check of a shared disk, that works even if every communication between the two nodes is broken.

to obtain this, I've grouped every Xen resource with a correspondent sfex one.

Everything works fine, until I try to live-migrate a domU resource from node A to B: it seems that sfex causes the domU shutdown on node A, and restart to the B (instead of live migrate from A to B).

If I use only standard RA for Xen instead (without sfex), the domU is correctly live-migrated from node A to B, without shutdown.

Any suggestion (or any working cluster setup to recommend)?

Massimiliano De Ruosi

SEMEL (SErvizio di Messaging ELettronico) - CSIT -Universita' di Udine

Xen-users mailing list

<Prev in Thread] Current Thread [Next in Thread>
  • [Xen-users] xen cluster with heartbeat and resource fencing (SFEX), Massimiliano De Ruosi <=