From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mark Fasheh Date: Wed Sep 6 09:15:11 2006 Subject: [Ocfs2-devel] heartbeat. In-Reply-To: <44FE73D5.2020602@seanodes.com> References: <44FE73D5.2020602@seanodes.com> Message-ID: <20060906161507.GD8792@ca-server1.us.oracle.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: ocfs2-devel@oss.oracle.com Hi Mathieu, On Wed, Sep 06, 2006 at 09:08:05AM +0200, Mathieu Avila wrote: > Hello OCFS2 team, > > I'm currently looking at the OCFS2 code in linux-2.6.17.11, and i wander > why OCFS2 performs its heartbeat on a disk region unlike on the network > like many clustered services stack do. What is the requirement for a > disk heartbeat ? Is there any way to tune this behaviour and change it > into a network heartbeat ? The OCFS2 cluster stack heartbeats on both disk and network. Disk is generally thought to have the final say as to whether a node is alive or not. The reason we heartbeat on disk is because we want to be sure a connection to the disk is maintained. Otherwise, the cluster could hang indefinitely if the disk cable from one mounted node is unplugged - it wouldn't be able to write out shared meta data for other nodes to read. That all said, we're likely to remove the disk heartbeat in favor of a read only mechanism in a future release of OCFS2. --Mark -- Mark Fasheh Senior Software Developer, Oracle mark.fasheh@oracle.com