i'm not quite sure if i got your question correctly but you can use the STONITH approach to "officially" degrade the master node from a failover cluster to prevent both nodes from accessing an exclusive shared resource (can't find my old post in the PLUG archives)....
it can be as simple as a watchdog device or you can disable the other node by disabling its corresponding port on the switch that is being used to access the shared resource or by disabling the AC outlet on the UPS power strip (overkill?), both via SNMP (only if the device supports it). however, i haven't used STONITH to protect a shared resource that requires exclusivity in R/W mounting since the nodes that i've handled/handling have their own storage or the SAN uses a "cluster-aware" filesystem wherein multiple hosts can mount the same block device. but i did use/using it to degrade an "ailing" node, a machine that's not really dead but no longer responds properly. you may not want to just perform a hard reset on the ailing node especially if the problem is a recurring one. you can't imagine the machines to keep on switching from one another because that alone accumulates a downtime. :) the other approach to handle a failover is via a load-balancing cluster (sayang daw kasi kuryente kapag hindi productive yun isang node. you know, global warming issue. hehe) which i still have to see some linux-ha _director_ deployments since most of which i've seen recently uses a turn-key solution from F5 networks. On 2/15/07, Orlando Andico <[EMAIL PROTECTED]> wrote:
It is not immediately obvious to me how to avoid split-brain without a shared storage device.
_________________________________________________ Philippine Linux Users' Group (PLUG) Mailing List [email protected] (#PLUG @ irc.free.net.ph) Read the Guidelines: http://linux.org.ph/lists Searchable Archives: http://archives.free.net.ph
