Howto setup 3-node heartbeat cluster

This probably isn't a big help, but this has been discussed extensively of late over at the Pacemaker and Linux-HA mailing lists.

I'm not very good with heartbeat, but with pacemaker I would set up a constraint that caused the cluster resource manager to flush disks with a write lock to the disk (or down mysql temporarily) before trying to switch over, and then releasing the lock once the switch had been completed.

i assume you're interested in simple active-passive setup.

ucarp & heartbeat in such setup do pretty much the same thing. in essence - they run provided scripts when machine is elected to be master / hot-standby.

heartbeat might look much more complicated [ since it can help you autoamte drdb mounts, restarting multiple services etc ] but at the end - you can script all of this and let ucarp invoke it].

personally - i run heartbeat with single resource - that is script that does following:

• [un]binds appropriate ip address
• runs couple of arp-broadcasts
• starts[stops] required services

my very simplistic setup [ heartbeat 2.1.3-6 under debian lenny ]: i have two servers:

• ser0 [preferred active node] with permanently assigned 10.0.0.2/24 at eth0
• ser0b [hot-standby node waiting to replace master] with permanently assigned 10.0.0.3/24 at eth0

'floating ip' - assigned to the active node is 10.0.1.1/24 assigned to eth1

in this case service that gets high availability is apache. i separately sync apache's configs and content that is served from ser0 to ser0b.

files below are identical on both machines with one marked exception:

/etc/ha.d/authkeys:

auth 1
1 md5 somethingrandom

/etc/ha.d/haresources

ser0 ha.sh

/etc/ha.d/ha.cf

keepalive 2
deadtime 10
udpport        694 
; below - address permanently assigned to the peer node . this is for master:
ucast eth1 10.0.0.3
; and on slave i have 
; ucast eth1 10.0.0.2
udp     eth0
logfacility     local0
auto_failback on

node    ser0
node    ser0b

/etc/init.d/ha.cf [ it can as well be in /etc/ha.d/resources.d/ha.cf ]

#!/bin/bash
case "$1" in
  start)
        ip link set dev eth1 up
        # bind 'floating' ip to the interface
        ip a a 10.0.1.1/24 dev eth1
        # you might want to add some route-changes here if needed
        /usr/lib/heartbeat/send_arp -r 10 eth1 10.0.0.1 auto 10.0.0.255 255.255.255.0
        # to make sure apache reloads it's config when machine becomes master
        /etc/init.d/apache2 restart
  ;;

  stop)
        # we are no longer active, un-bind 'floating' ip from the interface
        ip a d 10.0.1.1/24 dev eth1
        # you could stop it as well or just skip this step
        /etc/init.d/apache2 restart
  ;;
esac
exit 0