IP Fail Over-Back Prog
jalmasi at partner-banka.tel.hr
Wed Sep 9 05:18:05 MDT 1998
> What I thought was-
> On Primary Server. At boot up First look for weather Master IP is
> existing or not ->If no ->take HDD ->create Alias for Master IP ->
> Wait & check for some kind of signal ->if that is there OK(means I am
> On Secondary Server. At boot up First it will check for Master IP->if
> yes-> then fork Sniffer to Check NETWORK and not some server(here I
> guess the eth0 of Secondary is loaded which in any case was sitting
> IDLE) ->Tell Primary that it's working bye just giving some kind of
Did you try to mount shared disks on both machines? If that is possible,
complete cluster software could be located on shared disks, and so could be
cluster configuration. Then, each node should just mount shared disks read-only
and start cluster software with node's priority level as parameter... result is,
on any server:
1. Mount shared disks read-only and start, say, rc.cluster with a parameter.
Parameter is node's id within cluster, may be priority level (0 = default
master). Note that rc.cluster is found on shared disks.
2. Determine available nodes and their priorities.
3. If node has higher priority than other nodes, start takeover. Otherwise, just
join the cluster, whatever that means...:)
1. Determine is there a current master node.
2. If there is, tell him to go slave and wait for response.
3. Remount shared disks as read-write.
4. Reconfigure network cards.
5. Start some applications, if needed.
Note that prefered master node is not master node, and prefered master ip
address is different than master node's ip address. Than, each cluster member's
ip address is written somewhere in cluster config on shared disks, and so is
master node's ip address (this is actualy cluster's ip address).
> Also when I tried last time(first aswell) I was unable to get to
> Master IP using eth1(heartbeat). Am going to try again. Please tell me
> the configuration & the route's to get packets reaching Master-IP
> using heartbeat to check for it's working. Moreover do we plan to
> create heavy traffic! My personal thinking was to use eth1 as backup
> of eth0
> eth0 = ??.0.0.20/255.0.0.0
> eth0:0 = 188.8.131.52/255.0.0.0
> eth1 = ??.0.0.20/255.0.0.0
route add ??.0.0.30 netmask 255.0.0.0 eth1
> eth0 = ??.0.0.30/255.0.0.0
> eth1 = ??.0.0.30/255.0.0.0
route add ??.0.0.20 netmask 255.0.0.0 eth1
You may add those ip's to /etc/hosts on both machines.
More information about the Linux-HA