[Linux-cluster] IP Relocate Error / IP Restart error

[Lon Hohberger]

On Tue, Jul 10, 2007 at 04:50:44PM +0200, Dan Deshayes wrote:
> Jul 10 16:22:32 asl012 clurgmgrd[412]: <notice> Stopping service 
> service:www-project1
> Jul 10 16:22:32 asl012 clurgmgrd: [412]: <err> script:psql-db: stop of 
> /etc/init.d/postgresql failed (returned 1)
> Jul 10 16:22:32 asl012 clurgmgrd[412]: <notice> stop on script "psql-db" 
> returned 1 (generic error)

for this one:

http://sources.redhat.com/cluster/faq.html#rgm_wontrestart

> I'm using bond0 interface configured the following:
> DEVICE=bond0
> USERCTL=no
> ONBOOT=yes
> BROADCAST=<broadcast>
> NETWORK=<network>.32
> NETMASK=255.255.255.224
> IPADDR=<external ip 1>
> GATEWAY=<gw ip>

Ah - rgmanager won't start the *interface* for you.  You need an
already-running IP address on bond0 that is different from the one
rgmanager is trying to start.

So, if you have IPs on:

node1 192.168.1.2/24
node2 192.168.1.3/24
node3 192.168.1.4/24

You can have the IP address be on the 192.168.1.0/24 subnet, but not on,
say, 10.1.2.0/24.

That is, externally-facing IPs on the same subnet as the one you're
trying to manage need to be up (even if firewalled off) so rgmanager
can determine the device to use.  This is done so that different nodes
can use different devices (e.g. eth1 on 192.168.1.0/24 on nodes 1,2 but
bond0 on the same network on node 1...).

So, basically, you need a device on each node to already have an
external IP bound which is in the same subnet as the IP you're trying to
move between nodes.

-- 
Lon Hohberger - Software Engineer - Red Hat, Inc.