[Linux-HA] Only active node goes down!

[Kevin Dwyer]

Andreas Semt wrote:
> Hello list!
> 
> my configuration:
> two nodes (testnode1 active / testnode2 standby) heartbeat (1.04) + drbd 
> on Debian Woody.
> 
> testnode1 eth1: 192.168.100.140 (public interface)
> testnode2 eth1: 192.168.100.141 (public interface)
> testnode1 eth0: 10.1.1.1 (private GBit interface for drbd + heartbeat)
> testnode2 eth0: 10.1.1.2 (private GBit interface for drbd + heartbeat)
> cluster virtual IP (VIP): 192.168.100.142 (for public access)
> 
> Both nodes are crossover connected per GBit Ethernet (for Heartbeat and 
> drbd) and a serial cable (for Heartbeat only). I have the ipfail in 
> ha.cf but without ping nodes.

As Lars said, please don't do this.  I don't see any reason why it 
shouldn't work, but it is stated very clearly in the documentation that 
you must configure some ping nodes in order to use ipfail.

Additionally, using a more recent version of heartbeat is advisable.

>  In my ha-log file (on testnode1) i can read following:
> 
>  --- snip ---
>  heartbeat: 2004/04/19_14:52:44 WARN: node testnode2: is dead
>  heartbeat: 2004/04/19_14:52:44 info: Dead node testnode2 held no
>  resources.
>  heartbeat: 2004/04/19_14:52:44 info: Resources being acquired from 
> testnode2.
>  heartbeat: 2004/04/19_14:52:44 info: Link testnode2:/dev/ttyS0 dead.
>  heartbeat: 2004/04/19_14:52:44 info: Link testnode2:eth0 dead.
>  heartbeat: 2004/04/19_14:52:44 info: testnode1 wants to go standby

What is ipfail saying at this time?  I believe you left out the ipfail 
messages which are likely in your syslog somewhere.

> Now i see in ha-log on testnode1 following:
> 
> --- snip ---
> heartbeat: 2004/04/19_15:10:39 WARN: node testnode1: is dead
> heartbeat: 2004/04/19_15:10:39 ERROR: No local heartbeat. Forcing shutdown.

No idea here.

-- 
- kpd