|
> In my system. some problems can be seen in the condition of
> "nice_failback on", as mentioned below.
You've already noted the syntax change, but just wanted to stress it :).
http://wiki.trick.ca/linux-ha/ha.cf/AutoFailbackDirective
> 1)eth0:0 comes up on the both Active and Backup Servers.
Sounds like they lost communication and went split-brain. If you want
to find out exactly what went wrong you should post your ha-logs and
configs to the linux-ha list. I'm positive there are split-brain
scenarios that were fixed from 1.04 -> 1.2, so that's the option I would
choose.
> 2)Gratuitous arp are not broadcasted when failover.
> (I checked many times using tcpdump)
There are a few ARP bugs fixed in 1.2. I know it broadcasts on my 1.2,
but I'm pretty sure it did for my 1.04 too.
> To confirm whether it is my mis-configuration, I had tested
> same config
> files on my friend's system (heartbeat version 1.2.0-4,
> auto_failback off). Actually, it works quite well.
1.2 seems more stable to me in my usage. I think Alan Robertson and
Horms (builds Debian packages) stressed that this version was
recommended.
> Are these problems are typical in heartbeat 1.0.4?
> Is there any solution other than update heartbeat?
> Should I update it at any cost?
Yes, there are a few problems with 1.0.4. I don't know about any cost.
But in your situation -- split brain and arp -- it would make sense to
upgrade. There was definitely at least one split-brain bugfix, and
there were many others such as arp fixes. I think if you look at the
linux-ha changelog you will find the details.
http://www.linux-ha.org/download/ChangeLog
P
|
