[Sheepdog] Configuring Sheepdog environment

[MORITA Kazutaka]

At Mon, 20 Jun 2011 00:12:59 -0300,
Gustavo Callou wrote:
> Dear Sheepdog developers,
> 
> I’m a Brazilian PhD student from Federal University of Pernambuco and I'm
> trying to configure the Sheepdog in order to test its features. I have 4 PCs
> available to do such test, in which Ubuntu 10.04 64bits version are
> installed. I'm adopting the Sheepdog package
> “sheepdog_0.2.2-0ubuntu1_amd64.deb“, downloaded at Ubuntu repositories.
> However, I’m not sure about the configuration due to the fact that once I
> have installed the environment, machines sometimes are able to see each
> other and sometimes not. What I mean is when I execute the command “sudo
> collie node list”, there are machines that see that there is only one node
> in the cluster (i.e., the host sees itself), others can see 2 nodes, and
> other sometimes see 3 nodes. Another interesting fact is that, when I have
> restarted the corosync and the sheepdog in all machines, such seen
> configuration changes.
> 
> In a nutshell, I have attached in this message the logs from 3 machines
> (e.g., named sheep1, sheep2, sheep3). Besides, I attached the corosync.conf
> in which I have tried to configure the bindnetaddr with an ip like this
> 172.20.10.0 as well as adopting the machine ip such as 172.20.10.116.
> 
> Please, let me know as soon as possible if there is something that is not
> correctly configurated at the corosync.conf or any other issues that you may
> detect through the nodes logs.

Your corosync.conf looks correct.

From sheep[1-3].log, your Sheepdog directories contains wrong
information about node membership histories.  I'm not sure why it
happened, but it would work for you to clean the store directories and
run 'collie cluster format' again.

I highly recommend to upgrade Sheepdog to version 0.2.3 or use the
latest development code:
  https://github.com/collie/sheepdog

It includes many important bug fixes.

Thanks,

Kazutaka