[sheepdog-users] The right way to start the cluster

Liu Yuan namei.unix at gmail.com
Mon Jan 7 16:15:51 CET 2013


On 01/07/2013 10:04 PM, Valerio Pachera wrote:
> I stop the cluster because I wish to update sheepdog on all nodes.
> Once done, to run sheep on all nodes, I may use something like parallel-ssh.
> This will run sheep almost at the same time on all nodes and in a random order.
> 
> May this cause problems?
> Is it better to run sheep daemon one after the other and wait few
> seconds before running the next?

No. Concurrent start up is gracefully handled internally by sheepdog
cluster drivers. If some nodes fail to join the cluster due to some
reasons (for e.g, epoch mismatch), then users need simply re-join the
failed node.

Thanks,
Yuan




More information about the sheepdog-users mailing list