[sheepdog-users] Couple of Questions
namei.unix at gmail.com
Sun Dec 22 06:15:38 CET 2013
On Sat, Dec 21, 2013 at 07:37:46PM +0000, Brendan Johnson wrote:
> I am new to Sheepdog. I just setup a 10 node Sheepdog cluster and I have a couple of questions:
> 1. Currently each server has one disk partition mounted at /var/lib/sheepdog. If I want to use more than one partition do I run one sheepdog process per disk partition?
No, a single sheep daemon can manage intelligently unlimted disks.
Please see https://github.com/sheepdog/sheepdog/wiki/Multi-disk-on-Single-Node-Support
> 2. What is the best way to configure remote machines to access the Sheepdog cluster? Simply pointing the remote machines to one of the nodes would work but if that node goes down the remote machine will lose access to the Sheepdog cluster even though the cluster itself hasn't gone down. I was thinking haproxy might be used to provide HA and load balance requests between the Sheepdog nodes.
haproxy looks goode for HA. We also have implementated a auto-reconnection mechanism
in QEMU that connect to sheep daemon. This means, if the sheep dies for whatever
reason, you just need to restart the sheep daemon. The disconnected VM will try
to connect sheep daemon automatically and work as if there were nothing happened.
You need QEMU v1.7 and plus for this feature.
More information about the sheepdog-users