[sheepdog] issue of collie after sheepdog upgrade

[Liu Yuan]

On Wed, Aug 07, 2013 at 11:32:54AM +0800, 陈李粮 wrote:
> Once ago we used version 0.56 for cluster testing and IO performance testing.
> Everything working well but some times losing data (all or some vdi losing) because of zookeeper time out.
> we found you have new version so we decide to upgrade.
> no issue or error while configuration and make install the stable-0.6 version.
> and we start the new sheepdog with this script,using -u option to upgrade store.
> mountdir=/cloud
> zookeeper="192.168.10.45:2181 192.168.10.46:2181 192.168.10.41:2181 192.168.10.47:2181"
> sheep  /home/gateway -u  -b 192.168.10.47 -y 192.168.10.47 -p 7000 -g 64 -z 4 -D -c zookeeper:${zookeeper}
> for i in {0..6}
> do
> sheep -d ${mountdir}/${i} -u -b 192.168.10.47  -y  192.168.10.47 -p $((7001 + $i)) -D -c zookeeper:${zookeeper}
> done
> 
> We skip the MD support and Shepherd support because we have not move the data.

No, you can't skip MD. MD support is always enabled since v0.6.0. With this
feature, you can hot-plug/unplug the disks on the local node. See more info at:
https://github.com/collie/sheepdog/wiki/Multi-disk-on-Single-Node-Support

Thanks
Yuan