[sheepdog-users] re-start a cluster using md

Liu Yuan namei.unix at gmail.com
Fri May 3 15:50:23 CEST 2013


On 05/03/2013 09:37 PM, Valerio Pachera wrote:
> My testing cluster has born using a single disk.
> Now I have two nodes using 2 disks by (md).
> I stopped the cluster.
> This is what I see in the first disk of node 4:
> 
> root at test004:~# ls -lg /mnt/sheepdog/
> totale 1080
> -rw-r----- 1 root      40 apr  9 18:36 config
> drwxr-x--- 2 root      51 mag  3 14:31 epoch
> -rw-r----- 1 root       0 apr  9 18:30 lock
> drwxr-x--- 4 root   40960 mag  3 15:17 obj
> -rw-r--r-- 1 root 1031019 mag  3 15:30 sheep.log
> srwxr-xr-x 1 root       0 mag  3 14:28 sock
> 
> The second disk (/mnt/sdb1), contains only the object files.
> 
> To start sheep daemon, the right syntax is
>   sheep /mnt/sheepdog,/mnt/sdb1
> 
> Right?

No, you have to stick the multi-disk setup command format

  $ sheep /meta-store,/disk1,/disk2...

So in your case with multiple disks node, you need

 $ sheep /mnt/sheepdog,/mnt/sheepdog/obj,/mnt/sdb1

The first component is the path to the meta-store, which was created at
/mnt/sheepdog/ when you started it with a single store path.

Actually, for a single store setup, A and B both are equal
 A $ sheep /mnt/sheepdog
 B $ sheep /mnt/sheepdog,/mnt/sheepdog/obj

Thanks,
Yuan


More information about the sheepdog-users mailing list