[sheepdog-users] 'vdi check' no majority

Liu Yuan namei.unix at gmail.com
Thu Dec 19 08:36:44 CET 2013


On Thu, Dec 19, 2013 at 04:33:29PM +0900, Hitoshi Mitake wrote:
> At Thu, 19 Dec 2013 08:21:52 +0100,
> Valerio Pachera wrote:
> > 
> > [1  <multipart/alternative (7bit)>]
> > [1.1  <text/plain; UTF-8 (7bit)>]
> > 2013/12/19 Hitoshi Mitake <mitake.hitoshi at gmail.com>
> > 
> > > This message means that the corresponding objects which have ID
> > > 805ddf8800000000 don't have a majority. e.g. in a case of one lost
> > > object and one corrupted object in 3 replicated VDI can produce the
> > > above message.
> > >
> > 
> > Thank you for the answer.
> > In such case, what should the administrator do?
> 
> In such a case, administrators have to recovery corrupted objects
> manually (e.g. juddging a clean object and copy it by scp). "dog vdi
> track" is useful for such a case. The command can print which nodes
> have the objects on each epoch.
> 

Hitoshi,
	could you write a example in our wiki on how to fix the objects
manually in such cases? I guess not everyone is familiar with 'vdi track' command
and this is a good time to introduce it by example.

Thanks
Yuan



More information about the sheepdog-users mailing list