[sheepdog-users] Force cluster to start with a missing node

Liu Yuan namei.unix at gmail.com
Tue Jan 15 04:04:36 CET 2013


On 01/14/2013 06:52 PM, Valerio Pachera wrote:
> 2013/1/12 Liu Yuan <namei.unix at gmail.com>:
>> How did you create the vdi? It seems that object cache isn't flushed.
>> you can use 'vdi cache flush xxx' to assure vdi is flushed.
> 
> I repeated the test.with a fresh formated cluster (3 copies + quorum)
> and a single vdi.
> No cache option are give to sheep.
> The guest is not running.
> I stop the cluster and run 3 nodes of 4.
> Node 2 doesn't show the vdi.
> If iI run I vdi check (once recovery is completed) i get lots of error.
> 

Whatever mode you format the cluster, if the re-run nodes are less than
the number of nodes when the cluster is shutdown, you need force a
recovery to cluster to assure data are right placed at first by:

$ collie cluster recover force

Thanks,
Yuan



More information about the sheepdog-users mailing list