[sheepdog] segfault

[Valerio Pachera]

2012/12/7 Liu Yuan <namei.unix at gmail.com>:
>> re-format is the last resort. You can try 'collie node kill' the node
>> one by one and start up the nodes in a reverse order.

Does the order matter?
I tried, but no changes.

> Or you can run 'collie cluster recover force' to force the current nodes
> into a running state. I recommend this approach.

# collie cluster recover force
'recover' is not a valid subcommand
'cluster' supports the following subcommand:
info
format
shutdown

I also redownloaded and recomplied sheep now on all 3 nodes to be sure
to have the exat same version.