Should slave datanodes are not up, check if namespace ids are compatible.

If not, delete /app/hadoop/tmo (or generally, dfs.tmp.dir) and format namenode again!