what do you mean exactly by “the bucket sometimes down”? If the bucket is down, it’s probably true for other buckets in the same node (node failure rather than a strange bucket-level failure, in which case as @shiv4nsh stated what you probably need is node-level replication to ensure availability of your data.
Have a look at replica configuration.
Note: The Java SDK can optionally even read from a replica directly rather than waiting for the ops to failover a node in trouble.