Methods and Systems for Rapid Failure Recovery for a Distributed Storage System

A method comprising:

• determining that a first backend of a distributed storage system has become unavailable;
• identifying one or more buckets previously managed by the first backend;
• selecting one or more second backends to receive the one or more buckets according to load information associated with the distributed storage system;
• transferring ownership of the one or more buckets to the one or more second backends;
• and
• initializing metadata for each transferred bucket before accepting input/output requests directed to the transferred bucket.

A distributed storage system comprising:

• a plurality of computing nodes;
• one or more storage devices;
• and
• one or more processors configured to:
• detect failure of a first backend,
• reassign one or more buckets previously managed by the first backend to one or more second backends,
• reconstruct metadata for each reassigned bucket using registry information and journal information, and
• enable servicing of requests for a reassigned bucket after completion of the metadata reconstruction.