Granular replica healing for distributed databases

A storage node in a distributed database configured to heal compromised data items stored in the distributed database, the storage node comprising:

• a metadata manager configured to locally store metadata in a local data storage for a plurality of data items stored in the distributed database, the metadata including, for each data item, an identifier associated with the data item and an associated replica location list indicating one or more storage nodes in the distributed database storing replicas of the data item;
• and
• a healing manager configured to:
• iterate, responsive to a trigger indicating a fault in a second storage node, over the metadata in the local data storage to autonomously identify, based on the replica location list for each data item, compromised data items for which the first storage node is designated as a healing node;
• and
• perform a healing procedure for each compromised data item where the first storage node is designated as the healing node.

A storage node in a distributed database configured to heal compromised data items stored in the distributed database, the storage node comprising:

• communication circuitry configured for communication with client devices and with other storage nodes in the distributed database;
• data storage for storing replicas of data items;
• processing circuitry configured to:
• locally store, in the data storage, metadata for a plurality of data items stored in the distributed database, the metadata including, for each data item, an identifier associated with the data item and an associated replica location list indicating one or more storage nodes in the distributed database storing replicas of the data item
• iterate, responsive to a trigger indicating a fault in a second storage node, over the metadata in the data storage to autonomously identify, based on the replica location list for each data item, compromised data items for which the first storage node is designated as a healing node;
• and
• perform a healing procedure for each compromised data item where the first storage node is designated as the healing node.