MASSIVELY PARALLEL AUTONOMOUS REIMAGING OF NODES IN A COMPUTING CLUSTER

摘要

Systems for management of a large computing cluster. A set of computing nodes within the cluster are identified. The identified nodes are to be reimaged with an operating system image and parameters. One or more reimaging algorithms in the form of computer code are stored in a networked storage device within a storage pool that comprises the networked storage device and a plurality of local storage devices that correspond to respective individual ones of the set of computing nodes. An agent that can execute the reimaging algorithm code is configured to run on any one of the set of computing nodes, wherein the agent is further configured to retrieve reimaging data from a storage device that is within the storage pool. An agent running on a leader node initiates reimaging of a second node. The second node initiates reimaging of a third node. The third node reimages the leader node.