Coordinator in cluster membership management protocol

A computer system, comprising:

• a processor in communication with a memory storing instructions that when executed by the processor provide:
• a coordinator for managing a membership of a cluster, the coordinator being configured to:
• when the coordinator is in a stable state:
• provide a member list to any node in a node list, nodes in the node list being or wishing to be members of the cluster, nodes in the member list being the members of the cluster;
• periodically determine if the node list is equal to the member list;
• and
• when the node list is not equal to the member list, advance to a reconfiguration state to change the membership of the cluster;
• when the coordinator is in the reconfiguration state:
• send a reconfiguration request to the nodes in the node list, the reconfiguration request including a unique change ID;
• when reconfiguration acknowledgements with the unique change ID are received from all the nodes in the node list within the timeout period:
• update the member list to be equal to the node list;
• persist the updated member list;
• after the updated member list is persisted:
• send a reconfiguration confirmation to all the nodes in the node list, the reconfiguration confirmation including the unique change ID and the updated member list;
• and
• return to the stable state.

A non-transitory, computer-readable storage medium encoded with instructions executable by a processor to implement a coordinator to management a membership of a cluster, the instructions comprising:

• when the coordinator is in a stable state:
• providing a member list to any node in a node list, nodes in the node list being or wishing to be members of the cluster, nodes in the member list being the members of the cluster;
• periodically determining if the node list is equal to the member list;
• when the node list is equal to the member list, remaining in the stable state;
• and
• when the node list is not equal to the member list, advancing to a reconfiguration state to change the membership of the cluster;
• when the coordinator is in the reconfiguration state:
• sending a reconfiguration request to the nodes in the node list, the reconfiguration request including a unique change ID;
• when reconfiguration acknowledgments with the unique change ID are not received from all the nodes in the node list within a timeout period:
• sending an abort message with the unique change ID to all the nodes in the node list;
• and
• returning to the stable state;
• and
• when the reconfiguration acknowledgements with the unique change ID are received from all the nodes in the node list within the timeout period:
• updating the member list to be equal to the node list;
• persisting the updated member list;
• after the updated member list is persisted:
• sending a reconfiguration confirmation to all the nodes in the node list, the reconfiguration confirmation including the unique change ID and the updated member list;
• and
• returning to the stable state.