System and/or method for maintaining highly-available, consistent, partition-tolerant clusters using client voters

A distributed computing system, comprising:

• a plurality of servers organized in a cluster, each said server including first processing resources including at least one first processor and a first memory;
• a plurality of client devices operable to connect to the cluster and issue requests to the servers, each client device including second processing resources including at least one second processor and a second memory;
• wherein one of the servers is elected as a leader and is responsible for maintaining consensus information among the other servers, the consensus information including data indicating a shared operating state of the servers in the cluster, which server in the plurality of servers is elected as the leader, and which client device(s) is/are eligible to vote;
• wherein the shared operating state relates to requests issuable by the client devices;
• and
• wherein the first processing resources of each server are configured to control the respective server to determine whether a new leader election is to take place and, in response to a determination that a new leader election is to take place:
• request votes for a new leader from the other server(s) in the plurality of servers;
• receive any votes sent from other server(s);
• request votes for a new leader from the client device(s) that is/are eligible to vote as indicated in the consensus information;
• receive any votes from the client device(s) that is/are eligible to vote;
• determine whether the respective server has won the election by receiving a clear majority of votes cast, including votes cast by the client device(s) that is/are eligible voters;
• and
• responsive to a determination that the respective server has won the election with a clear majority of votes cast, including votes cast by the client device(s) that is/are eligible voters, distribute consensus information to the other server(s) about the respective server now being the leader.

A method of managing a distributed computing system including a plurality of servers organized in a cluster and a plurality of client devices operable to connect to the cluster and issue requests to the servers, each said server including first processing resources including at least one first processor and a first memory, each client device including second processing resources including at least one second processor and a second memory, one of the servers being elected as a leader and being responsible for maintaining consensus information among the other servers, the consensus information including data indicating a shared operating state of the servers in the cluster, which server in the plurality of servers is elected as the leader, and which client device(s) is/are eligible to vote, wherein the shared operating state relates to requests issuable by the client devices, the method comprising, for each said server:

• determining whether a new leader election is to take place;
• and
• in response to a determination that a new leader election is to take place:
• requesting votes for a new leader from the other server(s) in the plurality of servers;
• receiving any votes sent from other server(s);
• requesting votes for a new leader from the client device(s) that is/are eligible to vote as indicated in the consensus information;
• receiving any votes from the client device(s) that is/are eligible to vote;
• determining whether the respective server has won the election by receiving a clear majority of votes cast, including votes cast by the client device(s) that is/are eligible voters;
• and
• responsive to a determination that the respective server has won the election with a clear majority of votes cast, including votes cast by the client device(s) that is/are eligible voters, distributing consensus information to the other server(s) about the respective server now being the leader.

A non-transitory computer readable storage medium including instructions that, when executed by a given server in a distributed computing system including a plurality of servers organized in a cluster and a plurality of client devices operable to connect to the cluster and issue requests to the servers in which one of the servers is elected as a leader, cause that given server to execute operations comprising:

• maintaining, when the given server is elected leader, consensus information among the other servers, the consensus information including data indicating a shared operating state of the servers in the cluster, which server in the plurality of servers is elected as the leader, and which client device(s) is/are eligible to vote, wherein the shared operating state relates to requests issuable by the client devices;
• determining whether a new leader election is to take place;
• and
• in response to a determination that a new leader election is to take place:
• requesting votes for a new leader from the other server(s) in the plurality of servers;
• receiving any votes sent from other server(s);
• requesting votes for a new leader from the client device(s) that is/are eligible to vote as indicated in the consensus information;
• receiving any votes from the client device(s) that is/are eligible to vote;
• determining whether the given server has won the election by receiving a clear majority of votes cast, including votes cast by the client device(s) that is/are eligible voters;
• and
• responsive to a determination that the given server has won the election with a clear majority of votes cast, including votes cast by the client device(s) that is/are eligible voters, distributing consensus information to the other server(s) about the given server now being the leader.