METHOD AND APPARATUS FOR REMOVING A COMPUTER FROM A COMPUTER CLUSTER OBSERVING FAILURE

A method, and a corresponding apparatus, implemented on a suitably programmed processor, selects an ideal computer to remove from a computer cluster observing a failure condition. The computer cluster includes multiple computers. The method includes the steps of recording, at each of the computers, health data for each of the computers, computing, at a health timeout, a first health score for each of the computers based on the health data, computing, at a fatal timeout, and based on the first health score, a second health score for each of the computers, and at each of the computers, selecting a computer having the highest health score for removal from the cluster.