Failure Detection

How can a process crash be reliably detected?

General mode:

• Each process has its own failure detection module.

• A process P probes another process, Q, for a reaction.

• If Q reacts -> Q is considered available (by P).

• If Q does not react in a t timeframe -> Q is suspected to have failed.

In a synchronous system:

• A suspicion of a crash is in reality, a sure thing.

In practice

If P does not receive a heartbeat from Q in a t timeframe: P will suspect Q.

If Q sends a message later on (which is received by P):

• P stops being suspicious of Q.

• P increases the t value.