adaptive-accrual-failure-detector
v1.0.0
Published
Failure detection for processes, connections and distributed systems
Downloads
25
Maintainers
Readme
adaptive-accrual-failure-detector
Failure detection for processes, connections and distributed systems. This is an implementation for JavaScript and TypeScript of a failure detector that uses an adaptive accrual algorithm. The theory of this detector is taken from the paper A New Adaptive Accrual Failure Detector for Dependable Distributed System authored by Benjamin Satzger, Andreas Pietzowski, Wolfang Trumler and Theo Ungerer.
This detector is useful for detecting things such as network failures between two nodes.
Usage
The failure detector is based on incoming heartbeats and has a few options that can be used when creating the detector:
sampleSize
- number of samples kept and used when calculating probability of a failure. A higher number of samples means the probability calculation is more stable but uses more memory. Default is1000
.scalingFactor
- factor used to scale failure probabilities, used to reduce overestimation of failure. Default is0.9
.failureThreshold
- the probability needed to to detect something as a failure. If the probability of failure is above this the thing being monitored is considered failed. Default is0.5
.
// Using ES Module environment
import { FailureDetector } from 'adaptive-accrual-failure-detector';
const detector = new FailureDetector({
failureThreshold: 0.6
});
When you receive a heartbeat you should call registerHeartbeat
on the detector:
detector.registerHeartbeat();
To calculate if a failure has occurred you can call checkFailure()
or
calculateFailureProbability()
:
// Check failure, will return true if failed
const isFailed = detector.checkFailure();
// Calculate the probability of failure between 0 and 1
const failureProbability = detector.calculateFailureProbability();