Kafka cluster abnormal node detection method based on normal distribution model

The invention belongs to the technical field of big data cloud computing, and discloses a Kafka cluster abnormal node detection method based on a normal distribution model, and the method comprises the following steps: S1, feature selection; selecting a thread index in a Kafka node process as a feature, and obtaining a corresponding feature variable D; s2, constructing normal feature space distribution; constructing a data set for normal feature space distribution according to the obtained feature variable, constructing a distribution function, and determining a normal operation interval of the feature variable D; s3, performing anomaly detection and warning; and judging whether abnormity occurs or not by judging whether the characteristic variable D of the current day falls into the normal operation interval, and giving an alarm if the abnormity occurs. Compared with the prior art, the method for automatically detecting the abnormity is provided, and it is avoided that the alarm indexes are manually determin