Detection of container incidents using machine learning techniques

Abstract

Methods, apparatus, and processor-readable storage media for detecting container incidents using machine learning techniques are provided herein. An example method includes generating first and second representations of a telemetry dataset associated with a software container in an edge computing environment, the telemetry dataset including values for a set of parameters for each of a plurality of timestamps; providing the first representation to a predictive model to obtain a predicted remaining lifetime of the software container; providing the second representation of the telemetry dataset to a first machine learning model to obtain a predicted behavior of the software container; determining, using a second machine learning model, whether the predicted behavior of the software container corresponds to a pattern of behavior that is associated with previous container incidents; and triggering an automated action for the software container in response to determining that the predicted behavior corresponds to the pattern of behavior.