Abstract: Disclosed embodiments provide a computer-implemented technique for creating a log chain from multiple software component logs. The log chain can include log information from multiple software components that interact with each other. In embodiments, a locality-sensitive hashing technique is used to create a signature of a log chain. Machine-learning systems are trained with log chains generated from test execution, and/or log chains from incidents with deployed software. In embodiments, logs are pre-processed by tokenizing and trimming. Logs from various components that interact with each other may be combined into a temporally sequential log chain. The signature of the log chain may be used to identify additional information about the error from a machine-learning process that was trained on previously generated log chains. In this way, the time required to identify a problem can be significantly reduced, resulting in increased reliability and availability of complex computer systems.

Abstract: Disclosed embodiments provide a computer-implemented technique for creating a log chain from multiple software component logs. The log chain can include log information from multiple software components that interact with each other. In embodiments, a locality-sensitive hashing technique is used to create a signature of a log chain. Machine-learning systems are trained with log chains generated from test execution, and/or log chains from incidents with deployed software. In embodiments, logs are pre-processed by tokenizing and trimming. Logs from various components that interact with each other may be combined into a temporally sequential log chain. The signature of the log chain may be used to identify additional information about the error from a machine-learning process that was trained on previously generated log chains. In this way, the time required to identify a problem can be significantly reduced, resulting in increased reliability and availability of complex computer systems.