LogGrouper: Making Sense of Failure Logs

Processing and reviewing nightly test execution failure logs for large industrial systems is a tedious activity. Furthermore, multiple failures might share one root cause during test execution sessions, and the review might, therefore require redundant efforts. The LogGrouper approach for automated grouping of failure logs aids root cause analysis and enables the processing of each log group as a batch.

The LogGrouper approach uses the following steps to enable root cause analysis and make sense of failure logs:

• The approach pre-process failure log messages and lemmatize them
• The approach uses clustering to group similar failure logs together
• The approach then uses Rapid Application Keyword Extraction to summarize each failure group
• The approach visualize failure groups as word clouds

spaCy, NLTK, Numpy, Pandas

DevOps Engineers, Software Testers

Licensing and permission required.