Tracelogstotraining
What is Tracelogstotraining?
Tracelogstotraining is a project that transforms trace logs into training sets for machine learning models. Our training sample sets will be used for the training of the microservice fault diagnosis model.
How to use Tracelogstotraining？
- Import trace logs (.json)
- Setting parameters
- starttime: The earliest timestamp of the target trace logs
- endtime: The latest timestamp of target trace logs
- fstarttime: The start timestamp of faults injection
- fendtime: The end timestamp of faults injection
- Modify the value of fnum
- Select function (by modifying the ftype parameters)
- 0: Generate the normal request training sets
- 1: Generate the faults request training sets
- 2: Filter original trace logs
- Get the training sets
No trace logs？
We give a sample in the example/trace logs directory, you just need to move it to the root of the project.