Microservices using Petri Nets Sulochan Naik sulochan.naik@iiitb.org 0 Meenakshi D'Souza meenakshi@iiitb.ac.in 0 Workshop Proceedings 0 International Institute of Information Technology , Bangalore , India 2023

Despite higher adoption of Microservices worldwide, there is a lack of work on possible fault detection during early phases such as at requirements and design phase. We propose a novel way to detect faults using Petri nets' traversal analysis and model checking CTL specifications. Requirements of a system involving microservices are mapped to Netflix's Conductor specification which is later converted into Petri nets for further analysis and detection of faults. We tried our methodology on the TrainTicket benchmark system and were able to detect seven faults.

 Figure 1 gives complete description of our proposed methodology Based on our analysis on -

1. Introduction > 0). • P3 - Decentralized property can be verified by: AF(end is reached) for all microservices. • P4 - Feasible execution path can be verified by: AF(end is reached). • P5 - Verification of deadlock state can be done by: EF(end is not reached).

• P6 - Bounded response time can be verified using : AG(some state) < desired token. Using our methodology, we were able to detect seven faults on a benchmark microservices application called TrainTicket [ 2 ] which is depicted in Table 1. nEvelop-O

Microservice Errors [ 3 ]

F1 (Messages are displayed in wrong order) F2 (Some information displayed in a report is wrong)

F4 (The response time for some requests is very long) F5 (A service sometimes returns timeout exceptions for user requests)

F7 (The payment service of the system fails)

F16 (The file-uploading process fails ) F20 (Nothing is returned upon workflow data request ) Mapping

P1 P3 P6 P6 P2 P2 P5 3. Conclusion and Future Work We have verified our methodology on the train ticket benchmark system with satisfactory results. We plan to extend this work as a part of formal verification of microservices.

 [1] Netflix , Inc., Conductor , 2016 . URL: https://netflix.github.io/conductor/. [2] Microservice System Benchmark , Trainticket, 2018 . URL: https://github.com/FudanSELab/ train-ticket/. [3] Microservice System Benchmark , Train ticket fault replicate , 2018 . URL: https://github.com/ FudanSELab/train-ticket -fault-replicate.