Service Fabric Unhandled Exceptions和最佳实践

Question

Just curious if anyone has experience with unhandled exceptions in Service Fabric and what the best practices surrounding them are. Mainly curious about faulted state of services. Do the services get recycled if they are in a faulted state? Or should there be global exception handling for unhandled exceptions, if that concept even exists in SF. I haven't found much on this topic doing searches.

Answer 1

At my company we've built some re-usable code following the guidelines for using ITransaction , that allows us to run any arbitrary code and have it wrapped up with the appropriate exception handling and retry policies. Within that documentation, there is guidance on how the various types of exceptions should affect the RunAsync method, as well as how they should affect methods that are part of a remoting endpoint.

Generally, the handling of any exceptions outside of what is discussed in the documentation, is up to you to decide. For myself, I will simply let the exception bubble up and let the service fail if its an unexpected exception. Expected exceptions should always be handled with an appropriate retry policy.

Answer 2

Your question is quite broad, and the answer changes depending on the type of micro-service. Once good point to start, also getting an idea of the vastitude of this subject, is to look at the Service Fabric health monitoring article and related subtopics. As you can see there are plenty of configuration choices, and in my experience (I'm a Microsoft Azure Architect) there is no silver bullet, no "one size fits all" solution. The only best practice is to design the exception management that serves the best to your project, leveraging the Fabric health policies as much as possible.