How to implement a custom retry handler when all useful information is lost?

[thomaseyde]

In #2697 it is explained that activity exceptions will not be serialized.

With this decision, how are we supposed to implement custom retry handlers when there is no useful information to act upon?

I have a scenario where I call an Sql Database where timeouts are expected due to cold starts. I want to retry in these circumstances, but not anywhere else. I also do not want to litter my orchestration with try...catch, which needs their own workarounds also explained in #2476.

As an example, I have a test activity throwing a TimeoutException and a custom retry handler implemented with TaskOptions.FromRetryHandler:

var taskOptions = TaskOptions.FromRetryHandler(
    async retryContext =>
    {
        // Actual exception in debugger:
        //  TaskFailedException: The operation has timed out.
        //  InnerException: null
        //  FailureDetails: same as retryContext.LastFailure
        
        // retryContext.LastFailure: 
        //  ErrorMessage: "One or more errors occurred. (The operation has timed out.)"
        //  ErrorType: "System.AggregateException"
        //  InnerFailure: null
        
        return false;  // not a useful result
    });

RetryContext contains no useful information. I cannot reliably depend on the error message from an AggregateException.

In this case it it no use to add a try...catch in my orchestration, as the exception thrown is a TaskFailedException with just as little information.