Skip to main content

Expressions and serialization

Job methods have a few requirements because GUSTO stores the method call for later execution.

Supported methods​

Job methods must return Task. Synchronous and ValueTask methods can be called from a Task-returning job method when needed.

public Task GenerateAsync(Guid reportId)
{
    return reportGenerator.GenerateAsync(reportId);
}

await queue.EnqueueAsync<ReportJobs>(jobs =>
    jobs.GenerateAsync(reportId));

Avoid overloaded job method names because persisted jobs identify the method by name.

Use a concrete job class in the generic overload. Its constructor dependencies are resolved from dependency injection.

Argument evaluation​

Argument values are captured and serialized when the job is enqueued. Do not put side effects in argument expressions.

Every CancellationToken argument is replaced with default before serialization. During execution, the worker replaces it with the job execution token.

Serialized data​

Only your application should be able to write stored job arguments. Do not accept arbitrary serialized job records from untrusted sources.

ArgumentsJson for new jobs is a versioned GUSTO-owned JSON envelope containing version, arguments, and optional W3C traceparent and tracestate values. Argument serialization retains Newtonsoft.Json type metadata so heterogeneous argument arrays can be reconstructed. The worker also recognizes the earlier top-level JSON array format. Unsupported envelope versions are reported through the provider's normal execution-failure callback. Trace baggage is not stored.

Persisted jobs depend on the job type, method, and argument types. They can fail after changes such as:

  • renaming or moving the job type;
  • renaming the method;
  • changing the method overloads;
  • changing argument types incompatibly;
  • removing an assembly required by a serialized argument.

Plan deployment compatibility around the maximum time a job may remain queued. For long-lived queues, retain compatible entry methods or migrate pending records before deploying a breaking change.