Will try catch with branch

[smalljimmy]

Suppose I have following pipeline definition.

Will catch block be executed when test_processor fails?

pipeline:
  processors:
        - try:
          - branch:
              request_map: |
                root = this
              processors:
                - resource: test_processor
              result_map: |
                root.verified = this
        - catch:
          - bloblang: |
              root.meta.error = error()
              root.status = 400

[mihaitodor]

Yep! But don't trust me and try it out yourself.

input:
  generate:
    mapping: |
      root = {}
    interval: 0s
    count: 1

pipeline:
  processors:
    - try:
        - branch:
            request_map: |
              root = this
            processors:
              - bloblang: |
                  throw("kaboom!")
            result_map: |
              root.verified = this

        - log:
            message: "The rest of the processors in `try` won't get executed"

    - catch:
        - bloblang: |
            root.meta.error = error()
            root.status = 400

    - log:
        message: "The catch processor clears the error: ${! error() }"

output:
  stdout:
    codec: lines

[pokstad]

When I try this pipeline, it still logs the kaboom error even though the branch is wrapped with a try. Is that expected? I would expect that any try or catch logic would suppress normal error logging and let me choose how to surface the exception. Is there a way to suppress errors like kaboom?

[mihaitodor]

Yep, that's by design. You shouldn't perform work through exceptions by using the throw keyword to change behaviour and this log indicates a bug in your code that you need to fix. Ideally, pure processors which don't perform any IO should never fail. Instead, they should set some metadata or some field which indicates that the happy flow can't be performed and it's up to the subsequent processors to pick that up and handle it appropriately. For processors with side-effects, things can always go wrong, but we want to make sure that users are aware when that happens. In some cases, it might be desirable to suppress logging for certain components, but that would require an enhancements on a case-by-case basis.

[smalljimmy]

Thank you. In my case, the test_processor is a HTTP processor

  - label: test_processor
    http:
      url: 'http://api:8000/v2.0/private/verify_preapply'
      verb: PUT
      headers:
        Content-Type: application/json
        X-Product-Blacklist: '${! meta("X-Product-Blacklist")}'
      timeout: 20s

And I notice the catch blocker is NOT executed if the http call failed with 500 error. Is there a way to capture such error using try..catch?

[Jeffail]

@smalljimmy the http processor will retry on 500 errors, with a default number of 3 retry attempts per request: https://www.benthos.dev/docs/components/processors/http#retries, is it possible you're seeing logs showing the first request fail followed by a successful retry attempt? Otherwise I'd expect your catch block to execute.