[medium] fix: return a structured error when the remote event fetch fails

[elhoim]

BLUF

• Priority: medium
• push_event_by_uuid crashed with UnboundLocalError whenever the remote MISP server was unreachable, instead of returning its error dict.
• Cause: the except block dereferenced response, which is only bound by the very call that raised.
• Impact: DNS/connection/TLS/timeout failures killed the Celery push task with a traceback pointing at the handler, hiding the real cause; the UI got nothing useful.
• Fix: return a fixed 502 plus the exception text, and log with logger.exception.
• Covered by a new fixture-free regression test in app/tests/repositories/test_servers.py.

What was wrong

In api/app/repositories/servers.py, the handler guarding the remote event fetch read response.status_code and response.json():

    except Exception as ex:
        logger.warning(
            "Failed downloading the event {} from remote server {}".format(
                event_uuid, server.id
            ),
            ex,
        )
        return {
            "status": response.status_code,
            "message": "Failed downloading the event",
            "response": response.json(),
        }

response is bound only by remote_misp._prepare_request(...) inside the try — the call that can raise. Three defects in those lines:

1. Any transport-layer failure (DNS resolution, connection refused, bad TLS, timeout raised by requests inside _prepare_request) leaves response unbound, so the handler itself raises UnboundLocalError.
2. Even when response is bound — _check_json_response or MISPEvent.load raising on a 200 — response.json() can raise a second time on a non-JSON body.
3. logger.warning(msg, ex) passes ex as a printf-style argument to an already-.format()-ed message, which logging reports as a formatting error rather than logging the exception; and warning discards the traceback.

Concrete impact

Pushing an event to an unreachable, DNS-unresolvable or bad-TLS server never returns the structured error dict. The Celery task push_event_by_uuid dies with a misleading traceback anchored in the error handler, so the actual connection failure is lost from the logs, and the push endpoint returns a 500 rather than a readable message.

What this changes

One exception handler:

    except Exception as ex:
        logger.exception(
            "Failed downloading the event {} from remote server {}".format(
                event_uuid, server.id
            )
        )
        return {
            "status": 502,
            "message": "Failed downloading the event: %s" % ex,
        }

Notes on the shape chosen:

• It mirrors the sibling handler further down the same function, which already returns a fixed status and a message with no response dereference.
• The response key is dropped rather than reconstructed: nothing consumes it. The router returns the dict verbatim, the Celery task discards it, and the frontend reads only message (frontend/src/components/events/EventActions.vue:58).
• "...: %s" % ex is the idiom already used in this module for remote-connection failures.
• logger.exception keeps the traceback and removes the stray positional argument.

This also removes the second latent crash (response.json() on a non-JSON body) without a separate change.

How it was verified

• New regression test TestPushEventByUuid::test_unreachable_server_returns_an_error in api/app/tests/repositories/test_servers.py, following the existing fixture-free class pattern in that module. It patches get_remote_misp_connection so _prepare_request raises ConnectionError, asserts the fetch was actually reached, then asserts status == 502 and that the exception text reaches message. It errors with UnboundLocalError before this change and passes after it, and needs no database, OpenSearch or network.
• python3 -m compileall-equivalent syntax sweep over api/ is clean.
• No schema, route or migration change, so docs/features/api/openapi.json is unaffected.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 84.40%. Comparing base (adf4f0c) to head (1fdb90a).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #391      +/-   ##
==========================================
+ Coverage   84.33%   84.40%   +0.06%     
==========================================
  Files         208      208              
  Lines       19391    19402      +11     
==========================================
+ Hits        16354    16376      +22     
+ Misses       3037     3026      -11     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[righel]

Makes sense, thanks!