Mark Payne created NIFI-16305:
---------------------------------
Summary: Fix startup error propagation, flow snapshot metadata,
and secret lookup in Connector mock framework
Key: NIFI-16305
URL: https://issues.apache.org/jira/browse/NIFI-16305
Project: Apache NiFi
Issue Type: Bug
Components: Core Framework
Reporter: Mark Payne
The nifi-connector-mock-bundle test framework currently hides failed startup
attempts, produces inaccurate parameter snapshots, and cannot resolve secrets
using the fully qualified names returned by discovery.
h3. Startup failures are discarded
*Reproduction*
Build a runner for an invalid Connector, such as the existing
MissingBundleConnector, and call startConnector(). The call returns normally
even though the Connector cannot start.
StandardConnectorMockServer.startConnector() discards the Future returned by
ConnectorNode.start(). Validation and configuration-resolution exceptions are
captured in that Future, leaving the test without the failure cause.
*Expected behavior*
The runner reports a failed startup with its underlying cause instead of
appearing successful or requiring a later ingestion timeout to reveal the
failure.
*Acceptance tests*
* An invalid Connector reports its startup failure through the runner.
* A Connector whose configuration cannot resolve reports its startup failure
through the runner.
* A valid Connector still starts successfully.
h3. Flow snapshots lose asset references and misreport parameter metadata
*Reproduction*
Snapshot a parameter context containing the following:
* A sensitive parameter with provided=false. The snapshot reports provided=true.
* A nonsensitive parameter with provided=true. The snapshot reports
provided=false.
* A parameter referencing an asset. The snapshot loses referencedAssets and
instead contains the resolved local file path.
StandardConnectorMockServer.createVersionedParameterContext() sets provided
from the sensitivity flag and does not map referenced assets. Both active and
working snapshots use this helper.
*Expected behavior*
Snapshots preserve the actual provided flag and asset references consistently
with the production flow mapper, while continuing to omit sensitive values.
*Acceptance tests*
* Both snapshot APIs preserve asset identifiers and names.
* Both snapshot APIs preserve all combinations of sensitive and provided
metadata.
* Sensitive values remain omitted.
h3. Discovered secret names cannot be resolved
*Reproduction*
Register password using addSecret(). Discovery returns a secret with the fully
qualified name Default.password. Resolve a SecretReference using that provider
and fully qualified name. Lookup returns no secret.
ConnectorTestRunnerSecretProvider.getAllSecrets() adds the Default. prefix, but
getSecrets() looks up the supplied fully qualified name directly in a map keyed
by the unqualified name. The createSecretReference() convenience path masks
this inconsistency by using the bare name.
*Expected behavior*
A secret returned by discovery can be resolved using its advertised provider
and fully qualified name. References produced by createSecretReference() also
resolve consistently.
*Acceptance tests*
* Cover discovery-to-reference-to-lookup round trips.
* Cover references produced by the runner helper.
* Cover secret names containing dots.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
