You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@nifi.apache.org by "Mark Payne (Jira)" <ji...@apache.org> on 2026/09/05 21:13:00 UTC
[jira] [Created] (NIFI-16304) Connector verification results expire prematurely when another verification is submitted
Mark Payne created NIFI-16304:
---------------------------------
Summary: Connector verification results expire prematurely when another verification is submitted
Key: NIFI-16304
URL: https://issues.apache.org/jira/browse/NIFI-16304
Project: Apache NiFi
Issue Type: Bug
Environment: Reproduced on Apache NiFi 2.12.0-SNAPSHOT with a system test against a running standalone NiFi instance.
Reporter: Mark Payne
Submitting a connector configuration verification request can prematurely remove another completed verification request. Subsequent attempts to retrieve that result return HTTP 404, although verification succeeded.
{{ConnectorResource}} initializes its verification request manager with an expiration value of {{1L}}. The {{AsyncRequestManager}} constructor interprets this value as milliseconds. Each new submission removes completed requests older than that interval.
Ordinary submit -> poll -> delete usage works. The issue occurs when another verification is submitted before the client finishes retrieving or deleting an earlier result.
*Steps to reproduce:*
# Create two connectors.
# Submit configuration verification A and wait for successful completion, retaining the request.
# Wait 100 ms and confirm A remains retrievable.
# Submit configuration verification B for the other connector.
# Retrieve A again.
*Actual result:* HTTP 404, "Could not find a Request with identifier ...".
*Expected result:* A remains available for the intended retention period unless explicitly deleted.
This sequence was reproduced with a system test against a running NiFi instance. Both verifications succeeded; retrieving A after submitting B failed with HTTP 404.
*Proposed fix:* Pass {{TimeUnit.MINUTES.toMillis(1L)}} to the request manager. The connector purge request manager uses the same incorrect {{1L}} value and should also be corrected; the system-test reproduction currently covers verification.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)