[jira] [Updated] (FLINK-40662) 'Run kubernetes Materialized Table test' fails on release-2.3 because the SQL Gateway returns a null session handle

["Martijn Visser (Jira)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/FLINK-40662?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Martijn Visser updated FLINK-40662:
-----------------------------------
    Labels: test-stability  (was: )

> 'Run kubernetes Materialized Table test' fails on release-2.3 because the SQL Gateway returns a null session handle
> -------------------------------------------------------------------------------------------------------------------
>
>                 Key: FLINK-40662
>                 URL: https://issues.apache.org/jira/browse/FLINK-40662
>             Project: Flink
>          Issue Type: Bug
>          Components: Table SQL / Gateway
>    Affects Versions: 2.3.0
>            Reporter: Martijn Visser
>            Priority: Major
>              Labels: test-stability
>
> The e2e script waits for the gateway, gets {{SqlGateway REST endpoint is up.}}, and then the session it opens comes back empty, so every following request is sent to the literal path {{null}}:
> {code}
> Sep 14 10:00:03 SqlGateway REST endpoint is up.
> Sep 14 10:00:03 [INFO] Create a new session
> Sep 14 10:00:05 [INFO] Session Handle null
> Sep 14 10:00:05 [INFO] Create catalog & database
> Sep 14 10:00:05 Configure session null create catalog if not exists test_catalog with (...) failed: {"errors":["org.apache.flink.runtime.rest.handler.RestHandlerException: Bad request, could not parse parameters: Cannot resolve path parameter (session_handle) from value \"null\"."]}
> {code}
> {code}
> Sep 14 10:00:16 [FAIL] Test script contains errors.
> Sep 14 10:00:16 [FAIL] 'Run kubernetes Materialized Table test' failed after 1 minutes and 21 seconds! Test exited with exit code 1 and the logs contained errors, exceptions or non-empty .out files
> {code}
> Two earlier gateway starts in the same run came up in one to two seconds and worked, so the endpoint being reachable is not enough on its own here.
> First seen on the 2026-09-12 nightly, the 2026-09-11 nightly was still clean. Five of the twelve release-2.3 Azure runs in the window hit it:
> https://dev.azure.com/apache-flink/apache-flink/_build/results?buildId=79037
> https://dev.azure.com/apache-flink/apache-flink/_build/results?buildId=79029
> https://dev.azure.com/apache-flink/apache-flink/_build/results?buildId=79020
> https://dev.azure.com/apache-flink/apache-flink/_build/results?buildId=79010
> https://dev.azure.com/apache-flink/apache-flink/_build/results?buildId=79002
> FLINK-40123 and FLINK-39132 are the same script failing for other reasons.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)