The GitHub Actions job "Flink Kubernetes Operator CI" on 
flink-kubernetes-operator.git/main has failed.
Run started by GitHub user rmetzger (triggered by rmetzger).

Head commit for run:
5ca04ac7ebfb3d1d3358aa9b5dd613755cd12579 / Purushottam Sinha 
<[email protected]>
[FLINK-40246] Migrate logging stack to SLF4J 2.x and bump Logback to 1.5.34 
(#1170)

* [FLINK-40246] Migrate logging stack to SLF4J 2.x and bump Logback to 1.5.34

Logback 1.2.13 is the terminal release of an EOL line, so the CVEs filed
against it (CVE-2024-12798, CVE-2024-12801, CVE-2025-11226, CVE-2026-1225,
CVE-2026-9828, CVE-2026-10532) have no fix on that branch. Every patched
Logback release targets SLF4J 2.x, so clearing them requires moving the
whole logging stack off slf4j-api 1.7.36.

The 1.7.36 pin is independently incorrect. JOSDK 5.5.0 and fabric8 7.8.0
both declare slf4j-api 2.0.18, and JOSDK's
ReconciliationDispatcher.handleErrorStatusHandler calls
Logger.atLevel(Level), which does not exist in 1.7.36. A FlinkStateSnapshot
whose reconcile fails and whose error-status patch then also fails throws
NoSuchMethodError from inside the error handler, masking the original
failure.

- slf4j-api 1.7.36 -> 2.0.18, logback 1.2.13 -> 1.5.34
- log4j-slf4j-impl -> log4j-slf4j2-impl across all modules, including the
  dependency-plugin item that stages the binding into the image
- drop log4j-slf4j-impl from the operator NOTICE (provided since
  FLINK-39501, so never bundled) and refresh the slf4j-api versions
- drop the "Logging Library Version Overrides" hint from the logging docs:
  swapping the shipped logging jars is not a supported user operation

* [FLINK-40246] Pin slf4j-api and prune Flink's SLF4J 1.7 binding

Flink brings slf4j-api 1.7.36 in transitively, and without a
dependencyManagement pin four modules resolve it rather than 2.0.18:
flink-kubernetes-standalone, flink-kubernetes-operator-api,
flink-autoscaler and flink-autoscaler-plugin-jdbc.

The first two then pair 1.7.36 with the 2.x-only log4j-slf4j2-impl. SLF4J
does not fail on a binding/API mismatch: it installs a NOP logger and
discards every log line, so their test output was silently dead. The other
two silently stayed on the complete 1.7 stack, because
flink-test-utils-junit supplies the matching 1.7 binding - which is why the
migration appeared to have succeeded there.

Excluding Flink's log4j-slf4j-impl belongs to the same change rather than
being cleanup: once the pin moves those modules to 2.0.18, the 1.7 binding
inherited from the Flink test utils would be the only one left, flipping
them from working-on-1.7 to NOP.

The shipped image was not affected. flink-kubernetes-operator declares
slf4j-api directly, so the shaded jar always bundled 2.0.18, and the other
modules' classes are caller-side compatible at runtime.

- pin slf4j-api in dependencyManagement
- exclude Flink's log4j-slf4j-impl where flink-test-utils(-junit) puts it on
  a test classpath, and declare log4j-slf4j2-impl there instead
- test_logback_logging.sh: fail on the NOP fallback ("No SLF4J providers
  were found", "Failed to load class StaticLoggerBinder") and match SLF4J
  2.x's "providers" wording alongside 1.7.x's "bindings"

All 14 reactor modules now resolve slf4j-api 2.0.18, and none resolves the
1.7 binding.

* [FLINK-40246] Empty commit to re-trigger CI

The e2e_ci matrix failed on a Maven Central 403 while fetching the ASF
parent POM inside the Docker build, unrelated to this change. The re-run
carried the old result forward instead of rebuilding, so this triggers a
fresh run.

Report URL: 
https://github.com/apache/flink-kubernetes-operator/actions/runs/31718254824

With regards,
GitHub Actions via GitBox

Reply via email to