[Assembly] Use Log4j 2 configuration in wayang-submit (#536)

[mallesh-bot]

Closes #536

Problem

Wayang uses Log4j 2 (log4j-bom 2.20.0), but bin/wayang-submit passed the Log4j 1 property when FLAG_LOG=true:

-Dlog4j.configuration=file://${WAYANG_CONF}/log4j.properties

Log4j 2 ignores that property, and conf/ did not ship a logging configuration, so FLAG_LOG had no effect. Running WordCount from the distribution with FLAG_LOG=true on current main prints:

ERROR StatusLogger Reconfiguration failed: No configuration found for '289d1c02' at 'null' in 'null'

and only errors are logged. A file created manually in $WAYANG_HOME/conf is also lost on the next clean install, because that directory is regenerated from the source conf/ directory by the assembly.

Change

• bin/wayang-submit: pass -Dlog4j2.configurationFile=file://${WAYANG_CONF}/log4j2.properties when FLAG_LOG=true.
• conf/log4j2.properties (new): default configuration, INFO and above to stderr, so logs do not mix with application output on stdout. The assembly already copies conf/** into the distribution.
• wayang-assembly/README.md: document FLAG_LOG next to FLAG_WAYANG.

Behaviour without FLAG_LOG is unchanged.

Testing

All workflows of this repository passed on this commit in my fork (Backend incl. license check, Maven and Python tests; Docker incl. WordCount smoke test; CodeQL).

In addition I ran an end-to-end check on the built distribution image (Linux, JDK 17):

• FLAG_LOG=true: Log4j 2 loads /opt/wayang/conf/log4j2.properties without StatusLogger warnings or errors; Wayang's INFO/WARN logs (e.g. INFO Job: Preparing plan...) appear on stderr in the configured format; stdout only contains the WordCount result.
• FLAG_LOG unset: same output as before.
• Same run with the script from main: reproduces the No configuration found error above.

Notes (not changed here)

While checking which logging jars end up in libs/, I noticed the distribution also ships log4j-1.2.17.jar, reload4j-1.2.22.jar, slf4j-reload4j-1.7.36.jar and slf4j-simple-2.0.16.jar next to the Log4j 2 jars, which causes the SLF4J: Class path contains multiple SLF4J providers warning. wayang-api-sql also has two log4j.properties files that reference an undefined appender A1 (visible as log4j:ERROR Could not find value for key log4j.appender.A1 in CI). I kept this PR focused on #536 and can follow up on these separately if that is useful.

The contribution is my original work and I license the work to the project under the project's open source license.

AI assistance: parts of this change were prepared with an AI tool (Claude Code), disclosed in the commit with a Generated-by: trailer, following the project's AI-assisted contributions guidelines.

🤖 Generated with Claude Code