-
[OPIK-7491] [BE] perf: trim E2E backend image compile (maven.test.skip, drop ineffective -T 1C) (#7615)
frostbyte_neo 发布于
2026-07-27 10:46:05 +00:00 - [OPIK-7491] [BE] perf: parallelize Maven compile with -T 1C in E2E backend build
Add -T 1C (one build thread per core) to the mvn dependency:go-offline and
mvn clean package invocations in apps/opik-backend/Dockerfile so the
multi-module compile uses the runner's cores instead of a single thread.
The produced image is functionally identical; only build wall-clock changes.
MAVEN_OPTS heap is left at -Xmx1G so CI can verify whether parallel modules
need more headroom before bumping.Implements OPIK-7491: parallelize Maven compile in E2E backend image build
Co-Authored-By: Claude Opus 4.8 (1M context) noreply@anthropic.com
- perf(backend-image): trim package stage via maven.test.skip + skip source/javadoc jars
CI proved -T 1C is a no-op here: opik-backend is a single-module jar (no
reactor), so Maven's MultiThreadedBuilder had nothing to parallelize and the
compile stage was marginally slower (116.7s -> 121.3s on a real E2E build).Replace it with the actual deterministic wins for this Dockerfile stage:
- -Dmaven.test.skip=true drops compiling 427 test sources (~28s) that the image
never uses; nothing consumes a test-jar. - -Dmaven.source.skip / -Dmaven.javadoc.skip drop the sources/javadoc jars the
runtime image doesn't ship.
The produced shaded main jar and openapi.yaml are unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) noreply@anthropic.com
Co-authored-by: Claude Opus 4.8 (1M context) noreply@anthropic.com
下载附件