talayash/spring-fleet

spring-fleet-logging-setup

Use when fleet logs are console-only or scattered and cross-service correlation needs a stable location - installs a logback convention so every service writes / .log with OTel traceid / spanid and legacy sessionId / requestId in the pattern.

소스 보기

사용 방법

  1. 설치 명령을 복사하고 자신의 에이전트 환경에 설치하기 전에 소스 패키지를 확인하세요.
  2. 필요한 도구, 인증 정보, 지원 환경은 원본 문서에서 확인하세요.
  3. 입력과 원하는 산출물을 설명하고 출처와 결과를 대조하세요.
npx skills add https://github.com/talayash/spring-fleet --skill spring-fleet-logging-setup

시작 전 확인

  • 설치에는 모델 구독, API 한도, 파일 접근 권한이 포함되지 않습니다.
  • 별점 수와 설치 수는 활동 지표이며 독립적인 품질·보안 검증이 아닙니다.
원본 Skill 문서

원본 저장소의 제목, 예시, 코드, 표, 링크, 이미지를 유지해 표시합니다.

Fleet Logging Setup (logback convention)

Cross-service log correlation needs two things: every service writing to one known directory, and trace keys (trace_id, span_id, sessionId, …) present in each line. This skill installs a logback convention that guarantees both. It is idempotent and opt-in per service.

What trace keys to use (2026)

  • Primary: OpenTelemetry / Micrometer Observation. Add the

spring-boot-starter-actuator + micrometer-tracing-bridge-otel (or, on Spring Boot 4, the dedicated spring-boot-starter-opentelemetry). The W3C traceparent header is auto-propagated across HTTP/gRPC, and the Observation API automatically populates MDC keys trace_id (32-char hex) and span_id (16-char hex). These are the modern, framework-default keys.

  • Fallback / legacy: sessionId, requestId. Keep them in the pattern for

fleets that pre-date the Observation API or that propagate an application- level session id (auth, X-Request-Id) the user wants to correlate on.

When to use

  • /debug reports missing log files, or logs only go to the IDE console.
  • Logs exist but lack a trace key, so correlation falls back to timestamps.

Procedure

  1. Read the config for logDir and traceKeys.
  1. Start from the template: ${CLAUDE_PLUGIN_ROOT}/logback/logback-spring.xml.

For each service you want covered, place it at <service-repo>/src/main/resources/logback-spring.xml (or merge its appender

  • pattern into the existing one), then customize three things:
  • SERVICE_NAME → the service's name from the config (used as the log file

base name, so /debug can find it).

  • The %X{...} keys in FLEET_PATTERN → one per entry in traceKeys.
  • SPRING_FLEET_LOG_DIR → the config logDir, or set that env var when

launching the service.

  1. Ensure the trace keys are in MDC. A pattern can only print %X{trace_id}

if something put trace_id into the MDC.

  • trace_id / span_id: come for free once Micrometer Tracing or the OTel

starter is on the classpath; Spring Boot wires the Observation API which populates MDC automatically. Verify by hitting any endpoint and tailing the log; the keys should already be there.

  • sessionId / requestId (legacy): usually require a filter/interceptor

that reads the incoming header (or generates the id), calls MDC.put(...), and clears it in a finally. If propagation across services is missing, the same id must be forwarded as a header on outbound proxy-lib calls. (Modern fleets should rely on W3C traceparent instead and treat sessionId as a business id, not a correlation id.)

  1. Verify. Run the service, exercise one request, then:
   python "${CLAUDE_PLUGIN_ROOT}/scripts/tail_logs.py" \
     --config ./spring-fleet.config.json --service <name> --lines 20

Confirm lines land in <logDir>/<service>.log and carry the trace key.

Notes

  • Keep the same pattern across services so timestamps sort cleanly when

merged. The template's yyyy-MM-dd HH:mm:ss.SSS prefix is what the correlator parses.

  • The template also keeps a CONSOLE appender, so IDE output is unchanged.
  • For services you will not modify, prefer the (future) /run tee approach

instead of editing their logback config.

Outside Claude Code

  • ${CLAUDE_PLUGIN_ROOT} is the spring-fleet plugin root. Claude Code fills it in; in Codex it is the folder two directories above this SKILL.md.