跳到主要内容

plantpulse-warehouse (数据仓库 / 湖仓)

角色​

将 Cassandra 中的时序数据以 Apache Iceberg 格式归档到 MinIO 的湖仓模块。用于长期保留、分析查询性能优化和与其他 BI 工具集成。

项目值
模块名plantpulse-warehouse
安装路径/opt/kopens/plantpulse-platform/plantpulse-warehouse/
端口9600
格式Apache Iceberg V2
存储MinIO (S3)

架构​

目录结构​

plantpulse-warehouse/
├── app/                          # 웨어하우스 서비스
├── bin/start.sh stop.sh log-viewer.sh
├── config/
├── etc/                          # PID 관리
├── lib/                          # Spark JAR
├── logs/system.log
└── s3/
    ├── archive.sh                # 아카이빙
    └── optimize.sh               # 테이블 최적화

归档 (archive.sh)​

# 전일 데이터 아카이빙 (기본)
/opt/kopens/plantpulse-platform/plantpulse-warehouse/s3/archive.sh

# 특정 날짜
/opt/kopens/plantpulse-platform/plantpulse-warehouse/s3/archive.sh 2026-05-29
  • 基准性能：100万条记录约 60秒
  • 源：Cassandra tm_tag_point → MinIO plantpulse-iceberg/warehouse/tag_point/
  • 分区：dt=YYYY-MM-DD/hh=HH/

Iceberg 表优化 (optimize.sh)​

合并小文件、删除过期快照、清理元数据。

/opt/kopens/plantpulse-platform/plantpulse-warehouse/s3/optimize.sh

推荐频率：每周一次夜间执行。可通过 Kestra 工作流自动化。

运维命令​

cd /opt/kopens/plantpulse-platform/plantpulse-datalake-cli/bin

./restart-warehouse.sh
./pd status                # 9600 RUNNING 확인

# 표준 헬스체크 (익명, 순수 liveness — 공통 3개 필드 + 시스템/JVM 메트릭 포함)
curl -fsS http://127.0.0.1:9600/api/health
# → {"status":"UP","service":"plantpulse-warehouse-s3","ts":..., ...}

常见问题​

症状原因处置
归档 Job 失败Spark / Hive 未就绪pd restart analytics 后重试
小文件激增optimize.sh 未执行每周定期运行
MinIO 磁盘不足未设置保留周期Iceberg 过期快照删除 + 生命周期策略
Kyuubi 查询 OOM大表 / 缺少分区谓词显式指定 WHERE dt = '...'

SQL 查询示例​

USE iceberg;

-- 메타데이터
DESCRIBE TABLE EXTENDED tag_point;

-- 5월 집계
SELECT tag_id, count(*) AS rows, min(time), max(time)
FROM tag_point
WHERE dt BETWEEN '2026-05-01' AND '2026-05-31'
GROUP BY tag_id
ORDER BY rows DESC
LIMIT 100;

-- 스냅샷 이력
SELECT * FROM tag_point.snapshots;
SELECT * FROM tag_point.history;

相关文档​