<clickhouse>
  <storage_configuration>
    <disks>
      <!-- S3 오브젝트 스토리지 disk. {replica} 매크로로 replica마다 경로 분리(shared-nothing) -->
      <s3_disk>
        <type>object_storage</type>
        <object_storage_type>s3</object_storage_type>
        <metadata_type>local</metadata_type>   <!-- part 매핑 메타데이터는 로컬(EBS)에 상주 → §5.1 -->
        <endpoint>https://rum-clickhouse-cold.s3.ap-northeast-2.amazonaws.com/s3_disk/{replica}/</endpoint>
        <use_environment_credentials>1</use_environment_credentials>  <!-- IRSA(§3) -->
        <region>ap-northeast-2</region>
        <metadata_path>/var/lib/clickhouse/disks/s3_disk/</metadata_path>
      </s3_disk>
      <!-- S3 disk 위에 로컬(EBS) LRU 캐시를 얹는다. cold 쿼리 지연 방어의 핵심 -->
      <s3_cache>
        <type>cache</type>
        <disk>s3_disk</disk>
        <path>/var/lib/clickhouse/disks/s3_cache/</path>
        <max_size>150Gi</max_size>                          <!-- LRU 상한. EBS 소비항(§5.1). 실값은 스테이징 튜닝 -->
        <cache_on_write_operations>1</cache_on_write_operations>  <!-- TTL MOVE 시점에 프리페치 → 첫 조회 완화 -->
      </s3_cache>
    </disks>
    ...
  </storage_configuration>
</clickhouse>

    <policies>
      <rum_hot_cold>
        <volumes>
          <hot>
            <disk>default</disk>          <!-- = /var/lib/clickhouse = gp3/io2 PVC(내장 disk, 선언 불필요) -->
          </hot>
          <cold>
            <disk>s3_cache</disk>         <!-- cache로 감싼 S3 -->
          </cold>
        </volumes>
        <move_factor>0.1</move_factor>    <!-- 안전판(여유<10%=~90% 찼을 때만 개입). 주 이동은 시간 TTL. §1.3 -->
        <!-- prefer_not_to_merge 미설정(기본 false 유지): S3 위 작은 part 폭증 방지 -->
      </rum_hot_cold>
    </policies>

apiVersion: "clickhouse.altinity.com/v1"
kind: "ClickHouseInstallation"
metadata:
  name: rum-observability
  namespace: clickhouse
spec:
  defaults:
    storageManagement:
      provisioner: Operator          # EBS는 무중단 확장 가능 → Operator 선택 이점(로컬 NVMe와 다른 점)
      reclaimPolicy: Retain
    templates:
      podTemplate: ch-ebs
      dataVolumeClaimTemplate: data-ebs      # → /var/lib/clickhouse (hot=default 디스크)
      logVolumeClaimTemplate:  log-ebs
      serviceTemplate: ch-svc
  configuration:
    zookeeper:
      keeper:
        name: rum-keeper            # CHK 참조(Keeper 상세는 05)
    clusters:
      - name: main
        pdbManaged: "yes"
        pdbMaxUnavailable: 1
        layout:
          shardsCount: 1
          replicasCount: 2           # RF2. cold(S3)도 replica별 사본(shared-nothing) → {replica} 경로 분리
    files:
      config.d/storage_configuration.xml: |
        <clickhouse>
          <storage_configuration>
            <disks>
              <s3_disk>
                <type>object_storage</type>
                <object_storage_type>s3</object_storage_type>
                <metadata_type>local</metadata_type>
                <endpoint>https://rum-clickhouse-cold.s3.ap-northeast-2.amazonaws.com/s3_disk/{replica}/</endpoint>
                <use_environment_credentials>1</use_environment_credentials>
                <region>ap-northeast-2</region>
                <metadata_path>/var/lib/clickhouse/disks/s3_disk/</metadata_path>
              </s3_disk>
              <s3_cache>
                <type>cache</type>
                <disk>s3_disk</disk>
                <path>/var/lib/clickhouse/disks/s3_cache/</path>
                <max_size>150Gi</max_size>
                <cache_on_write_operations>1</cache_on_write_operations>
              </s3_cache>
            </disks>
            <policies>
              <rum_hot_cold>
                <volumes>
                  <hot><disk>default</disk></hot>       <!-- EBS gp3/io2 PVC -->
                  <cold><disk>s3_cache</disk></cold>
                </volumes>
                <move_factor>0.1</move_factor>
              </rum_hot_cold>
            </policies>
          </storage_configuration>
        </clickhouse>
    settings:
      # storage_policy는 보통 테이블별 SETTINGS로 지정(§4). 서버 기본으로 강제하려면:
      # merge_tree/storage_policy: rum_hot_cold
      max_concurrent_queries: 200
  templates:
    podTemplates:
      - name: ch-ebs
        spec:
          serviceAccountName: clickhouse-s3     # ← IRSA SA(§3.2)
          nodeSelector: { workload: clickhouse }
          tolerations:
            - { key: dedicated, operator: Equal, value: clickhouse, effect: NoSchedule }
          containers:
            - name: clickhouse
              image: clickhouse/clickhouse-server:24.8   # ClickStack 병용 24.8 LTS+
              resources:
                requests: { cpu: "4", memory: "32Gi" }
                limits:   { cpu: "4", memory: "32Gi" }
    volumeClaimTemplates:
      - name: data-ebs
        spec:
          accessModes: [ReadWriteOnce]
          storageClassName: gp3                 # 또는 io2 — 선택 기준은 02
          # prod 노드당 order ~1TB. hot 데이터 + part metadata + s3_cache(150Gi) + 머지 여유를 모두 포함(§5.1).
          # 정확한 사이징은 06이 기준 문서. 스테이징은 소규모(예 100Gi).
          resources: { requests: { storage: 1000Gi } }
      - name: log-ebs
        spec:
          accessModes: [ReadWriteOnce]
          storageClassName: gp3
          resources: { requests: { storage: 50Gi } }
    serviceTemplates:
      - name: ch-svc
        spec:
          type: ClusterIP
          ports:
            - { name: http, port: 8123 }
            - { name: tcp,  port: 9000 }

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Sid": "clickhouse-s3-cold-object",
      "Effect": "Allow",
      "Action": ["s3:GetObject", "s3:PutObject", "s3:DeleteObject"],
      "Resource": "arn:aws:s3:::rum-clickhouse-cold/*"
    },
    {
      "Sid": "clickhouse-s3-cold-list",
      "Effect": "Allow",
      "Action": ["s3:ListBucket", "s3:GetBucketLocation"],
      "Resource": "arn:aws:s3:::rum-clickhouse-cold"
    }
  ]
}

apiVersion: v1
kind: ServiceAccount
metadata:
  name: clickhouse-s3
  namespace: clickhouse
  annotations:
    eks.amazonaws.com/role-arn: arn:aws:iam::123456789012:role/clickhouse-cold-s3

-- 0) 정책 연결. storage_policy 변경은 "볼륨 추가 방향"(hot 유지 + cold 추가)이라 허용.
--    sessions는 의도적으로 rum_hot_cold를 붙이지 않는다(§4.4) → 기본 default 정책(EBS only) 유지.
ALTER TABLE default.otel_logs              MODIFY SETTING storage_policy = 'rum_hot_cold';
ALTER TABLE default.otel_traces            MODIFY SETTING storage_policy = 'rum_hot_cold';
ALTER TABLE default.otel_metrics_gauge     MODIFY SETTING storage_policy = 'rum_hot_cold';
ALTER TABLE default.otel_metrics_sum       MODIFY SETTING storage_policy = 'rum_hot_cold';
ALTER TABLE default.otel_metrics_histogram MODIFY SETTING storage_policy = 'rum_hot_cold';
ALTER TABLE default.otel_metrics_summary   MODIFY SETTING storage_policy = 'rum_hot_cold';

-- 1) logs: hot 14일 → S3, 90일 DELETE  (6개월 지평=180 / 1년=365)
ALTER TABLE default.otel_logs MODIFY TTL
    toDateTime(Timestamp) + INTERVAL 14 DAY TO VOLUME 'cold',
    toDateTime(Timestamp) + INTERVAL 90 DAY DELETE;

-- 2) traces: hot 14일 → S3, 90일 DELETE  (6개월=180 / 1년=365)
ALTER TABLE default.otel_traces MODIFY TTL
    toDateTime(Timestamp) + INTERVAL 14 DAY TO VOLUME 'cold',
    toDateTime(Timestamp) + INTERVAL 90 DAY DELETE;

-- 3) metrics: hot 30일 → S3, 180일 DELETE  (1년 지평=365). gauge/sum/histogram/summary 동일 패턴 반복
ALTER TABLE default.otel_metrics_gauge MODIFY TTL
    toDateTime(TimeUnix) + INTERVAL 30 DAY TO VOLUME 'cold',
    toDateTime(TimeUnix) + INTERVAL 180 DAY DELETE;
-- ALTER TABLE default.otel_metrics_sum       MODIFY TTL ... (동일)
-- ALTER TABLE default.otel_metrics_histogram MODIFY TTL ... (동일)
-- ALTER TABLE default.otel_metrics_summary   MODIFY TTL ... (동일)

-- 4) sessions: S3에 안 내린다. hot(EBS)만, 30일 DELETE만. (TO VOLUME 'cold' 없음)
ALTER TABLE default.hyperdx_sessions MODIFY TTL
    TimestampTime + INTERVAL 30 DAY DELETE;

-- 과거 파티션(예: 2주 넘은 날짜)을 cold 볼륨으로 즉시 이동
ALTER TABLE default.otel_logs   MOVE PARTITION '2026-06-14' TO VOLUME 'cold';
ALTER TABLE default.otel_traces MOVE PARTITION '2026-06-14' TO VOLUME 'cold';

-- 개별 part 단위 이동(특정 disk 지목: 긴급 hot 확보·재조정)
ALTER TABLE default.otel_logs   MOVE PART 'all_12345_12345_0' TO DISK 's3_cache';

-- 어느 파티션이 아직 hot에 있는지 확인 후 스크립트로 순차 이동
SELECT table, partition, disk_name, sum(bytes_on_disk)
FROM system.parts WHERE database='default' AND active AND disk_name='default'
GROUP BY table, partition ORDER BY partition;