출처 — HyperDX 내재화(실전 배포) 조사 자료

이 페이지는 개요, 스택 토폴로지·MongoDB 최소 규모, hot 스토리지(EBS), S3 cold 티어링, operator 토폴로지·다운타임, Keeper, 복제·멀티마스터·failover, 용량 산정, 블록 온리 튜닝, 버전 호환·업그레이드 열 페이지가 인용한 1차 조사(R1~R9)의 ## 출처 섹션을 모은 목록입니다. 중복을 걷어내고 주제별로 분류했습니다. 조사 기준일은 2026-07-16입니다.

각 조사가 다룬 범위는 ClickStack 4컴포넌트 배포 토폴로지와 MongoDB 최소 규모 운영, EBS(gp3/io2/io2 Block Express) hot 스토리지 선택, S3 cold 티어링 worked example, Altinity operator 기반 replication·sharding·다운타임 시나리오, ClickHouse Keeper 상세와 “큐가 아니다” 정정, 0.7TB/월 RUM 워크로드의 용량 산정입니다.

개별 URL의 등급(✓/≈/?/Ⓥ/Ⓑ)은 각 페이지 본문의 인라인 태그를 따릅니다 — 이 목록은 출처 목록일 뿐이고 등급을 재판정하지 않습니다. 기존 study-hugo 코퍼스(content/clickhouse/*, content/rum/*)와 선행 research(research/hyperdx-clickhouse/*) 참조는 각 페이지에서 relref로 인라인 연결하며 외부 URL이 아니므로 아래 목록에는 싣지 않았습니다.

ClickStack · HyperDX 공식

HyperDX·ClickStack 배포 문서와 스키마/TTL 설정, Browser SDK, 실물 Helm 차트 값·매니페스트입니다.

• ClickStack Helm 배포 가이드(2-차트 구조·PVC 보호) — clickhouse.com/docs/…/clickstack/deployment/helm
• ClickStack OTel Collector(역할·포트·배치·큐·사이징) — clickhouse.com/docs/…/clickstack/ingesting-data/otel-collector
• ClickStack Tables and schemas(otel_logs/traces/metrics·hyperdx_sessions ORDER BY·TTL·partition) — clickhouse.com/docs/…/clickstack/ingesting-data/schemas
• ClickStack Managing TTL(기본 3일·${TABLES_TTL}·ALTER MODIFY TTL·MATERIALIZE TTL) — clickhouse.com/docs/…/clickstack/ttl
• ClickStack Configuration options — clickhouse.com/docs/…/clickstack/config
• ClickStack Session replay(rrweb → hyperdx_sessions·Body/LogAttributes) — clickhouse.com/docs/…/clickstack/session-replay
• ClickStack OSS 소개(otel_* + hyperdx_sessions·ZSTD) — clickhouse.com/blog/clickstack-a-high-performance-oss-observability-stack
• HyperDX Browser SDK(4318 인제스트) — hyperdx.io/docs/install/browser
• HyperDX LOCAL.md(포트 8080/8000/8002/4317/4318) — github.com/hyperdxio/hyperdx/…/LOCAL.md
• ClickStack-helm-charts README(2-차트 구조·PVC 보호) — github.com/ClickHouse/ClickStack-helm-charts/…/README.md
• ClickStack-helm-charts values.yaml(members:1·mongo 5.0.32·storage 10Gi·otel mode:deployment·이미지 태그) — raw.githubusercontent.com/…/clickstack/values.yaml
• ClickStack-helm-charts UPGRADE.md — github.com/ClickHouse/ClickStack-helm-charts/…/UPGRADE.md

ClickHouse 공식 (docs · 엔지니어링)

ClickHouse Inc.의 스토리지·복제·insert·Keeper 문서, 압축·비용 엔지니어링 자료, GitHub PR입니다.

• External disks for storing data(type=s3/object_storage·cache disk·metadata_type/path) — clickhouse.com/docs/operations/storing-data
• Integrating S3(type=s3 예제·ReplicatedMergeTree on S3) — clickhouse.com/docs/integrations/s3
• Observability managing data(TTL·ttl_only_drop_parts·merge_with_ttl_timeout·MATERIALIZE TTL) — clickhouse.com/docs/observability/managing-data
• ClickHouse Keeper(NuRaft/Raft·ZK 호환·저장 대상·3노드) — clickhouse.com/docs/guides/sre/keeper/clickhouse-keeper
• Replicated* table engines(Keeper=메타만·데이터 직송·~10 znode/INSERT·블록 dedup) — clickhouse.com/docs/engines/table-engines/mergetree-family/replication
• Asynchronous inserts(wait 1/0·메모리 버퍼 크래시 유실·flush 트리거·async dedup off) — clickhouse.com/docs/optimize/asynchronous-inserts
• Selecting an insert strategy(동기 멱등·기본 dedup·배칭 1k~100k·at-least-once) — clickhouse.com/docs/best-practices/selecting-an-insert-strategy
• Deduplicating inserts on retries(replicated_deduplication_window(_seconds)·insert_deduplication_token) — clickhouse.com/docs/guides/developer/deduplicating-inserts-on-retries
• Using the Kafka table engine(Kafka engine→MV→MergeTree·파괴적 읽기·offset) — clickhouse.com/docs/integrations/kafka/kafka-table-engine
• Insert quorum parallel by default (PR #17567·insert_quorum_parallel 기본 1) — github.com/ClickHouse/ClickHouse/pull/17567
• nginx 로그 압축 170x(52~178x·낙관 상한 명시) — clickhouse.com/blog/log-compression-170x
• 압축 codec/ratio 개관 — clickhouse.com/resources/engineering/database-compression
• Observability cost optimization playbook(10~20x·7일 hot/cold·365 DELETE 예시) — clickhouse.com/resources/…/observability-cost-optimization-playbook
• Data Replication(멀티마스터·비동기·“only source data transferred”·SELECT은 Keeper 안 탐·블록 dedup) — clickhouse.com/docs/engines/table-engines/mergetree-family/replication
• system.replicas(is_leader “multiple leaders”·is_readonly·log_pointer vs log_max_index·absolute_delay) — clickhouse.com/docs/operations/system-tables/replicas
• system.replication_queue(type enum GET_PART/MERGE_PARTS/MUTATE_PART/DROP_RANGE·num_tries·postpone_reason) — clickhouse.com/docs/operations/system-tables/replication_queue
• Distributed 테이블 엔진(load_balancing·native connection failover) — clickhouse.com/docs/engines/table-engines/special/distributed
• Read consistency KB(insert_quorum·select_sequential_consistency·read-after-write) — clickhouse.com/docs/knowledgebase/read_consistency
• leader election 제거 — issue #10367 / PR #11639·#11795(20.6+ multiple leaders) — github.com/ClickHouse/ClickHouse/pull/11639
• TABLE_IS_READ_ONLY(에러 242) — Keeper/ZK 미가용 시 RMT read-only(issue #65424·#31052) — github.com/ClickHouse/ClickHouse/issues/65424

AWS 공식 (EBS · EC2 · S3)

EBS 볼륨 스펙·요금·내구성, gp3 상향 발표, io2 Block Express, EBS-optimized 인스턴스 대역, Multi-Attach입니다.

• EBS General Purpose SSD (gp3) User Guide(baseline 3,000/125·max 80,000 IOPS/2,000 MiB/s·64 TiB·durability·Outposts 16,000/1,000 한계) — docs.aws.amazon.com/ebs/…/general-purpose.html
• EBS Volume Types(io2 BE 256,000 IOPS/4,000 MB/s·99.999%·<500µs·gp3·io1) — aws.amazon.com/ebs/volume-types
• EBS Provisioned IOPS SSD (io2/io2 Block Express) User Guide — docs.aws.amazon.com/ebs/…/provisioned-iops.html
• EBS pricing(gp3 $0.08/GB·$0.005/IOPS·throughput·io2 tiered IOPS·$0.125/GB) — aws.amazon.com/ebs/pricing
• What’s New — gp3 max size & provisioned performance 상향(2025-09·80,000/2,000/64TiB) — aws.amazon.com/…/2025/09/amazon-ebs-size-provisioned-performance-gp3-volumes
• Storage Blog — larger and faster gp3(스트라이핑 대체·요금 동일·단일 볼륨 내구성) — aws.amazon.com/blogs/storage/improve-your-application-resiliency-with-larger-and-faster-gp3-volumes
• EBS-optimized instance types(r7g/m7g baseline vs burst 대역·IOPS) — docs.aws.amazon.com/AWSEC2/…/ebs-optimized.html
• EC2 r7g instances(EBS Gbps by size) — aws.amazon.com/ec2/instance-types/r7g
• EC2 memory-optimized specifications — docs.aws.amazon.com/ec2/…/mo.html
• Storage Blog — io2 Block Express higher DB performance(<500µs) — aws.amazon.com/blogs/storage/achieve-higher-database-performance-using-amazon-ebs-io2-block-express-volumes
• EBS Multi-Attach(io1/io2 전용·같은 AZ·최대 16 Nitro·cluster-aware FS 필수) — docs.aws.amazon.com/ebs/…/ebs-volumes-multi.html
• What’s New — io2 Block Express 전 리전 지원(2025-07) — aws.amazon.com/…/2025/07/amazon-ebs-io2-block-express
• What’s New — 128 volume attachments per instance(7세대·기본 32 dedicated) — aws.amazon.com/…/2023/08/amazon-ebs-128-volume-attachments-ec2-instance

Altinity · clickhouse-operator

Altinity operator/backup 문서와 GitHub 이슈, EBS/S3 KB, CRD install bundle, tiered-S3 예제 YAML, dedup/Keeper KB입니다.

• KB — AWS EC2 Storage(gp3 native·1-3 volume·7,000 IOPS/1,000 MiB/s safe·io2 500 MiB/s 낡은 주장) — kb.altinity.com/…/aws-ec2-storage
• Blog — Managing EBS gp3 Volumes in EKS(StorageClass ebs.csi.aws.com·throughput·allowVolumeExpansion·WaitForFirstConsumer) — altinity.com/blog/managing-ebs-gp3-volumes-in-eks
• Blog — reclaimPolicy(volumeClaimTemplate reclaimPolicy: Retain·라벨·수동삭제) — altinity.com/blog/preventing-clickhouse-storage-deletion-…reclaimpolicy
• operator docs/storage.md(volumeClaimTemplates·storageClassName) — github.com/Altinity/clickhouse-operator/…/storage.md
• operator CRD install bundle(reconcile.host.wait.probes·statefulSet.update·runtime·pdbManaged·podDistribution enum·suspend) — raw.githubusercontent.com/…/clickhouse-operator-install-bundle.yaml
• chi-example 03-persistent-volume-08-tiered-s3.yaml(storage_configuration.xml·s3_disk/s3_cache/정책 3종·hot=default) — raw.githubusercontent.com/…/03-persistent-volume-08-tiered-s3.yaml
• KB — AWS S3 Recipes(IAM 정책·IRSA SA annotation·use_environment_credentials·endpoint 포맷) — kb.altinity.com/…/aws-s3-recipes
• KB — S3Disk — kb.altinity.com/…/s3disk
• KB — Insert Deduplication / Idempotency(블록 체크섬을 Keeper /blocks znode 파티션별 저장) — kb.altinity.com/altinity-kb-schema-design/insert_deduplication
• KB — Using clickhouse-keeper — kb.altinity.com/…/clickhouse-keeper
• KB — DDLWorker and DDL queue problems(task_queue/ddl znode·직렬 sequential·task_max_lifetime) — kb.altinity.com/…/altinity-kb-ddlworker
• operator issue #1619 — reclaimPolicy Retain not honored — github.com/Altinity/clickhouse-operator/issues/1619
• operator issue #1385 — PVC volume template 확장 시 데이터 손실 — github.com/Altinity/clickhouse-operator/issues/1385
• operator issue #1408 — reclaimPolicy in ClickHouseKeeper — github.com/Altinity/clickhouse-operator/issues/1408
• clickhouse-backup issue #798 — IRSA self-assume 버그 — github.com/Altinity/clickhouse-backup/issues/798
• clickhouse-backup issue #1025 — use_environment_credentials IRSA — github.com/Altinity/clickhouse-backup/issues/1025
• KB — Load balancers(native 9000 TCP 한계·HTTP 8123 chproxy/HAProxy·클라 측 LB 3전략) — kb.altinity.com/…/load-balancers
• KB — Check replication & DDL queue(system.replicas/replication_queue 점검·SYSTEM RESTART/SYNC/RESTORE REPLICA) — kb.altinity.com/…/altinity-kb-check-replication-ddl-queue

MongoDB · MCK

HyperDX 메타데이터 스토어(MongoDB)의 최소 규모 운영, WiredTiger 캐시, 백업, Community/Enterprise 통합 operator입니다.

• mongodb/mongodb-kubernetes(MCK — community+enterprise 통합 리네임) — github.com/mongodb/mongodb-kubernetes
• Community Operator 샘플(pod resources·SCRAM·members:3) — github.com/mongodb/mongodb-kubernetes-operator/…/specify_pod_resources.yaml
• WiredTiger 스토리지 엔진(캐시 산식) — mongodb.com/docs/manual/core/wiredtiger
• MongoDB 백업 방법(mongodump vs Ops Manager) — mongodb.com/docs/manual/core/backups
• Community Operator 백업 부재 → mongodump(forums) — mongodb.com/community/forums/t/backups-from-kubernetes/161702

OpenTelemetry

Collector exporter·persistent queue·resiliency 문서입니다. CH 앞단에서 유실을 막는 근거로 씁니다.

• ClickHouse exporter README(create_schema·ttl_days) — github.com/open-telemetry/opentelemetry-collector-contrib/…/clickhouseexporter/README.md
• exporterhelper README(sending_queue.storage·queue_size 1000·num_consumers 10·block_on_overflow·WAL 재개) — github.com/open-telemetry/opentelemetry-collector/…/exporterhelper/README.md
• Resiliency(persistent queue·at-least-once·file_storage) — opentelemetry.io/docs/collector/resiliency

Kubernetes · 노드/스토리지 운영 (다운타임)

노드 급사·비정상 종료·PVC AZ 고정·force-detach·Karpenter disruption 등 EBS 기반 다운타임 시나리오의 근거입니다.

• Non-Graceful Node Shutdown GA(1.28·out-of-service taint·StatefulSet+PV 무한 Terminating 해결) — kubernetes.io/blog/2023/08/16/kubernetes-1-28-non-graceful-node-shutdown-ga
• Node Shutdowns 개념 — kubernetes.io/docs/concepts/cluster-administration/node-shutdown
• aws-ebs-csi-driver FAQ(6분 force-detach·노드 급사 reattach 지연·Multi-Attach error·완화책) — github.com/kubernetes-sigs/aws-ebs-csi-driver/…/faq.md
• rack2cloud — EBS PVC AZ lock-in(topology.ebs.csi.aws.com/zone nodeAffinity·WaitForFirstConsumer·volume node affinity conflict) — rack2cloud.com/kubernetes-pvc-stuck-volume-node-affinity
• Kubernetes 기본 pod eviction(unreachable/not-ready toleration 300s·node-monitor-grace-period) — devops-notes.com/kubernetes/eviction.html
• Karpenter — Disruption(voluntary/consolidation·do-not-disrupt·≥v1.0 VolumeAttachment 삭제 대기) — karpenter.sh/docs/concepts/disruption

커뮤니티 · 2차 자료 · 사례

third-party 기술 블로그와 실사례, RUM 리플레이 벤더 문서, EBS 실패율 실측입니다. 편향 가능성을 감안해 읽습니다.

• OneUptime — S3 cold storage worked example(disk/cache/policy/TTL/모니터링) — oneuptime.com/…/clickhouse-s3-cold-storage
• OneUptime — Storage policies(move_factor 정밀 정의·max_data_part_size·system.storage_policies) — oneuptime.com/…/clickhouse-storage-policies
• OneUptime — S3 disk storage — oneuptime.com/…/clickhouse-s3-disk-storage
• OneUptime — How ClickHouse Replication Protocol Works(복제 log/queue znode·block_numbers) — oneuptime.com/…/clickhouse-replication-protocol
• OneUptime — Configure quorum inserts(insert_quorum) — oneuptime.com/…/clickhouse-configure-quorum-inserts
• OneUptime — Estimate cluster size(5~10x·8x 예시·30% 헤드룸·노드/RAM/CPU 공식) — oneuptime.com/…/clickhouse-estimate-cluster-size
• OneUptime — ClickStack unified observability(세션/로그/트레이스/메트릭 스키마·기본 TTL 서술) — oneuptime.com/…/clickstack-unified-observability
• OneUptime — MongoDB Community Operator on k8s — oneuptime.com/…/mongodb-how-to-use-the-mongodb-community-operator
• OneUptime — WiredTiger cache size 구성 — oneuptime.com/…/mongodb-how-to-configure-wiredtiger-cache-size
• PlanetScale — The Real Failure Rate of EBS(실측 degradation 빈도·AZ 상관 장애·io2도 예외 아님) — planetscale.com/blog/the-real-fail-rate-of-ebs
• langfuse discussion #13013 — EKS S3 tiered storage(object_storage 신문법·move_factor 0.2·325GB/66일) — github.com/orgs/langfuse/discussions/13013
• Cloudflare — ClickHouse Capacity Estimation Framework(디스크 1순위·shardgroup 단일사본·Prophet 예보) — blog.cloudflare.com/clickhouse-capacity-estimation-framework
• Character.AI GPU 관측성(1520x 압축·샘플 후 34개월 150TB) — clickhouse.com/blog/scaling-observabilty-for-thousands-of-gpus-at-character-ai
• PostHog — session replay ingestion(스냅샷=S3 blob·CH엔 집계만 — HyperDX와 대비) — posthog.com/docs/how-posthog-works/recordings-ingestion
• rrweb(DOM 이벤트 기록·재구성·영상 아님) — github.com/rrweb-io/rrweb
• rrweb — optimize storage(샘플링·blocking으로 볼륨 절감) — rrweb.com/docs/recipes/optimize-storage
• Datadog — 브라우저 세션 리플레이 대역(<100KB/분) — docs.datadoghq.com/real_user_monitoring/session_replay/browser
• Datadog — 모바일 세션 리플레이 대역(iOS 12KB/s·Android 1.22KB/s) — docs.datadoghq.com/real_user_monitoring/session_replay/mobile/app_performance
• GlassFlow — OpenTelemetry to ClickHouse: Do You Need Kafka?(Kafka가 푸는 문제·안 필요한 경우·배칭) — glassflow.dev/blog/opentelemetry-to-clickhouse-do-you-need-kafka
• InfoQ — ClickHouse Keeper: Efficient ZooKeeper Alternative(C++·Raft·NuRaft 배경) — infoq.com/news/2023/12/clickhouse-keeper-raft
• matduggan — ClickHouse is winning the Observability Wars(1014x 압축·ES 23x 대비) — matduggan.com/clickhouse-is-winning-the-observability-wars
• Tasrie — ClickStack setup(docker/helm/k8s) — tasrieit.com/blog/clickstack-setup-guide-docker-helm-kubernetes-2026

R8·R9 추가 — 블록 온리 튜닝·버전/업그레이드

블록 스토리지 온리(08)·버전 호환·업그레이드(09) 조사에서 추가된 1차 출처입니다.

ClickHouse 공식 (설정·업그레이드·백업)

• MergeTree table settings(background/merge 풀·ttl_only_drop_parts·merge_with_ttl_timeout·parts_to_throw_insert) — clickhouse.com/docs/operations/settings/merge-tree-settings
• Server settings(background_pool_size·background_merges_mutations_concurrency_ratio) — clickhouse.com/docs/operations/server-configuration-parameters/settings
• ClickHouse Upgrades(compatibility·업그레이드 정책·채널) — clickhouse.com/docs/manage/updates
• compatibility 설정(정의·미변경 설정만 영향) — clickhouse.com/docs/operations/settings/settings#compatibility
• 2025 changelog(v25.12 JSON·v25.11 String 직렬화 → 다운그레이드 하한) — clickhouse.com/docs/whats-new/changelog/2025
• Backup/Restore 공식(BACKUP ... TO) — clickhouse.com/docs/operations/backup/overview
• issue #86837(25.8→25.3 marks 롤백 불가)·#68198·#68408(업그레이드 후 broken parts) — github.com/ClickHouse/ClickHouse/issues/86837

Altinity · operator (스토리지 확장·업그레이드)

• KB — Aggressive merges(background_pool_size 36·ratio·max_bytes_to_merge_at_max_space_in_pool ~150GB·주의) — kb.altinity.com/…/altinity-kb-aggressive_merges
• KB — Using the Kubernetes Operator(storageManagement provisioner·allowVolumeExpansion) — kb.altinity.com/altinity-kb-kubernetes
• Blog — What’s New in Altinity operator(provisioner Operator=STS 재생성 없이 온라인 확장) — altinity.com/blog/whats-new-in-altinity-clickhouse-operator
• operator 릴리즈노트(0.27.0 Keeper GA·async_replication 기본→Keeper 25.3+) — docs.altinity.com/releasenotes/…
• operator 0.27.1 — Artifact Hub(CH 21.11+·K8s 1.25+) — artifacthub.io/…/altinity-clickhouse-operator
• KB — ClickHouse 버전/업그레이드(스테이징 다운그레이드 리허설·혼합버전 증상) — kb.altinity.com/upgrade
• clickhouse-backup(업그레이드 전 백업·restore) — github.com/Altinity/clickhouse-backup
• issue #1263·#457 — CHI 재생성/AKS PVC 리사이즈 실패 사례 — github.com/Altinity/clickhouse-operator/issues/1263

AWS EBS (Elastic Volumes·스냅샷 롤백)

• Modify an EBS volume using Elastic Volumes — docs.aws.amazon.com/ebs/…/ebs-modify-volume.html
• Requirements for EBS volume modifications(OPTIMIZING 상태·수정 제약) — docs.aws.amazon.com/ebs/…/modify-volume-requirements.html
• EBS Elastic Volumes 6시간 쿨다운 폐지·24h당 4회(2026-01-15) — dev.classmethod.jp/…/ebs-elastic-volumes-4-modifications
• EBS 스냅샷 생성·복원(create-volume --snapshot-id --volume-type gp3) — docs.aws.amazon.com/ebs/…/ebs-restoring-volume.html

ClickStack (배포·업그레이드)

• ClickStack HyperDX-only 배포(외부 CH·MONGO_URI) — clickhouse.com/docs/…/clickstack/deployment/hyperdx-only

추가 조사 — S3 백엔드·티어링 비용 경로 (2026-08)

self-host에서 S3를 어디까지 쓸 수 있는지 다시 조사한 결과 가운데 이 챕터로 들어온 출처입니다. 티어링의 요청 단가, HyperDX가 붙는 CH 계정 권한, Express One Zone 경계가 여기에 속합니다.

• ClickStack Going to production(HyperDX용 CH 계정은 readonly + max_rows_to_read(최소 100만)·read_overflow_mode·cancel_http_readonly_queries_on_client_close·wait_end_of_query 변경 권한 필요, 기본 계정 대신 전용 사용자 권고) — clickhouse.com/docs/…/clickstack/production
• AWS Price List Bulk API — ap-northeast-2 S3 오퍼 index.json(서울 S3 Standard GB 단가·PUT/GET/DELETE 요청 단가를 직접 조회한 경로. gp3 GB 단가는 같은 API의 EC2 오퍼를 같은 방식으로 조회한다) — pricing.us-east-1.amazonaws.com/…/AmazonS3/current/ap-northeast-2/index.json
• GitHub issue — S3 Express One Zone 디렉터리 버킷 엔드포인트 IncompleteBody #72078 — github.com/ClickHouse/ClickHouse/issues/72078

아래 넷은 본문이 인용하지만 1차 URL이 원 조사 산출물에 남아 있지 않아 이 목록에 아직 싣지 못했습니다 — S3 Gateway VPC Endpoint의 무료 정책과 NAT Gateway 데이터 처리요금(서울), S3 Express One Zone 제공 리전 목록, wide part 파일 수 증폭 실측 예시(컬럼 109개 → part당 227 파일, ClickHouse 공식 KB), Packed storage min_level_for_full_part_storage와 system.parts.part_storage_type 문서. URL을 확보하는 대로 이 절에 채웁니다.

시점 기준 2026-08.