DongHyeonkaandClaude Opus 5
b2963105a8
The keycloak project ended with four open questions that design could not
settle. A two-VM lab was built to answer them by measurement, and this is
that material: 26 experiments, 125 raw command outputs, 22 browser captures.
Follows the import procedure in README.md.
source/ the originating repository verbatim — 78 documents, 28 SVGs,
8 manifests, plus .source-revision recording the commit
final/ the SSOT
document.md 729 lines written from the 29 experiment documents, not
concatenated: what was predicted, what was measured, and
where the measurement itself was wrong
evidence/raw 125 outputs, flattened to <experiment>__<file> because
the originals collided (01-baseline.txt appeared three
times) and the audit only globs the top level
evidence/meta one per raw file; command and exitCode are null and the
README says why rather than inventing them
evidence/browser 22 captures
assets/ three diagrams through techviz
.techviz/ their VizSpecs
A separate project rather than an addition to keycloak: the B-layer answers
that project's four questions, but the A, C and D layers are about cluster
failure, SSO and operations, and one document.md should hold one subject.
The four question records there can point here through 관계.
Recorded rather than papered over: only three of the 28 diagrams were
remade. The repository forbids hand-drawn SVG and forbids titles inside the
canvas; all 28 originals carry both, so converting them is redrawing, not
reformatting. They stay in source/ and the gap is written into the document.
verify-pipeline.py passes. audit-records.py reports no issues.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
54 lines
3.2 KiB
Markdown
54 lines
3.2 KiB
Markdown
---
|
|
kind: CONCEPT
|
|
slug: adapter-outbound-cache-redis-c13
|
|
title: 등록되지 않은 스크립트는 서버에 닿을 방법이 없다
|
|
topic: transaction-and-consistency-models
|
|
project: clean-architecture-backend-template
|
|
status: 게시 전
|
|
sourceRevision: 21234e38cdb9a926cbc92bb97a2aee2e4a7d2916
|
|
rootTreeNode: concept:adapter-outbound-cache-redis-c13
|
|
evidenceCapturedOn: 2026-09-01
|
|
assets:
|
|
- key: adapter-outbound-cache-redis-c13
|
|
file: ../../../final/evidence/rendered/adapter-outbound-cache-redis-c13.svg
|
|
- key: adapter-outbound-cache-redis-c13-diagram
|
|
file: ../../../final/assets/diagrams/adapter-outbound-cache-redis-c13.svg
|
|
evidence:
|
|
- ../../../final/evidence/raw/adapter-outbound-cache-redis-c13.txt
|
|
source:
|
|
- 원본 분석 절은 analysis/10-adapter-outbound-cache-redis.md#L701 이다.
|
|
module: adapter-outbound-cache-redis
|
|
---
|
|
|
|
# 등록되지 않은 스크립트는 서버에 닿을 방법이 없다
|
|
|
|
`RedisScriptRegistry`가 등록을 배포 단계로 두고, 트랜잭션 창은 감시 키에서 정한 노드에 고정된다.
|
|
|
|
## 본문
|
|
|
|
<!-- body:start -->
|
|
|
|
`RedisScriptRegistry`의 규칙 — "Registration is a deployment step, not a request-time one. A script that was never registered has no digest and therefore **no way to reach the server**, which is what makes 'only reviewed scripts run' a structural property rather than a convention." 같은 identity에 다른 body를 등록하면 거부하고, 실행 시점에도 body가 등록본과 같은지 다시 본다.
|
|
|
|
## 등록과 실행의 분리
|
|
|
|
:::evidence key="adapter-outbound-cache-redis-c13-diagram" alt="배포 시 스크립트 등록과 digest 대조와 EVALSHA 실행이 왼쪽에서 오른쪽으로 이어지는 구조" caption="등록과 실행의 분리" zoom="false"
|
|
:::
|
|
|
|
README가 주장하는 복구 사슬 `EVALSHA → NOSCRIPT → SCRIPT LOAD → digest verify → EVALSHA`는 **실재한다** — `LettuceRedisScriptOperations:26` javadoc이 "NOSCRIPT is the one failure retried automatically"라고 적고, `:129-133`이 `RedisNoScriptException` 또는 메시지 접두 `NOSCRIPT`를 잡아 `:100`에서 `registry.forget(script.id())`를 호출한다. 다음 호출이 `digest(...)`에서 다시 `SCRIPT LOAD`한다.
|
|
|
|
## RedisScriptRegistry 참조 위치
|
|
|
|
:::evidence key="adapter-outbound-cache-redis-c13" alt="코드베이스에서 RedisScriptRegistry 를 검색한 출력 7줄. 이 기록이 세는 참조가 그 출력에 그대로 보인다." caption="RedisScriptRegistry 코드베이스 검색 — 7줄 · exit 0" zoom="true"
|
|
:::
|
|
|
|
## Cluster에서 MULTI 창을 노드에 고정하는 방법
|
|
|
|
`RedisTransactionRunner`는 Cluster에서의 `MULTI` 문제를 정면으로 다룬다. javadoc이 문제와 해법을 적는다 — 다른 레인은 명령마다 슬롯 소유 노드로 라우팅하는데 "that is exactly what a `MULTI` window must not do: the queued commands would be spread across nodes and none of them would be part of the same window." 해법은 연결이 아니라 **라우팅 결정**이었다 — 감시 키(또는 명시적 슬롯 태그)에서 노드를 정해 레인을 고정한다.
|
|
|
|
## 감시 키가 없는 Cluster 트랜잭션을 거부하는 이유
|
|
|
|
"the keys the callback will queue are not known until the callback runs, which is after the window is open."
|
|
|
|
<!-- body:end -->
|