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>
3.7 KiB
kind, slug, title, topic, project, status, sourceRevision, rootTreeNode, evidenceCapturedOn, assets, evidence, source, module
| kind | slug | title | topic | project | status | sourceRevision | rootTreeNode | evidenceCapturedOn | assets | evidence | source | module | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CONCEPT | adapter-outbound-persistence-mongo-c06 | 더 자주 갱신해도 고쳐지지 않는 문제라서 fencing을 얹는다 | state-machines-and-ownership | clean-architecture-backend-template | 게시 전 | 21234e38cdb9a926cbc92bb97a2aee2e4a7d2916 | concept:adapter-outbound-persistence-mongo-c06 | 2026-09-01 |
|
|
|
adapter-outbound-persistence-mongo |
더 자주 갱신해도 고쳐지지 않는 문제라서 fencing을 얹는다
MongoMigrationLock.fence()의 javadoc이 lease 만료와 보유자 정지가 다르다는 것을 적는다 — 첫 runner는 갱신했어야 할 그 순간에 돌고 있지 않다.
본문
MongoMigrationLock.fence()의 javadoc이 이 sub-scope에서 가장 정확한 문장을 담고 있다.
A lease expiring is not the same as its holder stopping. A runner paused inside a long
execute— a stop-the-world pause, a stalled network write — loses the lease on the server while its thread is still alive and still writing… Refreshing more often does not fix that: the first runner is not running at the moment it would refresh.
그래서 lease 위에 monotonic fencing token을 얹고, MongoCollectionMigrationLock.tryAcquire가 그 token을 lease를 부여하는 같은 조건부 update 안에서 서버가 증가시킨다("A token handed out anywhere else could be handed out twice"). held()는 owner 이름이 같아도 fence가 다르면 false를 반환한다 — 프로세스가 재시작했거나 운영자가 owner 문자열을 재사용한 경우다.
MongoCollectionMigrationLock 참조 위치
:::evidence key="adapter-outbound-persistence-mongo-c06" alt="코드베이스에서 MongoCollectionMigrationLock 를 검색한 출력 15줄. 이 기록이 세는 참조가 그 출력에 그대로 보인다." caption="MongoCollectionMigrationLock 코드베이스 검색 — 15줄 · exit 0" zoom="true" :::
heartbeat이 migration에게 넘겨진 이유
matchedCount를 쓰는 이유(같은 값을 다시 쓰면 modifiedCount가 0이라 소유권 판정이 뒤집힌다)도 두 곳에 적혀 있다. MongoMigrationHeartbeat은 이미 고쳐진 결함의 산물이다 — runner가 execute가 반환된 뒤에 한 번만 refresh했으므로, 40분짜리 execute는 35분 동안 만료된 lease를 들고 있었고 그 사이 두 번째 runner가 정당하게 획득해 같은 migration을 동시에 돌렸다. 이제 heartbeat이 migration에게 넘겨진다 — batch 경계를 아는 것은 migration뿐이기 때문이다.
첫 checkpoint가 전부 거부되던 두 결함
MongoCollectionMigrationLedger.saveCheckpoint에는 두 개의 결함 이력이 주석으로 남아 있다. upsert 하나로는 "매치할 게 없었다"와 "fence filter가 배제했다"를 구분할 수 없어 모든 migration의 첫 checkpoint가 "a newer migration runner owns the lease"로 거부됐고, 동시에 진짜 배제 경로는 unique index의 duplicate-key로 죽어 그 문장을 만드는 분기가 도달 불가였다. 지금은 replace-then-insert로 두 경우를 분리한다.
rollback이 없는 것도 명시적 결정이다
"A rollback method implies the reverse operation is always safe and always possible, and for a backfill that dropped a column's old values it is neither." 실패한 production 변경은 forward-fix migration으로 고친다.