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>
70 lines
3.9 KiB
Plaintext
70 lines
3.9 KiB
Plaintext
# javadoc 은 넣기를 upsert 로 정의한다
|
|
19: * <p>An enqueue is an upsert: the same file reported twice updates the open item rather than adding
|
|
20: * a second one. Reconciliation runs on a schedule and re-raises whatever it still cannot settle, so
|
|
21: * an append-only queue would grow one row per sweep per unresolved file and bury the distinct
|
|
22: * problems under repetitions of the same one.
|
|
# 저장소 javadoc 은 최신 이유가 이긴다고 적는다
|
|
24: /**
|
|
25: * Re-raises an open item instead of adding a second one.
|
|
26: *
|
|
27: * <p>Reconciliation is retried on a schedule, so the same file reaches the queue repeatedly. One
|
|
28: * open item per file keeps the queue a worklist rather than a failure log; the newest reason wins
|
|
29: * because it describes the most recent evidence.
|
|
30: */
|
|
|
|
# 구현은 갱신하고, 0 이면 저장한다
|
|
42: @Override
|
|
43: public void enqueue(FileId fileId, String reasonCode) {
|
|
44: Instant now = clock.instant();
|
|
45: if (items.refreshPending(fileId.value(), reasonCode, now) > 0) {
|
|
46: return;
|
|
47: }
|
|
48: items.save(
|
|
49: new RecoveryItemEntity(UUID.randomUUID(), fileId.value(), reasonCode, STATUS_PENDING, now));
|
|
50: }
|
|
# 그 저장이 persist 로 갈지 merge 로 갈지 정하는 것
|
|
# 엔티티의 @Version 필드: 0
|
|
# Persistable 구현: 0
|
|
|
|
# 데이터베이스가 파일당 열린 항목을 하나로 강제한다
|
|
52--- At most one open item per file: the queue is a worklist, not a log of every sweep.
|
|
53:CREATE UNIQUE INDEX uq_fs_recovery_open
|
|
54- ON fs_recovery_item (file_id)
|
|
55- WHERE status = 'PENDING';
|
|
56-
|
|
|
|
# 호출자가 실제로 받는 예외
|
|
195: private <T> T executeLegacy(TransactionTemplate template, Supplier<T> action) {
|
|
196- try {
|
|
197- return template.execute(status -> action.get());
|
|
198- } catch (RuntimeException failure) {
|
|
199- throw exceptionTranslator
|
|
200- .translate(failure)
|
|
201- .map(RuntimeException.class::cast)
|
|
202- .orElse(failure);
|
|
203- }
|
|
204- }
|
|
20: "23505", OperationalError.DB_UNIQUE_VIOLATION,
|
|
|
|
# 이 넣기를 부르는 프로덕션 코드와 그 경계
|
|
application-core/src/main/java/dev/caskeleton/application/fileserver/upload/DefaultFinalizeUploadService.java:323: recoveryQueue.enqueue(verifying.fileId(), "READY_COMMIT_UNCONFIRMED");
|
|
application-core/src/main/java/dev/caskeleton/application/fileserver/recovery/DefaultFileReconciliationService.java:220: recoveryQueue.enqueue(record.fileId(), reasonCode);
|
|
} catch (RuntimeException exception) {
|
|
// The object is on the volume and the record does not say so — the definition of an ambiguous
|
|
// completion. Enqueued in its own boundary, because the one that failed took nothing with it.
|
|
transactions.inWrite(
|
|
() -> {
|
|
recoveryQueue.enqueue(verifying.fileId(), "READY_COMMIT_UNCONFIRMED");
|
|
});
|
|
throw new AmbiguousCompletionException(
|
|
"content was published but the READY commit could not be confirmed",
|
|
# 조정 쪽 진입점을 부르는 프로덕션 코드: 0
|
|
# 조정이 집어 올리는 대상의 조건
|
|
27: private static final Duration STALE_AFTER = Duration.ofMinutes(5);
|
|
71: clock.instant().minus(STALE_AFTER),
|
|
|
|
# 같은 저장소가 승패를 데이터베이스에 물어보는 자리
|
|
application-core/src/main/java/dev/caskeleton/application/notification/platform/dispatch/NotificationRequestInsertOutcome.java:16: * anything: {@code ON CONFLICT DO NOTHING RETURNING} returns a row when the caller won and no rows
|
|
messaging/messaging-inbox-jdbc-postgresql/src/main/java/dev/caskeleton/messaging/inbox/JdbcInboxRepository.java:20: * <p>Reservation is an {@code INSERT ... ON CONFLICT DO NOTHING} whose affected-row count is the
|
|
messaging/messaging-inbox-jdbc-postgresql/src/main/java/dev/caskeleton/messaging/inbox/JdbcInboxRepository.java:36: ON CONFLICT (message_id, consumer_id) DO NOTHING
|