feat: verify durable client spool recovery

This commit is contained in:
2026-09-06 06:41:38 +00:00
parent 62c0052979
commit fe9fadbb01
5 changed files with 199 additions and 1 deletions
+1 -1
View File
@@ -86,7 +86,7 @@ func (store *Store) AcceptCommand(ctx context.Context, command Command, accepted
if err != nil {
return Acceptance{}, err
}
_, err = tx.ExecContext(ctx, `INSERT INTO commands(issue_uuid, immutable_sha256, command_revision, phase, terminal, total_charged_bytes, closeout_remaining_bytes, accepted_at) VALUES (?, ?, ?, ?, ?, ?, ?, ?)`, command.IssueUUID[:], command.ImmutableSHA256[:], command.Revision, command.Phase, boolInt(command.Terminal), decision.CommandTotalCharged, decision.CloseoutRemaining, acceptedAt.UnixNano())
_, err = tx.ExecContext(ctx, `INSERT INTO commands(issue_uuid, immutable_sha256, command_revision, phase, terminal, base_charged_bytes, total_charged_bytes, closeout_remaining_bytes, accepted_at) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)`, command.IssueUUID[:], command.ImmutableSHA256[:], command.Revision, command.Phase, boolInt(command.Terminal), charge, decision.CommandTotalCharged, decision.CloseoutRemaining, acceptedAt.UnixNano())
if err != nil {
return Acceptance{}, err
}
+1
View File
@@ -61,6 +61,7 @@ CREATE TABLE commands (
phase INTEGER NOT NULL CHECK(phase BETWEEN 1 AND 11),
terminal INTEGER NOT NULL DEFAULT 0 CHECK(terminal IN (0, 1)),
output_charged_bytes INTEGER NOT NULL DEFAULT 0 CHECK(output_charged_bytes >= 0),
base_charged_bytes INTEGER NOT NULL CHECK(base_charged_bytes > 0),
total_charged_bytes INTEGER NOT NULL CHECK(total_charged_bytes >= 0),
closeout_remaining_bytes INTEGER NOT NULL CHECK(closeout_remaining_bytes >= 0),
next_local_ordinal INTEGER NOT NULL DEFAULT 1 CHECK(next_local_ordinal > 0),
+137
View File
@@ -0,0 +1,137 @@
package spool
import (
"bytes"
"context"
"database/sql"
"errors"
"fmt"
"github.com/rvbox/rvbox/internal/domain"
)
var (
ErrIntegrityCheck = errors.New("client spool SQLite integrity check failed")
ErrQuotaCounterMismatch = errors.New("client spool quota counter mismatch")
ErrStoredPayloadChecksum = errors.New("client spool stored payload checksum mismatch")
)
type RecoveryReport struct {
EventsChecked uint64
ScriptsChecked uint64
}
// Check verifies state that SQLite's own structural check cannot express. The
// daemon runs it asynchronously during startup and exposes failure as dirty
// local health; it never invents a replacement identity or counter value.
func (store *Store) Check(ctx context.Context) (RecoveryReport, error) {
if store == nil || store.db == nil {
return RecoveryReport{}, sql.ErrConnDone
}
if err := quickCheck(ctx, store.db); err != nil {
return RecoveryReport{}, err
}
if err := checkQuotaCounters(ctx, store.db); err != nil {
return RecoveryReport{}, err
}
return checkPayloads(ctx, store.db, store.maxScriptBytes)
}
func quickCheck(ctx context.Context, database *sql.DB) error {
rows, err := database.QueryContext(ctx, `PRAGMA quick_check(100)`)
if err != nil {
return fmt.Errorf("%w: query failed", ErrIntegrityCheck)
}
defer rows.Close()
count := 0
for rows.Next() {
var result string
if err := rows.Scan(&result); err != nil {
return fmt.Errorf("%w: unreadable result", ErrIntegrityCheck)
}
count++
if result != "ok" {
return fmt.Errorf("%w: %s", ErrIntegrityCheck, result)
}
}
if err := rows.Err(); err != nil || count != 1 {
return fmt.Errorf("%w: incomplete result", ErrIntegrityCheck)
}
return nil
}
func checkQuotaCounters(ctx context.Context, database *sql.DB) error {
var invalid int
err := database.QueryRowContext(ctx, `SELECT count(*) FROM commands AS c
WHERE c.total_charged_bytes != c.base_charged_bytes +
COALESCE((SELECT sum(charged_bytes) FROM events WHERE issue_uuid = c.issue_uuid), 0) +
COALESCE((SELECT charged_bytes FROM scripts WHERE issue_uuid = c.issue_uuid), 0)
OR c.output_charged_bytes !=
COALESCE((SELECT sum(charged_bytes) FROM events WHERE issue_uuid = c.issue_uuid AND output = 1), 0)`).Scan(&invalid)
if err != nil {
return err
}
var stored, computed uint64
if err := database.QueryRowContext(ctx, `SELECT client_total_charged_bytes, COALESCE((SELECT sum(total_charged_bytes) FROM commands), 0) FROM spool_counters WHERE singleton = 1`).Scan(&stored, &computed); err != nil {
return err
}
if invalid != 0 || stored != computed {
return ErrQuotaCounterMismatch
}
return nil
}
func checkPayloads(ctx context.Context, database *sql.DB, maxScriptBytes uint64) (RecoveryReport, error) {
var report RecoveryReport
rows, err := database.QueryContext(ctx, `SELECT issue_uuid, payload, payload_sha256 FROM events ORDER BY issue_uuid, local_ordinal`)
if err != nil {
return report, err
}
for rows.Next() {
var owner, payload, digest []byte
if err := rows.Scan(&owner, &payload, &digest); err != nil {
_ = rows.Close()
return report, err
}
computed := immutableDigest(payload)
if _, valid := copiedUUID(owner); !valid || len(digest) != 32 || !bytes.Equal(computed[:], digest) {
_ = rows.Close()
return report, ErrStoredPayloadChecksum
}
report.EventsChecked++
}
if err := rows.Close(); err != nil {
return report, err
}
scripts, err := database.QueryContext(ctx, `SELECT issue_uuid, received_raw_bytes, stored_bytes, stored_data FROM scripts ORDER BY issue_uuid`)
if err != nil {
return report, err
}
defer scripts.Close()
for scripts.Next() {
var owner, stored []byte
var received, storedBytes uint64
if err := scripts.Scan(&owner, &received, &storedBytes, &stored); err != nil {
return report, err
}
if _, valid := copiedUUID(owner); !valid || storedBytes != uint64(len(stored)) {
return report, ErrScriptState
}
if _, err := decompressScript(stored, received, maxScriptBytes); err != nil {
return report, err
}
report.ScriptsChecked++
}
return report, scripts.Err()
}
// copiedUUID is intentionally kept here to make a corrupted BLOB type obvious
// at the recovery boundary rather than relying on SQLite's dynamic conversions.
func copiedUUID(value []byte) (domain.UUID, bool) {
var result domain.UUID
if len(value) != len(result) {
return result, false
}
copy(result[:], value)
return result, validUUID(result)
}
+48
View File
@@ -0,0 +1,48 @@
package spool
import (
"context"
"errors"
"path/filepath"
"testing"
"time"
)
func TestCheckSpoolPayloadsAndCounters_HP_CLIENT_09(t *testing.T) {
t.Parallel()
ctx := context.Background()
store := openTestStore(t, ctx, filepath.Join(t.TempDir(), "spool"), DefaultTombstoneLimit)
issue := testUUID(t, "019c46f1-1d02-7000-8000-000000000041")
if _, err := store.AcceptCommand(ctx, testCommand(issue, []byte("recovery")), time.Now().UTC()); err != nil {
t.Fatal(err)
}
if _, err := store.AppendEvent(ctx, issue, EventInput{Kind: 1, Compression: 1, Payload: []byte("durable event"), CreatedAt: time.Now().UTC()}); err != nil {
t.Fatal(err)
}
report, err := store.Check(ctx)
if err != nil || report.EventsChecked != 1 || report.ScriptsChecked != 0 {
t.Fatalf("initial Check = %#v, %v", report, err)
}
if _, err := store.db.ExecContext(ctx, `UPDATE events SET payload = ? WHERE issue_uuid = ?`, []byte("tampered"), issue[:]); err != nil {
t.Fatal(err)
}
if _, err := store.Check(ctx); !errors.Is(err, ErrStoredPayloadChecksum) {
t.Fatalf("tampered event Check error = %v, want ErrStoredPayloadChecksum", err)
}
}
func TestCheckSpoolRejectsCounterDrift_BH_CLIENT_03(t *testing.T) {
t.Parallel()
ctx := context.Background()
store := openTestStore(t, ctx, filepath.Join(t.TempDir(), "spool"), DefaultTombstoneLimit)
issue := testUUID(t, "019c46f1-1d02-7000-8000-000000000042")
if _, err := store.AcceptCommand(ctx, testCommand(issue, []byte("recovery")), time.Now().UTC()); err != nil {
t.Fatal(err)
}
if _, err := store.db.ExecContext(ctx, `UPDATE spool_counters SET client_total_charged_bytes = client_total_charged_bytes + 1`); err != nil {
t.Fatal(err)
}
if _, err := store.Check(ctx); !errors.Is(err, ErrQuotaCounterMismatch) {
t.Fatalf("counter-drift Check error = %v, want ErrQuotaCounterMismatch", err)
}
}
+12
View File
@@ -233,6 +233,18 @@ layer = "unit"
status = "implemented"
tests = ["internal/client/spool/script_test.go:TestScriptUploadDurableExactReplayAndCommit_HP_SCRIPT_01"]
[[requirements]]
id = "HP-CLIENT-09"
layer = "unit"
status = "implemented"
tests = ["internal/client/spool/recovery_test.go:TestCheckSpoolPayloadsAndCounters_HP_CLIENT_09"]
[[requirements]]
id = "BH-CLIENT-03"
layer = "unit"
status = "implemented"
tests = ["internal/client/spool/recovery_test.go:TestCheckSpoolRejectsCounterDrift_BH_CLIENT_03"]
[[requirements]]
id = "HP-WINCTX-02"
layer = "integration"