feat(account): retention TTL reaper (2-year legal hold)

Daily background reaper purges the deletion dossier past its two-year TTL: every
retained_identities row by detached_at (covering unlink/change on live accounts too),
and — for accounts tombstoned before the cutoff — the retained feedback thread plus the
dossier PII (deleted_display_name, last_login_ip). Chat is kept (a shared game artifact)
and the tombstone row stays. Started from main next to the guest reaper. Integration
test covers the cutoff boundary and the deleted-account feedback/PII purge.
This commit is contained in:
Ilia Denisov
2026-07-03 12:08:55 +02:00
parent 52e6378f40
commit d889edfdb9
3 changed files with 162 additions and 0 deletions
+9
View File
@@ -162,6 +162,15 @@ func run(ctx context.Context, cfg config.Config, logger *zap.Logger) error {
zap.Duration("interval", cfg.GuestReapInterval), zap.Duration("interval", cfg.GuestReapInterval),
zap.Duration("retention", cfg.GuestRetention)) zap.Duration("retention", cfg.GuestRetention))
// Purge the account-deletion legal dossier past its retention TTL: the
// retained-identities journal, and the feedback thread + dossier PII of long-deleted
// accounts (chat is kept). Checked daily; the TTL is a two-year policy constant.
retentionReaper := account.NewRetentionReaper(accounts, account.RetentionTTL, logger)
go retentionReaper.Run(ctx, 24*time.Hour)
logger.Info("retention reaper started",
zap.Duration("interval", 24*time.Hour),
zap.Duration("retention", account.RetentionTTL))
// Re-evaluate moderated-chat write access when a temporary block self-expires: // Re-evaluate moderated-chat write access when a temporary block self-expires:
// no operator action fires then, so the sweeper emits the chat-access-changed // no operator action fires then, so the sweeper emits the chat-access-changed
// event for lapsed blocks and the gateway re-pushes the chat-gate command. // event for lapsed blocks and the gateway re-pushes the chat-gate command.
+89
View File
@@ -8,10 +8,15 @@ import (
"github.com/go-jet/jet/v2/postgres" "github.com/go-jet/jet/v2/postgres"
"github.com/google/uuid" "github.com/google/uuid"
"go.uber.org/zap"
"scrabble/backend/internal/postgres/jet/backend/table" "scrabble/backend/internal/postgres/jet/backend/table"
) )
// RetentionTTL bounds how long the account-deletion legal dossier is kept before the
// reaper purges it: two years from the detach/deletion event (owner policy, 2026-07-03).
const RetentionTTL = 2 * 365 * 24 * time.Hour
// Reasons recorded on a retained_identities row: what detached the credential from its // Reasons recorded on a retained_identities row: what detached the credential from its
// account. The row is written just before the live identities row is removed, preserving // account. The row is written just before the live identities row is removed, preserving
// the legal dossier (which email/vk/tg was linked, and when) even as the identity frees // the legal dossier (which email/vk/tg was linked, and when) even as the identity frees
@@ -63,3 +68,87 @@ func (s *Store) StampLastLogin(ctx context.Context, accountID uuid.UUID, ip stri
} }
return nil return nil
} }
// ReapExpiredRetention purges retention data whose event is older than cutoff: every
// retained_identities row by its detached_at (covering unlink/change on live accounts as
// well as deleted ones), plus — for accounts tombstoned before cutoff — the retained
// feedback thread and the dossier PII (deleted_display_name, last_login_ip). Chat is kept
// (a shared game artifact), and the tombstone account row itself stays (its no-cascade
// foreign keys). It returns how many journal rows and feedback messages were removed.
func (s *Store) ReapExpiredRetention(ctx context.Context, cutoff time.Time) (identities, feedback int64, err error) {
cut := postgres.TimestampzT(cutoff)
delJournal := table.RetainedIdentities.DELETE().
WHERE(table.RetainedIdentities.DetachedAt.LT(cut))
res, err := delJournal.ExecContext(ctx, s.db)
if err != nil {
return 0, 0, fmt.Errorf("account: reap retained identities: %w", err)
}
identities, _ = res.RowsAffected()
expired := postgres.SELECT(table.Accounts.AccountID).
FROM(table.Accounts).
WHERE(table.Accounts.DeletedAt.IS_NOT_NULL().AND(table.Accounts.DeletedAt.LT(cut)))
delFeedback := table.FeedbackMessages.DELETE().
WHERE(table.FeedbackMessages.AccountID.IN(expired))
fbRes, err := delFeedback.ExecContext(ctx, s.db)
if err != nil {
return identities, 0, fmt.Errorf("account: reap deleted feedback: %w", err)
}
feedback, _ = fbRes.RowsAffected()
clearPII := table.Accounts.UPDATE(table.Accounts.DeletedDisplayName, table.Accounts.LastLoginIP).
SET(postgres.NULL, postgres.NULL).
WHERE(
table.Accounts.DeletedAt.IS_NOT_NULL().
AND(table.Accounts.DeletedAt.LT(cut)).
AND(table.Accounts.DeletedDisplayName.IS_NOT_NULL().
OR(table.Accounts.LastLoginIP.IS_NOT_NULL())),
)
if _, err := clearPII.ExecContext(ctx, s.db); err != nil {
return identities, feedback, fmt.Errorf("account: clear expired dossier PII: %w", err)
}
return identities, feedback, nil
}
// RetentionReaper periodically purges expired account-deletion retention data via
// Store.ReapExpiredRetention, mirroring GuestReaper: one background goroutine started once
// from main.
type RetentionReaper struct {
store *Store
ttl time.Duration
clock func() time.Time
log *zap.Logger
}
// NewRetentionReaper constructs a reaper purging retention data older than ttl. log may be
// nil.
func NewRetentionReaper(store *Store, ttl time.Duration, log *zap.Logger) *RetentionReaper {
if log == nil {
log = zap.NewNop()
}
return &RetentionReaper{
store: store,
ttl: ttl,
clock: func() time.Time { return time.Now().UTC() },
log: log,
}
}
// Run purges expired retention data on each tick until ctx is cancelled.
func (r *RetentionReaper) Run(ctx context.Context, interval time.Duration) {
ticker := time.NewTicker(interval)
defer ticker.Stop()
for {
select {
case <-ctx.Done():
return
case <-ticker.C:
idn, fb, err := r.store.ReapExpiredRetention(ctx, r.clock().Add(-r.ttl))
if err != nil {
r.log.Warn("retention reap failed", zap.Error(err))
} else if idn > 0 || fb > 0 {
r.log.Info("reaped expired retention", zap.Int64("identities", idn), zap.Int64("feedback", fb))
}
}
}
}
@@ -6,10 +6,12 @@ import (
"context" "context"
"database/sql" "database/sql"
"testing" "testing"
"time"
"github.com/google/uuid" "github.com/google/uuid"
"scrabble/backend/internal/account" "scrabble/backend/internal/account"
"scrabble/backend/internal/accountdelete"
) )
// lastLoginIP reads an account's stamped last-login IP ("" when unset). // lastLoginIP reads an account's stamped last-login IP ("" when unset).
@@ -126,3 +128,65 @@ func TestStampLastLoginThrottles(t *testing.T) {
t.Fatalf("throttled ip = %q, want unchanged 1.2.3.4", got) t.Fatalf("throttled ip = %q, want unchanged 1.2.3.4", got)
} }
} }
// TestReapExpiredRetention: the reaper keeps journal rows newer than the cutoff and purges
// older ones, and drops a long-deleted account's feedback thread + dossier PII.
func TestReapExpiredRetention(t *testing.T) {
ctx := context.Background()
store := account.NewStore(testDB)
deleter := accountdelete.NewDeleter(testDB)
// An unlinked provider leaves a journal row detached "now".
tgExt := "tg-" + uuid.NewString()
acc, err := store.ProvisionByIdentity(ctx, account.KindTelegram, tgExt)
if err != nil {
t.Fatalf("provision: %v", err)
}
if err := store.AttachIdentity(ctx, acc.ID, account.KindEmail, "keep-"+uuid.NewString()+"@example.com", true); err != nil {
t.Fatalf("attach email: %v", err)
}
if err := store.RemoveIdentity(ctx, acc.ID, account.KindTelegram); err != nil {
t.Fatalf("unlink: %v", err)
}
// A cutoff before the detach keeps the row.
if _, _, err := store.ReapExpiredRetention(ctx, time.Now().Add(-time.Hour)); err != nil {
t.Fatalf("reap (early cutoff): %v", err)
}
if got := retainedRows(t, acc.ID); len(got) != 1 {
t.Fatalf("journal after early-cutoff reap = %+v, want kept", got)
}
// A cutoff after the detach purges it.
if _, _, err := store.ReapExpiredRetention(ctx, time.Now().Add(time.Hour)); err != nil {
t.Fatalf("reap (late cutoff): %v", err)
}
if got := retainedRows(t, acc.ID); len(got) != 0 {
t.Fatalf("journal after late-cutoff reap = %+v, want purged", got)
}
// A deleted account past the cutoff loses its feedback thread and dossier PII.
del, _, err := store.ProvisionTelegram(ctx, "tg-"+uuid.NewString(), "ru", "", "Иван", "")
if err != nil {
t.Fatalf("provision deletee: %v", err)
}
if _, err := testDB.ExecContext(ctx,
"INSERT INTO feedback_messages (message_id, account_id, body, channel) VALUES ($1, $2, 'hi', 'web')",
uuid.New(), del.ID); err != nil {
t.Fatalf("insert feedback: %v", err)
}
if err := deleter.AnonymizeAndTombstone(ctx, del.ID); err != nil {
t.Fatalf("delete: %v", err)
}
if _, fb, err := store.ReapExpiredRetention(ctx, time.Now().Add(time.Hour)); err != nil || fb == 0 {
t.Fatalf("reap deleted = (fb %d, err %v), want fb>=1", fb, err)
}
if name, _ := deletedFields(t, del.ID); name != "" {
t.Errorf("deleted_display_name after reap = %q, want cleared", name)
}
var fbCount int
if err := testDB.QueryRowContext(ctx, "SELECT count(*) FROM feedback_messages WHERE account_id = $1", del.ID).Scan(&fbCount); err != nil {
t.Fatalf("count feedback: %v", err)
}
if fbCount != 0 {
t.Errorf("feedback rows after reap = %d, want 0", fbCount)
}
}