can1357/oh-my-pi · error · AIError.ConfigurationError

Failed to open auth database at

Error message

Failed to open auth database at '${dbPath}' after ${maxAttempts} attempts: ${lastBusyError?.message}

What it means

SqliteAuthCredentialStore opens its Bun SQLite database with a bounded retry loop for SQLITE_BUSY contention. If the database still cannot be opened after maxAttempts, it throws AIError.ConfigurationError including the db path, attempt count, and the last busy error's message (cause preserved).

Solutions

Ensure no other process is holding the DB: close other omp/CLI sessions and retry
Move the auth database off network/synced filesystems to local disk (check the configured dbPath)
Check file permissions and free disk space at dbPath; delete a corrupt DB only if re-auth is acceptable
Increase maxAttempts/baseDelayMs in environments with heavy concurrent access

Example fix

// before: auth dir on NFS
AUTH_DB=~/netdrive/omp/auth.db
// after: local disk
AUTH_DB=~/.omp/auth.db

Defensive patterns

Strategy: retry

Validate before calling

// pre-check: can we open the DB and is it on a suitable filesystem?
const dbPath = getAuthDbPath();
const stat = await fs.stat(path.dirname(dbPath)).catch(() => null);
if (!stat) throw new Error(`auth dir missing: ${path.dirname(dbPath)}`);
if (isNetworkMount(path.dirname(dbPath))) logger.warn("auth DB on network mount; locks may time out");

Try / catch

try {
  const store = new SqliteAuthCredentialStore(dbPath);
} catch (err) {
  if (err instanceof AIError.ConfigurationError && err.message.includes("Failed to open auth database")) {
    logger.error("Auth DB locked or unreadable", { path: dbPath, cause: err.cause });
    // surface to user: close other sessions or relocate dbPath
  } else throw err;
}

Prevention

Keep the auth database on local disk, not NFS/SMB/synced folders
Avoid running many concurrent CLI sessions against the same auth DB
Monitor disk space and file permissions on the auth directory
Treat this error as environmental: check for lock holders before retrying

When it happens

Trigger: The auth DB file is locked for longer than the retry window — another omp process holds a write lock, an antivirus/indexer has the file open, the DB lives on a network filesystem, or the file is corrupt so SQLite fails immediately on every attempt.

Common situations: Running multiple CLI sessions/workers that refresh auth concurrently; the auth directory on NFS/SMB/Dropbox-synced storage; stale lock from a crashed process; disk-full or permission issues surfacing as repeated open failures.

Related errors

AI-assisted analysis of can1357/oh-my-pi@9690622007 (2026-08-31). Data as JSON: /api/errors/98ce044ad138bd43. Report an issue: GitHub.

Appendix: source

Thrown at packages/ai/src/auth/sqlite-credential-store.ts:557

				try {
					await fs.chmod(dbPath, 0o600);
				} catch {
					// Ignore chmod failures (e.g., Windows)
				}
				SqliteAuthCredentialStore.#ensureAuthCredentialRefreshLeasesTable(db);
				return new SqliteAuthCredentialStore(db);
			} catch (err) {
				db?.close();
				if (!isSqliteBusyError(err)) {
					throw err;
				}
				lastBusyError = err instanceof Error ? err : new Error(String(err));
				if (attempt < maxAttempts - 1) {
					await Bun.sleep(baseDelayMs * 2 ** attempt);
				}
			}
		}
		throw new AIError.ConfigurationError(
			`Failed to open auth database at '${dbPath}' after ${maxAttempts} attempts: ${lastBusyError?.message}`,
			{ cause: lastBusyError },
		);
	}

	static #ensureAuthCredentialRefreshLeasesTable(db: Database): void {
		db.run(`
			CREATE TABLE IF NOT EXISTS auth_credential_refresh_leases (
				credential_id INTEGER PRIMARY KEY,
				owner TEXT NOT NULL,
				expires_at_ms INTEGER NOT NULL,
				updated_at INTEGER NOT NULL
			);
			CREATE INDEX IF NOT EXISTS idx_auth_credential_refresh_leases_expires ON auth_credential_refresh_leases(expires_at_ms);
		`);
	}

	/**

View on GitHub (pinned to 9690622007)