mssql uses tarn for pooling. When a pending acquire is aborted (e.g. pool teardown), tarn rejects with a plain Error('aborted'). Timeouts use TimeoutError, which is easy to detect with instanceof. Aborts are not, so callers must match message or stack traces to tell “shutting down / aborted wait” from real DB or network failures.
Upstream, a tarn change was proposed (TarnError, aborts using it, optionally TimeoutError extends it). Tarn looks unmaintained, so we should not assume that lands.
Request
Please either:
- In
mssql: rethrow/wrap these pool errors in some own error type, so we can catch and classify them without parsing the error message (message === 'aborted'); or
- Longer term: consider another maintained pool package (or internal pooling) that exposes explicit error types. We need a stable way to handle teardown without treating aborted acquires like ordinary query failures.
mssqluses tarn for pooling. When a pending acquire is aborted (e.g. pool teardown), tarn rejects with a plainError('aborted'). Timeouts useTimeoutError, which is easy to detect withinstanceof. Aborts are not, so callers must matchmessageor stack traces to tell “shutting down / aborted wait” from real DB or network failures.Upstream, a tarn change was proposed (
TarnError, aborts using it, optionallyTimeoutErrorextends it). Tarn looks unmaintained, so we should not assume that lands.Request
Please either:
mssql: rethrow/wrap these pool errors in some own error type, so we can catch and classify them without parsing the error message (message === 'aborted'); or