withPausedRequestHandling<T> method
Stops accepting new requests on the user-facing servers (the API
server and, if enabled, the web server), waits for in-flight requests
to settle, runs action, then resumes request handling.
Used during runtime operations that require the pod to be quiescent -
e.g. applying migrations through
InsightsEndpoint.applyMigrations - to provide the same safety
guarantees as a pod restart.
The operator-facing Insights server is not paused, so the call that
invoked action can complete and return its result.
Implementation
Future<T> withPausedRequestHandling<T>(
Future<T> Function() action,
) async {
await server.shutdown();
await _webServer?.stop();
final T result;
try {
result = await action();
} catch (_) {
// The action's error is the one to report.
await _resumeRequestHandling();
rethrow;
}
if (!await _resumeRequestHandling()) {
throw StateError(
'Failed to resume the Serverpod servers, see logs for details.',
);
}
return result;
}