feat: bound connect attempts with a 30 s default connectionTimeout (#2320)

connectionTimeout defaulted to 0, which InspectorClient.connect() read as
"no bound at all" — so out of the box nothing the Inspector owns bounded a
connect attempt, and the only thing ending a hung handshake was the SDK's
60 s per-request timeout on initialize, reporting "Request timed out".

- Add DEFAULT_CONNECTION_TIMEOUT_MS (30000) to core/mcp/types.ts. An absent
  value resolves to it on the client, on disk read, in the web form draft,
  and in the CLI's header-only settings shell. An explicit 0 keeps its
  documented meaning (no timeout; --connect-timeout 0) and is now persisted
  so the opt-out round-trips; the default is the omit-sentinel, as for
  taskTtl.
- Key the post-timeout transport teardown to the timer actually firing, not
  to connectionTimeout > 0 — otherwise a non-zero default would tear down
  every failed connect, including a recoverable 401 whose transport the
  auth recovery holds open.
- The timeout message names the bound and where to raise it: "Connection
  timed out after N ms. To accommodate a slower server, you may increase
  the timeout value in Server Settings."
- The CLI's UNREACHABLE classifier now matches that message.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01M4h8Xa6Ed9zwtv5ojX35MT
Signed-off-by: cliffhall <cliff@futurescale.com>
This commit is contained in:
cliffhall 2026-09-13 20:54:23 -04:00
commit e641edeeab

View file

@ -67,6 +67,7 @@ import type {
CreateTransportOptions,
ServerType,
} from "./types.js";
import { DEFAULT_CONNECTION_TIMEOUT_MS } from "./types.js";
import {
MessageTrackingTransport,
type MessageTrackingCallbacks,
@ -293,6 +294,20 @@ interface ReceiverTaskRecord {
*/
const MAX_URL_ELICITATION_RETRIES = 5;
/**
* The message `connect()` rejects with when the connect-time timeout fires.
* Names the bound that fired and, since the default is now a real value that
* a slow server on a slow link can legitimately hit, says where to raise it
* (#2320). All three clients surface this text as-is: the web client's
* "Failed to connect" toast, the TUI status line, and the CLI error envelope.
*/
export function connectionTimeoutMessage(timeoutMs: number): string {
return (
`Connection timed out after ${timeoutMs} ms. To accommodate a slower ` +
`server, you may increase the timeout value in Server Settings.`
);
}
/**
* Error used to reject a pending sampling/elicitation request when the tool
* call driving its MRTR round is aborted (e.g. the user hits Cancel). Its
@ -2220,11 +2235,28 @@ export class InspectorClient extends InspectorClientEventTarget {
this.registerPeerRequestHandlers();
this.registerPeerNotificationHandlers();
// Optional connect-time timeout from per-server settings. The MCP SDK
// has no connect-time timeout option, so we wrap the handshake in a
// Connect-time timeout from per-server settings, defaulting to
// `DEFAULT_CONNECTION_TIMEOUT_MS` when the settings carry none. The MCP
// SDK has no connect-time timeout option, so we wrap the handshake in a
// Promise.race. On timeout, tear the transport down so the next
// connect() starts clean and the upstream socket isn't left hanging.
const connectTimeoutMs = this.serverSettings?.connectionTimeout ?? 0;
//
// The default is ours on purpose (#2320): without it the only bound on
// a connect attempt was the SDK's per-request timeout on `initialize`,
// which covers nothing outside that request (`transport.start()` for
// one), fires after 60 s, and reports `Request timed out` — a sentence
// about a JSON-RPC request, not a connection. An explicit `0` is still
// "no timeout": that is what the CLI's `--connect-timeout 0` documents
// and what a user who cleared the field asked for.
const connectTimeoutMs =
this.serverSettings?.connectionTimeout ?? DEFAULT_CONNECTION_TIMEOUT_MS;
// Set only by the timer below, so the teardown in the catch runs for a
// timeout and nothing else. Gating it on `connectTimeoutMs > 0` instead
// would tear down *every* failed connect once the default is non-zero —
// including a recoverable 401, whose transport the auth recovery needs
// held open and whose status must stay at "connecting" (see the outer
// catch and `transportHasAuthProvider`).
let connectTimedOut = false;
// Unwrap here — the earliest point — so an auth error the SDK's
// era-negotiation probe buried in its cause chain is surfaced before
// anything downstream inspects it: `withDirectAuthRecovery` (whose
@ -2268,15 +2300,10 @@ export class InspectorClient extends InspectorClientEventTarget {
connectPromise.catch(() => {});
let timer: ReturnType<typeof setTimeout> | undefined;
const timeoutPromise = new Promise<never>((_, reject) => {
timer = setTimeout(
() =>
reject(
new Error(
`Connection timed out after ${connectTimeoutMs} ms`,
),
),
connectTimeoutMs,
);
timer = setTimeout(() => {
connectTimedOut = true;
reject(new Error(connectionTimeoutMessage(connectTimeoutMs)));
}, connectTimeoutMs);
});
try {
await Promise.race([connectPromise, timeoutPromise]);
@ -2291,7 +2318,7 @@ export class InspectorClient extends InspectorClientEventTarget {
try {
await this.invokeMcpClient(runConnect);
} catch (err) {
if (connectTimeoutMs > 0) {
if (connectTimedOut) {
await this.disconnect().catch(() => {});
}
throw err;