nexus

Troubleshooting

Diagnose connection, service, callback, and resource-lifecycle problems.

A Connection Never Completes

Check whether you supplied a target. A targetless connect waits passively for exactly one matching ready session and has no default timeout. It does not start the peer context or dial it. Supply a finite timeout while diagnosing the wait.

Verify both contexts exist, both runtimes are configured, the receiving endpoint is listening, and the target and where match the intended peer. For startup connectTo, the receiver must listen before the one-shot dial. Check adapter logs for transport or authentication failures.

See Establishing connections.

More Than One Connection Matches

E_SERVICE_AMBIGUOUS is also the current code for ambiguous passive connection acquisition. Narrow where, supply an exact target, or use connectMulticast when the operation intentionally involves multiple peers.

Connected, But The Service Is Unavailable

get checks the current catalog, not future availability. Verify the provider is registered on the intended runtime with the same Token ID. For live registration, retry safeGet after a business-ready signal. Publish the provider before updating the ready identity marker.

See Providing services and Connection lifecycle.

A Method Or Callback Does Not Execute

Consume its result. Nexus calls are lazy: service.run() followed by ignoring the return value, or void service.run(), does not send the operation. Use await, .catch(reportFailure), or a Promise helper.

For local event emitters that ignore returned values, wrap remote callbacks in a listener that consumes and observes their results. See Passing callbacks.

An Old Proxy Still Fails After Reconnection

Old proxies and refs remain bound to the terminated session. Get a fresh proxy from the new Connection. A retained ConnectionCollection also retains its old members; reacquire the collection for a new snapshot.

Release Did Not Stop A Subscription

Release gives up the remote reference; it does not invoke business cleanup. Call the subscription’s stop() or equivalent before releasing it. The provider also needs a business-owned shutdown path for abandoned work or disconnection. See Remote resources.

Only Some Sources Fail

A collection’s get reports one Result per connection. Inspect both acquisition and call errors with their source connection. A missing service is not a missing collection member. Decide whether to retry one source, take another snapshot, or display partial results.

See Multiple contexts.

A Timed-Out Operation Still Changed Remote State

The timeout ends local waiting; it is not a remote rollback or cancellation. If the operation may be retried, design an application-level idempotency or status protocol. See Errors and timeouts.

An Example Uses Removed APIs

These docs describe the Core 2.0 alpha contract under development. Old examples using create, select, default targets, provider waits, or remote assignment need migration. Follow the migration guide and check which package version the application uses.

Edit this page on GitHub