การลองใหม่และกลไกสำรอง

In real-time systems, reliable communication is key. Network glitches, server restarts, or temporary overloads can cause your WebSocket connection to drop.

This lesson explores how to make your applications resilient. We'll cover automatic reconnection strategies (retries) and alternative communication methods (fallbacks) to ensure a smooth user experience even when things go wrong.

When a WebSocket connection closes unexpectedly, the client shouldn't just give up. Implementing automatic reconnection logic on the client side is crucial for maintaining real-time interactions.

• The client detects a disconnection.
• It waits for a short period.
• It attempts to re-establish the WebSocket connection.
• This process repeats until successful or a maximum number of attempts is reached.

Here's a simple Java example simulating connection attempts with a fixed delay. Notice how it waits before each retry.

Try running it to see the retry process:

Repeatedly trying to reconnect with a fixed delay can overwhelm a recovering server. Exponential backoff is a smarter strategy:

• Start with a small delay.
• Double the delay after each failed attempt.
• Cap the delay at a maximum to prevent excessively long waits.

This gives the server more time to recover and reduces network traffic during outages.

Let's enhance our retry logic with exponential backoff. See how the delay increases with each failed attempt, up to a maximum.

Run this code to observe the growing delays:

Even with exponential backoff, if many clients disconnect and try to reconnect at the exact same doubled intervals, they might still create a 'thundering herd' problem.

Jitter adds a small, random amount of time to each delay. This spreads out reconnection attempts, preventing simultaneous bursts of requests and further easing server load during recovery.

Sometimes, WebSockets aren't just temporarily down; they might be completely unavailable due to network restrictions (e.g., corporate firewalls, old proxies) or server misconfiguration.

In such cases, a fallback mechanism provides an alternative communication channel. Common fallbacks include:

• Long Polling: Client repeatedly makes HTTP requests, server holds connection open until new data is available or timeout.
• Server-Sent Events (SSE): Server pushes data over a single, long-lived HTTP connection.

A robust client will first attempt to establish a WebSocket connection. If this consistently fails after a certain number of retries (and backoff), it can switch to a fallback method.

The logic typically looks like this:

• Try WebSocket connection.
• If WebSocket fails after N attempts, try Long Polling.
• If Long Polling also fails, consider showing an 'offline' message or degraded experience.

Libraries like SockJS automatically handle these fallbacks, simplifying client development.

For fallbacks to work, the server must also support the alternative communication protocols. For example, a Spring application configured for WebSockets often also provides HTTP endpoints for long polling or SSE.

Spring's STOMP over WebSocket support (using WebSocketMessageBrokerConfigurer) can automatically provide HTTP fallback options (like SockJS) if configured correctly, abstracting much of this complexity.

Consider a scenario where hundreds of clients disconnect simultaneously from a WebSocket server due to a brief network outage. The server quickly recovers.

Which of the following strategies, when combined, would best help these clients reconnect without overwhelming the recovering server and ensuring continued service?

Congratulations! You've learned how to build more reliable real-time applications.

We covered:

• The importance of automatic reconnection for clients.
• Implementing exponential backoff to manage retry delays gracefully.
• Adding jitter to prevent simultaneous reconnection storms.
• Using fallback mechanisms like long polling or SSE when WebSockets are not viable.

These techniques are essential for creating resilient and user-friendly real-time systems.