Prevent Service Failures: Question Third Parties, Guard Users, Perfect Your Code

For every programmer, failures loom like a sword of Damocles, and avoiding them is a constant pursuit. Most services depend on third‑party components, internal business logic, and user interactions, any of which can become a failure source.

1. Question Third Parties

1.1 Provide fallback and degradation plans

If a third‑party service goes down, your own service should have a degradation strategy, such as using cached hot items or maintaining snapshot indexes to continue operating with slightly stale data.

1.2 Apply fast‑failure principle with timeouts

Set reasonable timeout thresholds (e.g., 200 ms) for external calls; otherwise, slow third‑party responses can exhaust thread pools and cripple your service.

1.3 Choose retry mechanisms wisely

Only retry when business logic permits; indiscriminate retries can amplify load and worsen failures.

2. Guard Users

2.1 Design robust APIs (RPC/RESTful) to prevent misuse

Expose the minimal necessary interfaces.

Avoid requiring callers to make many repeated calls; provide batch methods when appropriate.

Enforce length limits and clearly document them.

Prefer strong typing over generic strings.

Return meaningful exceptions instead of swallowing errors.

2.2 Implement traffic control per service

Use rate‑limiting or circuit‑breaker patterns (like a fuse) to protect against sudden traffic spikes that could crash the system.

2.3 Strengthen your own service

2.3.1 Single‑responsibility principle

Apply SRP across requirements, architecture, and code to keep services modular and fault‑isolated.

2.3.2 Control resource usage

CPU

Optimize algorithms and avoid unnecessary locks.

Use thread pools to limit concurrency.

Fine‑tune JVM parameters.

Memory

Set JVM memory limits and pre‑size collections.

Use object pools and avoid unbounded caches.

Compress cached data and consider off‑heap storage.

Network

Batch remote calls and limit data transferred.

Support field‑level selection to reduce payload size.

Disk

Log only essential information, monitor log size, and rotate or purge logs regularly.

2.3.3 Avoid single points of failure

Deploy services across multiple zones, use horizontal scaling, and apply sharding or clustering techniques for stateful components.

Conclusion

To prevent failures, remember the three‑step mantra: “Question third parties, guard users, and perfect your own code.”

public List<Integer> test() {
    try {
        ...
    } catch (Exception e) {
        return Collections.emptyList();
    }
}