Mastering Sentinel: From Setup to Advanced Flow Control in Spring Cloud

1. Introduction

This article is the fifth in the "Spring Cloud Advanced" series and introduces Alibaba's open‑source traffic‑guard Sentinel, a mature solution tested during ten years of Double‑11 traffic.

2. What is Sentinel?

Sentinel (literally "sentinel") monitors traffic, provides flow control, circuit breaking, and system load protection to keep services stable.

3. Features

Rich scenarios from Alibaba's Double‑11 traffic (spike control, message smoothing, cluster flow control, real‑time circuit breaking).

Real‑time monitoring of single‑machine and cluster metrics.

Integration with Spring Cloud, Dubbo, gRPC, Quarkus, and multi‑language support.

Extensible SPI for custom logic.

4. Version selection

When using the Spring Cloud Alibaba dependencies version 2.2.1.RELEASE, Sentinel 1.7.1 is recommended.

5. Installing the Sentinel dashboard

Download the appropriate sentinel-dashboard-1.7.1.jar and run it with:

java -Dserver.port=8080 -Dcsp.sentinel.dashboard.server=localhost:8080 -Dproject.name=sentinel-dashboard -jar sentinel-dashboard-1.7.1.jar

Default credentials are sentinel/sentinel.

6. Integrating a microservice with the dashboard

Register the service with Nacos, add the Sentinel starter dependency, and configure the dashboard address in application.yml:

spring:
  cloud:
    sentinel:
      transport:
        dashboard: localhost:8080

Define a test controller and verify that the service appears in the dashboard.

7. Flow control configuration

Sentinel supports QPS and concurrent‑thread limits. A FlowRule consists of resource, count, grade, limitApp, strategy, and controlBehavior. Three control behaviors are available: fast‑fail, warm‑up, and queue‑waiting.

7.1 Fast‑fail

Requests exceeding the threshold are rejected immediately.

7.2 Warm‑up

Gradually increases traffic after a cold start using a token‑bucket algorithm.

7.3 Queue‑waiting

Uses a leaky‑bucket algorithm to pace requests.

8. Degrade (circuit‑breaker) rules

Sentinel provides three degrade strategies: average response time, exception ratio, and exception count. Rules are defined with DegradeRule and can be visualised in the dashboard.

9. Hot‑parameter flow control

Hot‑parameter limits protect frequently accessed parameters (e.g., product IDs in a flash‑sale). Rules are defined with ParamFlowRule and support parameter‑specific thresholds and exception items.

10. System adaptive flow control

Sentinel can protect the whole system based on load, CPU usage, average RT, entry QPS, or concurrent threads.

11. Custom block messages

Use @SentinelResource with blockHandler (or blockHandlerClass) to return custom responses when a request is blocked.

12. Exception fallback

Define fallback (or defaultFallback) methods to handle business exceptions separately from block exceptions.

13. Black‑/white‑list control

Configure AuthorityRule to allow or deny requests from specific origins (e.g., IP addresses).

14. Rule persistence

Sentinel rules are stored in memory by default. Using the Push mode with Nacos, rules can be persisted and synchronised. Example application.yml configuration is provided.

15. Pushing dashboard changes to Nacos

Modify Sentinel Dashboard source to use Nacos datasource for both pull and push, enabling rule changes made in the UI to be written back to Nacos.

16. Cluster flow control

Sentinel supports embedded or standalone token servers for cluster‑wide flow control. Configure one instance as the token server and others as clients, then add cluster flow rules via the dashboard or Nacos.

17. Gateway flow control

Gateway integration follows the same principles and can be configured in the dashboard.