Spring Boot中使用HikariCP连接池
2019-02-02
Spring Boot 2.x中使用HikariCP作为默认的数据连接池。
HikariCP使用Javassist字节码操作库来实现动态代理，优化并精简了字节码，同时内部使用com.zaxxer.hikari.util.FastList代替ArrayList、使用了更好的并发集合类com.zaxxer.hikari.util.ConcurrentBag，“号称”是目前最快的数据库连接池。
基本使用 #
在Spring Boot 2.x中使用HikariCP十分简单，只需引入依赖implementation 'org.springframework.boot:spring-boot-starter-jdbc'：
1pluginManagement {
2 repositories {
3 gradlePluginPortal()
4 }
5}
6rootProject.name = 'datasource-config'
7
8plugins {
9 id 'org.springframework.boot' version '2.1.3.RELEASE'
10 id 'java'
11}
12
13apply plugin: 'io.spring.dependency-management'
14
15group = 'spring-boot-guides'
16version = '0.0.1-SNAPSHOT'
17sourceCompatibility = '1.8'
18
19repositories {
20 mavenCentral()
21}
22
23dependencies {
24 implementation 'org.springframework.boot:spring-boot-starter-jdbc'
25 runtimeOnly 'com.h2database:h2'
26 testImplementation 'org.springframework.boot:spring-boot-starter-test'
27}
配置文件如下：
1spring:
2 datasource:
3 url: jdbc:h2:mem:demodb
4 username: sa
5 password:
6 hikari: # https://github.com/brettwooldridge/HikariCP (uses milliseconds for all time values)
7 maximumPoolSize: 10
8 minimumIdle: 2
9 idleTimeout: 600000
10 connectionTimeout: 30000
11 maxLifetime: 1800000
关于连接池的具体配置参数详见HikariCP。
示例代码如下：
1package springbootguides.datasourceconfig;
2
3import org.springframework.beans.factory.annotation.Autowired;
4import org.springframework.boot.CommandLineRunner;
5import org.springframework.boot.SpringApplication;
6import org.springframework.boot.autoconfigure.SpringBootApplication;
7
8import javax.sql.DataSource;
9import java.sql.Connection;
10
11@SpringBootApplication
12public class DatasourceConfigApplication implements CommandLineRunner {
13
14 @Autowired
15 private DataSource datasource;
16
17 @Override
18 public void run(String... args) throws Exception {
19 try(Connection conn = datasource.getConnection()) {
20 System.out.println(conn);
21 }
22 }
23
24 public static void main(String[] args) {
25 SpringApplication.run(DatasourceConfigApplication.class, args);
26 }
27
28}
实现原理 #
Spring Boot使用如下方式整合HikariCP：入口是org.springframework.boot.autoconfigure.jdbc.DataSourceAutoConfiguration，通过org.springframework.boot.autoconfigure.jdbc.DataSourceConfiguration.Hikari中的@Bean方式创建com.zaxxer.hikari.HikariDataSource：
1/**
2 * Hikari DataSource configuration.
3 */
4 @ConditionalOnClass(HikariDataSource.class)
5 @ConditionalOnMissingBean(DataSource.class)
6 @ConditionalOnProperty(name = "spring.datasource.type", havingValue = "com.zaxxer.hikari.HikariDataSource", matchIfMissing = true)
7 static class Hikari {
8
9 @Bean
10 @ConfigurationProperties(prefix = "spring.datasource.hikari")
11 public HikariDataSource dataSource(DataSourceProperties properties) {
12 HikariDataSource dataSource = createDataSource(properties,
13 HikariDataSource.class);
14 if (StringUtils.hasText(properties.getName())) {
15 dataSource.setPoolName(properties.getName());
16 }
17 return dataSource;
18 }
19
20 }
@ConfigurationProperties(prefix = "spring.datasource.hikari")会自动把spring.datasource.hikari.*相关的连接池配置信息注入到创建的HikariDataSource实例中。
HikariCP的监控和遥测 #
因为在我们的微服务体系中使用的监控系统是Prometheus，这里以Prometheus为例。
注意spring boot 2.0对spring boot 1.x的metrics进行了重构，不再向后兼容，主要是在spring-boot-acutator中使用了micrometer，支持了更多的监控系统：Atlas、Datadog、Ganglia、Graphite、Influx、JMX、NewRelic、Prometheus、SignalFx、StatsD、Wavefront。Spring boot 2.0的metrics对比spring boot 1.x除了引入micrometer外，更大的体现是支持了tag，这也说明Prometheus、Influx等支持Tag的时序监控数据模型的监控系统已经成为主流。
在前面示例中的build.gradle中加入如下依赖：
1implementation 'org.springframework.boot:spring-boot-starter-web'
2implementation 'org.springframework.boot:spring-boot-starter-actuator'
3implementation 'io.micrometer:micrometer-registry-prometheus'
配置文件applycation.yaml中加入对actuator的配置：
1management:
2 endpoints:
3 web:
4 exposure:
5 include: "health,info,prometheus"
6 server:
7 port: 8079
8 servlet:
9 context-path: /
注意这里引入了web和actuator依赖，通过配置management.server.port指定actuator的web端点为8089端口，通过
management.endpoints.include对外开放/actuator/prometheus，在引入io.micrometer:micrometer-registry-prometheus依赖之后，端点/actuator/prometheus当即生效。
1curl http://localhost:8079/actuator/prometheus | grep hikari
2# TYPE hikaricp_connections_acquire_seconds summary
3hikaricp_connections_acquire_seconds_count{pool="HikariPool-1",} 3.0
4hikaricp_connections_acquire_seconds_sum{pool="HikariPool-1",} 0.001230082
5# HELP hikaricp_connections_acquire_seconds_max Connection acquire time
6# TYPE hikaricp_connections_acquire_seconds_max gauge
7hikaricp_connections_acquire_seconds_max{pool="HikariPool-1",} 0.0
8# HELP hikaricp_connections_min Min connections
9# TYPE hikaricp_connections_min gauge
10hikaricp_connections_min{pool="HikariPool-1",} 2.0
11# TYPE hikaricp_connections_timeout_total counter
12hikaricp_connections_timeout_total{pool="HikariPool-1",} 0.0
13# HELP hikaricp_connections_pending Pending threads
14# TYPE hikaricp_connections_pending gauge
15hikaricp_connections_pending{pool="HikariPool-1",} 0.0
16# HELP hikaricp_connections_usage_seconds Connection usage time
17# TYPE hikaricp_connections_usage_seconds summary
18hikaricp_connections_usage_seconds_count{pool="HikariPool-1",} 3.0
19hikaricp_connections_usage_seconds_sum{pool="HikariPool-1",} 0.06
20# HELP hikaricp_connections_usage_seconds_max Connection usage time
21# TYPE hikaricp_connections_usage_seconds_max gauge
22hikaricp_connections_usage_seconds_max{pool="HikariPool-1",} 0.0
23# HELP hikaricp_connections_max Max connections
24# TYPE hikaricp_connections_max gauge
25hikaricp_connections_max{pool="HikariPool-1",} 10.0
26# HELP hikaricp_connections Total connections
27# TYPE hikaricp_connections gauge
28hikaricp_connections{pool="HikariPool-1",} 2.0
29# HELP hikaricp_connections_creation_seconds_max Connection creation time
30# TYPE hikaricp_connections_creation_seconds_max gauge
31hikaricp_connections_creation_seconds_max{pool="HikariPool-1",} 0.0
32# HELP hikaricp_connections_creation_seconds Connection creation time
33# TYPE hikaricp_connections_creation_seconds summary
34hikaricp_connections_creation_seconds_count{pool="HikariPool-1",} 1.0
35hikaricp_connections_creation_seconds_sum{pool="HikariPool-1",} 0.001
36# HELP hikaricp_connections_idle Idle connections
37# TYPE hikaricp_connections_idle gauge
38hikaricp_connections_idle{pool="HikariPool-1",} 2.0
39# HELP hikaricp_connections_active Active connections
40# TYPE hikaricp_connections_active gauge
41hikaricp_connections_active{pool="HikariPool-1",} 0.0