Spring Boot 3中如何向Prometheus Pushgateway推送指标？

核心问题分析 #

你当前的误区是混淆了Micrometer的MeterRegistry和Prometheus simpleclient的CollectorRegistry，同时没必要手动初始化PushGateway——Spring Boot已经提供了自动配置能力，能帮你完成指标的收集和推送。

1. 确认依赖配置 #

你添加的依赖是正确的，micrometer-registry-prometheus-simpleclient已包含simpleclient基础依赖，保留simpleclient_pushgateway也无影响：

<dependency>
    <groupId>io.micrometer</groupId>
    <artifactId>micrometer-registry-prometheus-simpleclient</artifactId>
    <version>${micrometer-registry-prometheus.version}</version>
</dependency>
<dependency>
    <groupId>io.prometheus</groupId>
    <artifactId>simpleclient_pushgateway</artifactId>
    <version>0.16.0</version>
</dependency>

2. 完善application.properties配置 #

仅开启推送不够，需补充PushGateway地址、任务标识等关键配置：

# 开启PushGateway推送
management.prometheus.metrics.export.pushgateway.enabled=true
# PushGateway服务地址
management.prometheus.metrics.export.pushgateway.base-url=http://your-pushgateway-host:9091
# 推送的任务名称（Prometheus中作为job标签）
management.prometheus.metrics.export.pushgateway.job=order-service
# 定期推送间隔（可选，默认1分钟）
management.prometheus.metrics.export.pushgateway.push-interval=1m
# 推送时保留实例标签（可选）
management.prometheus.metrics.export.pushgateway.instance=${spring.application.name}:${server.port}

3. 正确创建并使用指标 #

不要用Prometheus simpleclient的Counter.builder，改用Micrometer原生API，配合Spring自动注入的MeterRegistry（实际是PrometheusMeterRegistry实现类）：

方式一：提前初始化Counter（推荐）

import io.micrometer.core.instrument.Counter;
import io.micrometer.core.instrument.MeterRegistry;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.stereotype.Component;

@Component
public class MetricService {

    private final Counter timeoutCounter;

    // 构造器注入MeterRegistry，提前初始化Counter
    @Autowired
    public MetricService(MeterRegistry meterRegistry) {
        this.timeoutCounter = Counter.builder("order_service.cronjob.clean_orphan_order_drafts.timeouts.total")
                .description("Represents how many times the CleanOrphanOrderDraftsJob cronjob failed with timeout")
                .register(meterRegistry);
    }

    public void incrementCleanOrphanOrderDraftsJobTimeoutMetric() {
        timeoutCounter.increment();
    }
}

方式二：快速递增（适合简单场景）

import io.micrometer.core.instrument.Counter;
import io.micrometer.core.instrument.MeterRegistry;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.stereotype.Component;

@Component
public class MetricService {

    @Autowired
    private MeterRegistry meterRegistry;

    public void incrementCleanOrphanOrderDraftsJobTimeoutMetric() {
        Counter.increment(meterRegistry, "order_service.cronjob.clean_orphan_order_drafts.timeouts.total");
    }
}

4. 自动推送 vs 手动推送 #

• 自动推送：配置完成后，Spring Boot的PushGatewayManager会按设置的push-interval自动推送所有Micrometer收集的指标，无需额外代码。
• 手动推送：如果是定时任务执行后立即推送（如一次性任务），可注入PushGateway手动触发：

import io.prometheus.client.exporter.PushGateway;
import io.micrometer.prometheus.simpleclient.PrometheusMeterRegistry;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.stereotype.Component;

@Component
public class ManualPushService {

    @Autowired
    private PushGateway pushGateway;

    @Autowired
    private PrometheusMeterRegistry prometheusMeterRegistry;

    public void pushMetricsManually() throws Exception {
        // 推送当前所有指标到PushGateway
        pushGateway.push(prometheusMeterRegistry.getCollectorRegistry(), "order-service");
    }
}

内容的提问来源于stack exchange，提问作者Artem