Metric#
metric 使用比较流行的prometheus SDK接入
默认配置#
toml
env
[metric]
enable = true
provider = "prometheus"
endpoint = "/metrics"
[metric.api_stats]
enable = true
# 50线 90线 99线
request_summary = true
request_summary_name = "http_request_duration"
request_summary_objective = [0.5, 0.9, 0.99]
# 响应时长分布(单位毫秒)
request_histogram = false
request_histogram_name = "http_request_duration"
request_histogram_bucket = [100, 250, 500, 1000, 2500, 5000, 10000]
# 请求总数
request_total = true
request_total_name = "http_request_total"基本使用#
# HELP go_gc_duration_seconds A summary of the pause duration of garbage collection cycles.
# TYPE go_gc_duration_seconds summary
go_gc_duration_seconds{quantile="0"} 3.4167e-05
go_gc_duration_seconds{quantile="0.25"} 3.4167e-05
go_gc_duration_seconds{quantile="0.5"} 3.4167e-05
go_gc_duration_seconds{quantile="0.75"} 3.4167e-05
go_gc_duration_seconds{quantile="1"} 3.4167e-05
go_gc_duration_seconds_sum 3.4167e-05
go_gc_duration_seconds_count 1
...Gin框架#
// 非业务模块
_ "github.com/infraboard/mcube/v2/ioc/apps/metric/gin"$ mcube/examples/http_gin ‹master*› » go run main.go
...
2023-12-27T11:21:07+08:00 INFO metric/gin/metric.go:51 > Get the Metric using http://127.0.0.1:8080/metrics component:METRIC
...GoRestful框架#
// 非业务模块
_ "github.com/infraboard/mcube/v2/ioc/apps/metric/restful"$ mcube/examples/http_go_restful ‹master*› » go run main.go
...
2023-12-27T11:18:40+08:00 INFO metric/restful/metric.go:55 > Get the Metric using http://127.0.0.1:8080/metrics component:METRIC
...API Stats#
默认统计了 API响应时长的 50线 90线 99线
访问: http://127.0.0.1:8080/metrics/
# HELP http_request_duration Histogram of the duration of HTTP requests
# TYPE http_request_duration summary
http_request_duration{app="",method="GET",path="/metrics/",quantile="0.5"} 1.527666
http_request_duration{app="",method="GET",path="/metrics/",quantile="0.9"} 5.361042
http_request_duration{app="",method="GET",path="/metrics/",quantile="0.99"} 5.361042
http_request_duration_sum{app="",method="GET",path="/metrics/"} 7.5801240000000005
http_request_duration_count{app="",method="GET",path="/metrics/"} 3
# HELP http_request_total Total number of HTTP rquests
# TYPE http_request_total counter
http_request_total{app="",method="GET",path="/metrics/",status_code="200"} 3
自定义指标#
这里使用prometheus SDK提供的默认注册表, 因此只需要将编写好的自定义采集器 注册到 默认注册表即可
编写Collector#
package impl
import "github.com/prometheus/client_golang/prometheus"
func NewEventCollect() *EventCollect {
return &EventCollect{
errCountDesc: prometheus.NewDesc(
"save_event_error_count",
"事件入库失败个数统计",
[]string{},
prometheus.Labels{"service": "maudit"},
),
}
}
// 收集事件指标的采集器
type EventCollect struct {
errCountDesc *prometheus.Desc
// 需要自己根据实践情况来维护这个变量
errCount int
}
func (c *EventCollect) Inc() {
c.errCount++
}
// 指标元数据注册
func (c *EventCollect) Describe(ch chan<- *prometheus.Desc) {
ch <- c.errCountDesc
}
// 指标的值的采集
func (c *EventCollect) Collect(ch chan<- prometheus.Metric) {
ch <- prometheus.MustNewConstMetric(c.errCountDesc, prometheus.GaugeValue, float64(c.errCount))
}注册Collector#
在控制器初始化的时候完成注册:
// 控制器初始化
func (i *HelloServiceImpl) Init() error {
// 从Ioc总获取GORM DB对象, GORM相关配置已经托管给Ioc
// Ioc会负责GORM的配置读取和为你初始化DB对象实例,以及关闭
i.db = datasource.DB()
// 将采集器注册到默认注册表
i.colector = NewEventCollect()
prometheus.MustRegister(i.colector)
return nil
}自定义指标处理#
在业务控制器内，根据实际情况使用自定义指标纪录应用当前状态:
// 具体业务逻辑
func (i *HelloServiceImpl) Hello() string {
// 模拟存储失败报错, 然后调用采集器纪录状态
err := errors.New("save event error")
if err != nil {
i.colector.Inc()
}
return "hello world"
}自定义指标验证#
最后访问: http://127.0.0.1:8020/metrics
# HELP save_event_error_count 事件入库失败个数统计
# TYPE save_event_error_count gauge
save_event_error_count{service="maudit"} 0
目录