zsais / go-gin-prometheus Goto Github PK

View Code? Open in Web Editor NEW

439.0 2.0 131.0 39 KB

Gin Web Framework Prometheus metrics exporter

License: MIT License

Go 100.00%

go gin-gonic prometheus middleware

go-gin-prometheus's People

Contributors

Stargazers

Watchers

Forkers

gwik artushin alehano t3chguy pmoncadaisla depado yuankunzhang smikulcik natefinch mattel cjauvin vmware-archive droiplatform bdrich kadaan molenzhang kuiche1982 alex-grant-bose jim-lambert-bose manasi-bhandare-bose hellais banzaicloud etsangsplk mlvazqu2 nemo-g icook forging2012 0nese7en lscieux eldad87 godefault liaoxu josca eaglerayp zetaab codeplaytech pubgo aarushik93 blessmylovexy clannadxr lewisgj rrudduck ryannguyen89 stdioa dgsb taiyang-li rkuska huangguozhen hsh1237 tuan3w hzhaop easyops-cn tongxurt dtomasi shawnd genji1037 justxu 0x1a0b amperfii epguan likelihoood liminxufen lixianliang i0mm shivas blankidentifier ktakenaka subzerobo dxc520 akiolove zhangdekui vishnu333 jemuri luchenhua zoftdev vkstack big-wig davidzkeng euterper atij riseinrose even-april ljx0520 vothanhdien wangguanxiaoer sniper91 codysk virus20200317 zoelov hedemonde saberlee1987 wolf-nest xiong11feng wangganglab dmatosl column willthrom snowlyg rokt xuanyingcool

go-gin-prometheus's Issues

How to use custom metrics?

Hi, i read sourse code and could not understand how to use custom metrics? Does exist way do it, or i should rewrite this library?

Please add tag from the newest master branch

Hi,

I saw you fix some bug in master, please add a tag from the newest master.

We used v0.1.0 and met OOM issue, after we looked up the issues, we found you fix it in #35.

Thank you.

High cardinality caused by the url label

url label contains the url of every request that was ever made against gin webserver. This can cause very high memory usage for prometheus server. What's worse this is out of control for the operator of a service running go-gin-prometheus and someone sending a high number of requests with random url can easily cause prometheus server scraping that webserver to run out of memory. This is a common misuse of prometheus metrics and there's even a tool written to detect and mitigate the most serious cases of cardinality explosion - https://blog.freshtracks.io/bomb-squad-automatic-detection-and-suppression-of-prometheus-cardinality-explosions-62ca8e02fa32

This is very easy to reproduce with curl:

for i in {1..1000} ; do curl -so /dev/null "localhost:8080/xxx$i" ; done```

Resulting in:

[...]
gin_requests_total{code="404",handler="main.notFound",host="localhost:8080",method="GET",url="/xxx100"} 1
gin_requests_total{code="404",handler="main.notFound",host="localhost:8080",method="GET",url="/xxx10"} 1
gin_requests_total{code="404",handler="main.notFound",host="localhost:8080",method="GET",url="/xxx1"} 1
gin_requests_total{code="404",handler="main.notFound",host="localhost:8080",method="GET",url="/xxx11"} 1
gin_requests_total{code="404",handler="main.notFound",host="localhost:8080",method="GET",url="/xxx12"} 1
gin_requests_total{code="404",handler="main.notFound",host="localhost:8080",method="GET",url="/xxx13"} 1
[...]

This was already reported in #34 and partially fixed in #35 but those changes are not enough.
All metrics already have the handler label that points to the request handler method that the request was routed to, this is enough to identify requests for any debugging / performance tracking.

Please remove the url label from metrics.

Thread safe issue

The middleware pulls in this function: computeApproximateRequestSize from
https://github.com/DanielHeckrath/gin-prometheus

which inherits this bug:
DanielHeckrath/gin-prometheus#1

Type error on Use

./main.go:12:7: cannot use r (type *"gitlab.lojasrenner.com.br/bd-api-gateway-2/vendor/github.com/gin-gonic/gin".Engine) as type *"github.com/zsais/go-gin-prometheus/vendor/github.com/gin-gonic/gin".Engine in argument to p.Use

package main

import (
	"github.com/gin-gonic/gin"
	"github.com/zsais/go-gin-prometheus"
)

func main() {
	r := gin.New()

	p := ginprometheus.NewPrometheus("gin")
	p.Use(r)

	r.GET("/", func(c *gin.Context) {
		c.JSON(200, "Hello world!")
	})

	r.Run(":29090")
}

How to use my own custom metrics ?

How to use custom metrics ? i want to integrate with my personal apps

Total number of scrapes by HTTP status code.

Hi, i have a problem with the status code metrics, all the other metrics works except these:

promhttp_metric_handler_requests_total{code="200"} 0
promhttp_metric_handler_requests_total{code="500"} 0
promhttp_metric_handler_requests_total{code="503"} 0

Any idea?

Delete querystring in metrics labels.

This problem cause too many labels in metrics requests_total.

myname_requests_total{code="200",handler="myname",host="127.0.0.1:3000",method="GET",url="/handshake?token=1"} 1
myname_requests_total{code="200",handler="myname",host="127.0.0.1:3000",method="GET",url="/handshake?token=2"} 1
myname_requests_total{code="200",handler="myname",host="127.0.0.1:3000",method="GET",url="/handshake?token=3"} 1
myname_requests_total{code="200",handler="myname",host="127.0.0.1:3000",method="GET",url="/handshake?token=4"} 1
myname_requests_total{code="200",handler="myname",host="127.0.0.1:3000",method="GET",url="/handshake?token=5"} 1
myname_requests_total{code="200",handler="myname",host="127.0.0.1:3000",method="GET",url="/handshake?token=6"} 1

Support unix sockets

The SetListenAddress / SetListenAddressWithRouter is really nice, but unfortunately it only supports the "ADDRESS:PORT" scheme.
Do you can think of a good way to support also the other run functions like RunUnix?

Logrus import issue

It crashes complaining about
case-insensitive import collision: "github.com/Sirupsen/logrus" and "github.com/sirupsen/logrus"

requests_total metric not found

I am using go-gin-prometheus as a middleware with the default metrics and a couple of custom metrics that I registered but when I look at the metrics in my /metrics endpoint, I do not see a requests_total metric in there. Should I be implementing this like a custom metrics?

 func Use(engine *gin.Engine, port string) {
     prometheus.MustRegister(commonMetrics.ComponentProcessingTime)
     prometheus.MustRegister(commonMetrics.ComponentConsumerFailures)
     ginCol := ginprometheus.NewPrometheus("gin")
     ginCol.SetListenAddress(":" + port)
     ginCol.Use(engine)
}

Double Gzip of prometheus' metrics

Hi,
In case and you try to use this package along with an existing Gin router that have Gzip middleware, /metrics will get Gzipped twice and won't be readable.

For example:

import (
	"github.com/gin-gonic/gin"
	"github.com/zsais/go-gin-prometheus"
)

ginRouter := gin.New()
ginRouter.Use(gzip.Gzip(gzip.BestCompression))

// Add Prometheus
p := ginprometheus.NewPrometheus("gin")
p.Use(ginRouter)

// Add custom route
ginRouter.GET("/ping", func(c *gin.Context) {
		c.JSON(200, gin.H{
			"message": "pong",
		})
	})

// Run
ginRouter.Run(":8080")

Now browse:
http://localhost:8080/ping - {"message": "pong"}
http://localhost:8080/metrics - gibberish

Suggestion:
Please export hidden functionality to allow uses to have more control over this package.

please visualize it

hi, the metrics data can be get from http://host/metrics, but it's text, can you please visualize it?

pushIntervalSeconds refers to nanoseconds

Parameter in SetPushGateway

No longer relevant.

’requests_total‘ metirc oom issue

This metic default label ‘url’ will set url param,
if url param contain uuid or ramdon number.
label array give rise to oom.

Prometheus label naming

Hi peeps, according to Prometheus best practices docs, labels should have the http_ prefix upfront, is it possible to have it by any chance? Thank you

It would be nice to support multiple Prometheus registries

At the moment because the default registry is used this is not possible. For example if you have an application with two Gin servers the second gin-prometheus handler will fail with an error like:

time="2021-03-09T17:43:51Z" level=error msg="requests_total could not be registered in Prometheus" error="duplicate metrics collector registration attempted"
time="2021-03-09T17:43:51Z" level=error msg="request_duration_seconds could not be registered in Prometheus" error="duplicate metrics collector registration attempted"
time="2021-03-09T17:43:51Z" level=error msg="response_size_bytes could not be registered in Prometheus" error="duplicate metrics collector registration attempted"
time="2021-03-09T17:43:51Z" level=error msg="request_size_bytes could not be registered in Prometheus" error="duplicate metrics collector registration attempted"

Cheers!

Best practice for http and https together?

I have a server which listens for http (80) and https (443), with a gin instance for each. I'm trying to figure out how I should handle that setup with this package-- what is the best practice?

What I think I want is a single Prometheus /metrics endpoint on a separate port (let's say 3000) which exports metrics for both http and https endpoints, ideally as separate metrics sets, but merged would be OK. Is this possible? I couldn't find a way to get this to work, nor any examples which got to this level of detail.

The closest I got was something like this:

package main

import (
        "github.com/gin-gonic/gin"
        "github.com/zsais/go-gin-prometheus"
)

func main() {
        metrics := gin.New()
        r1 := gin.New()
        r2 := gin.New()

        p := ginprometheus.NewPrometheus("gin")
        p.SetListenAddressWithRouter(":3000", metrics)

        p.Use(r1)
        p.Use(r2)
        r1.GET("/", func(c *gin.Context) {
                c.JSON(200, "Hello world!")
        })
        r2.GET("/", func(c *gin.Context) {
                c.JSON(200, "Goodbye world!")
        })

        go r1.Run(":29090")
        go r2.Run(":29091")
        go metrics.Run(":3000")
}

Which panics as follows:

[GIN-debug] GET    /metrics                  --> github.com/zsais/go-gin-prometheus.prometheusHandler.func1 (3 handlers)
[GIN-debug] GET    /metrics                  --> github.com/zsais/go-gin-prometheus.prometheusHandler.func1 (3 handlers)
[GIN-debug] Listening and serving HTTP on :3000
panic: handlers are already registered for path '/metrics'

goroutine 1 [running]:
github.com/gin-gonic/gin.(*node).addRoute(0xc420070780, 0xc420024810, 0x8, 0xc42000d2c0, 0x3, 0x3)
        /home/dp/go/src/github.com/gin-gonic/gin/tree.go:235 +0x9c6
github.com/gin-gonic/gin.(*Engine).addRoute(0xc42022e480, 0xae34e2, 0x3, 0xc420024810, 0x8, 0xc42000d2c0, 0x3, 0x3)
        /home/dp/go/src/github.com/gin-gonic/gin/gin.go:252 +0x19e
github.com/gin-gonic/gin.(*RouterGroup).handle(0xc42022e480, 0xae34e2, 0x3, 0xae6ee0, 0x8, 0xc42000e188, 0x1, 0x1, 0xb68601, 0xc42000e188)
        /home/dp/go/src/github.com/gin-gonic/gin/routergroup.go:71 +0xe7
github.com/gin-gonic/gin.(*RouterGroup).GET(0xc42022e480, 0xae6ee0, 0x8, 0xc42000e188, 0x1, 0x1, 0x1, 0xb70f40)
        /home/dp/go/src/github.com/gin-gonic/gin/routergroup.go:99 +0x7c
github.com/zsais/go-gin-prometheus.(*Prometheus).setMetricsPath(0xc42023e180, 0xc42022e360)
        /home/dp/go/src/github.com/zsais/go-gin-prometheus/middleware.go:181 +0xab
github.com/zsais/go-gin-prometheus.(*Prometheus).Use(0xc42023e180, 0xc42022e360)
        /home/dp/go/src/github.com/zsais/go-gin-prometheus/middleware.go:339 +0x80
main.main()
        /home/dp/example1/main.go:16 +0xbf

Thanks for any ideas you might have.

Update to the Releases

I'm trying to use this package and I need the reqDur to be HistogramVec but in the current release(v0.1.0), it is a Summary. Can you fast track the next release?

My point being versions should be moving. The first and last version came two years back..

InternalMessageInfo

# gitlab.lojasrenner.com.br/bd-api-gateway-2/vendor/github.com/prometheus/client_model/go
vendor/github.com/prometheus/client_model/go/metrics.pb.go:96:31: undefined: proto.InternalMessageInfo
vendor/github.com/prometheus/client_model/go/metrics.pb.go:141:27: undefined: proto.InternalMessageInfo
vendor/github.com/prometheus/client_model/go/metrics.pb.go:179:29: undefined: proto.InternalMessageInfo
vendor/github.com/prometheus/client_model/go/metrics.pb.go:218:30: undefined: proto.InternalMessageInfo
vendor/github.com/prometheus/client_model/go/metrics.pb.go:265:29: undefined: proto.InternalMessageInfo
vendor/github.com/prometheus/client_model/go/metrics.pb.go:317:29: undefined: proto.InternalMessageInfo
vendor/github.com/prometheus/client_model/go/metrics.pb.go:357:31: undefined: proto.InternalMessageInfo
vendor/github.com/prometheus/client_model/go/metrics.pb.go:410:28: undefined: proto.InternalMessageInfo
vendor/github.com/prometheus/client_model/go/metrics.pb.go:461:28: undefined: proto.InternalMessageInfo
vendor/github.com/prometheus/client_model/go/metrics.pb.go:544:34: undefined: proto.InternalMessageInfo
vendor/github.com/prometheus/client_model/go/metrics.pb.go:544:34: too many errors

package main

import (
	"github.com/gin-contrib/gzip"
	"github.com/gin-gonic/gin"
	"github.com/zsais/go-gin-prometheus"
)

func main() {
	r := gin.New()


	r.Use(gzip.Gzip(gzip.DefaultCompression))

	p := ginprometheus.NewPrometheus("gin")
	p.Use(r)

	r.GET("/", func(c *gin.Context) {
		c.JSON(200, "Hello world!")
	})

	r.Run(":29090")
}

How to Increment Custom Metric

ok.. so im a bit stumped here.

i have the custom metrics implemented (so i can see them show up when i have my server running):

However, even with looking with the answer provided to nizarayari, i dont see how to apply the .Inc() method to the custom metric.

I have tried as well looking at the other example, however the prometheus.Register(numErrors) example never ends up showing up in the gin metrics list (as the test ones outlined above are).

any pointers would be wonderful!
D

zsais / go-gin-prometheus Goto Github PK

go-gin-prometheus's People

Contributors

Stargazers

Watchers

Forkers

go-gin-prometheus's Issues

Recommend Projects

Recommend Topics

Recommend Org

Jobs