vrischmann / go-metrics-influxdb Goto Github PK

InfluxDB will now panic on sending metrics because it defaults tags to nil and then tries to assign an entry in the nil map.

Default to nil is influxdb.go#L30 and assignment to nil map: influxdb.go#L152.

Do you have any example queries or a grafana dashboard that worked for you?

When creating tags for the purpose of bucketing histogram, meter, and timer metrics, the library improperly modifies the global tag set:

https://github.com/vrischmann/go-metrics-influxdb/blob/master/influxdb.go#L151
https://github.com/vrischmann/go-metrics-influxdb/blob/master/influxdb.go#L173
https://github.com/vrischmann/go-metrics-influxdb/blob/master/influxdb.go#L207

I'm sure the intent here was to copy the existing tag set, but maps are a reference type. The result is that the bucket tag that is added for histogram, meter, and timer metrics is also applied to counter and gauge metrics.

The correct way to handle this is to explicitly copy the existing map:

these_tags := map[string]string{}
for tk, tv := range r.tags {
    these_tags[tk] = tv
}

I'll open a PR with the fix.

@vrischmann If you don't mind, I can create a PR to add go modules to this project. Which version should I give it? v0.1.0?

Error:
"partial write: field type conflict: input field "total.request.meter" on measurement "table" is type integer, already exists as type float

https://github.com/vrischmann/go-metrics-influxdb/blob/master/influxdb.go#L164-L175

in InfluxDB ， the client use Default config ， the client don't support timeout

Hi,
is support for InfluxDB 2.x planned?

Authentication has changed so older clients can't connect anymore.

There is an official influxdb client for 1.8+ and 2.x:
https://github.com/influxdata/influxdb-client-go

Thanks
Jochen

InfluxDB function should expose errors

this repo still being maintained?

hello,i find if i get server info.after long time.linux show too many ESTABLISHED.

package main

import (
	"time"
	"github.com/rcrowley/go-metrics"
	"github.com/vrischmann/go-metrics-influxdb"
)


func main() {
	r := metrics.NewRegistry()
	metrics.RegisterDebugGCStats(r)
	metrics.RegisterRuntimeMemStats(r)

	go metrics.CaptureDebugGCStats(r, time.Second *5)
	go metrics.CaptureRuntimeMemStats(r, time.Second *5)

	for {
		go influxdb.InfluxDB(
			r,                             // metrics registry
			time.Second * 5,                 // how often
			"***", // InfluxDB url
			"***",                  // InfluxDB database name
			"",                            // InfluxDB user
			"",                            // InfluxDB password
		)
		time.Sleep(time.Second * 3)
	}
}

Thanks for your hard work on this lib.

I just want to ask if you have plans to support the new v2 client as the old one is deprecated: https://github.com/influxdata/influxdb/tree/master/client

Also the new client supports udp connections, but apparently there is no way to set this up with go-metrics-influxdb.

Hi,

It looks like when reporting histogram values - all percentiles will be set to same value.

Here's how go-metrics logger fetches the percentiles:
ps := h.Percentiles([]float64{0.5, 0.75, 0.95, 0.99, 0.999})

Here's influxdb.go:
"p999": m.Percentile(99.9),
"p9999": m.Percentile(99.99),

From traffic capture. Note the pNN values:

count=481999i,m1=999.9992707686223,m15=998.8201711297105,m5=999.589420684841,max=1181835i,mean=127654.4280155642,meanrate=999.9936221298891,min=81662i,p50=1181835,p75=1181835,p95=1181835,p99=1181835,p999=1181835,p9999=1181835,stddev=76492.17168321775,variance=5851052328.814858 1445863754555784700

when i use metrics.Histogram

总数一直相加，count 一直直线增长。 应该调用metrics.clear() 清除一下吧？

我可否提一个pr，合并一下？

according to https://community.influxdata.com/t/how-to-solve-type-conflicts/8473, one field under the same measurement (regardless of the different buckets) needs to have same field type.

otherwise, we might have errors as shown below

2020/01/21 12:28:10 unable to send metrics to InfluxDB. err={"error":"partial write: field type conflict: input field \"process-time.histogram\" on measurement \"item-pipeline\" is type float, already exists as type integer dropped=10"}

The fix would be to unify the same data type for each field. For example, for Histogram, we could do the following:

fields := map[string]interface{}{
	"count": float64(ms.Count()),
	"m1":    ms.Rate1(),
	"m5":    ms.Rate5(),
	"m15":   ms.Rate15(),
	"mean":  ms.RateMean(),
}

Please approve the problem and solution and I can create the MR for it.