google / go-metrics-stackdriver Goto Github PK

Hi.
In some cases, there is a graceful shutdown logic of the application (such as in Cloud Run), and I need a way to gracefully shutdown the Sink as well, so pending metrics will get sent upon closure and not lost.

It turns out that a counter metrics use a gauge metric type as here

MetricKind: metricpb.MetricDescriptor_GAUGE,

This doesn't seem to be right. According to documentation of Aligner I am unable to calculate deltas if the metric is not defined as delta or cumulative. This is a big limitation. Should we consider switching to cumulative in case of counters?

https://pkg.go.dev/github.com/google/go-metrics-stackdriver#NewSink

currently only accepts a https://pkg.go.dev/cloud.google.com/go/monitoring/apiv3#MetricClient

which says:

Deprecated: Please use cloud.google.com/go/monitoring/apiv3/v2.

This was also mentioned #16 but would be a breaking change to the package.

gometrics supports gauge/counter/histograms with all the same "key".

example: https://github.com/hashicorp/vault/blob/35667f93a7afce915c4cd5fed99682918c46f60e/sdk/database/dbplugin/databasemiddleware.go#L122-L134

Proposal: append count or gauge to the metric key name (and leave histograms naked).

go-metrics has added a Shutdown() method to the interface ( https://pkg.go.dev/github.com/armon/go-metrics?utm_source=godoc#MetricSink )

func (s *Sink) Shutdown() { s.Close(context.Background()) }

This library used to work, but lately it started to produce weird and inaccurate values.
If I run the full example and check in stackdriver UI for foo metric: m.AddSample([]string{"foo"}, 100), the result in stackdriver is 75 for some reason. It happens on every sample in the basic example.

The expectations in stackdriver_test.go would be much more readable of a package like protocmp was used for ensuring the requests had the correct data.

Failed to write time series data: rpc error: code = InvalidArgument desc = One or more TimeSeries could not be written: Field timeSeries[8] had an invalid value: Duplicate TimeSeries encountered. Only one point can be written per TimeSeries per request.: timeSeries[8]

Related Issues:
census-ecosystem/opencensus-go-exporter-stackdriver#73
census-ecosystem/opencensus-go-exporter-stackdriver#120

We are facing this error when using the library:

2021/02/12 18:23:36 Failed to write time series data: rpc error: code = InvalidArgument desc = One or more TimeSeries could not be written: Field timeSeries[11] had an invalid value: Duplicate TimeSeries encountered. Only one point can be written per TimeSeries per request.: timeSeries[11];