apisrv

Documentation

Overview

Copyright 2018 Google LLC

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

https://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

Index

• Variables
• type FrontendAPI
  • func New(cfg *viper.Viper, pool *redis.Pool) *FrontendAPI
  • func (s *FrontendAPI) Open() error

Variables

var (
	// API instrumentation
	FeGrpcRequests          = stats.Int64("frontendapi/requests_total", "Number of requests to the gRPC Frontend API endpoints", "1")
	FeGrpcStreamedResponses = stats.Int64("frontendapi/streamed_responses_total", "Number of responses streamed back from the gRPC Frontend API endpoints", "1")
	FeGrpcErrors            = stats.Int64("frontendapi/errors_total", "Number of errors generated by the gRPC Frontend API endpoints", "1")
	FeGrpcLatencySecs       = stats.Float64("frontendapi/latency_seconds", "Latency in seconds of the gRPC Frontend API endpoints", "1")

	// Logging instrumentation
	// There's no need to record this measurement directly if you use
	// the logrus hook provided in metrics/helper.go after instantiating the
	// logrus instance in your application code.
	// https://godoc.org/github.com/sirupsen/logrus#LevelHooks
	FeLogLines = stats.Int64("frontendapi/logs_total", "Number of Frontend API lines logged", "1")

	// Failure instrumentation
	FeFailures = stats.Int64("frontendapi/failures_total", "Number of Frontend API failures", "1")
)

OpenCensus Measures. These are exported as metrics to your monitoring system https://godoc.org/go.opencensus.io/stats

When making opencensus stats, the 'name' param, with forward slashes changed to underscores, is appended to the 'namespace' value passed to the prometheus exporter to become the Prometheus metric name. You can also look into having Prometheus rewrite your metric names on scrape.

 For example:
  - defining the promethus export namespace "open_match" when instanciating the exporter:
			pe, err := promethus.NewExporter(promethus.Options{Namespace: "open_match"})
  - and naming the request counter "frontend/requests_total":
			MGrpcRequests := stats.Int64("frontendapi/requests_total", ...
  - results in the prometheus metric name:
			open_match_frontendapi_requests_total
  - [note] when using opencensus views to aggregate the metrics into
    distribution buckets and such, multiple metrics
    will be generated with appended types ("<metric>_bucket",
    "<metric>_count", "<metric>_sum", for example)

In addition, OpenCensus stats propogated to Prometheus have the following auto-populated labels pulled from kubernetes, which we should avoid to prevent overloading and having to use the HonorLabels param in Prometheus.

• Information about the k8s pod being monitored: "pod" (name of the monitored k8s pod) "namespace" (k8s namespace of the monitored pod)
• Information about how promethus is gathering the metrics: "instance" (IP and port number being scraped by prometheus) "job" (name of the k8s service being scraped by prometheus) "endpoint" (name of the k8s port in the k8s service being scraped by prometheus)

var (
	// KeyMethod is used to tag a measure with the currently running API method.
	KeyMethod, _   = tag.NewKey("method")
	KeySeverity, _ = tag.NewKey("severity")
)

var (
	FeLatencyView = &view.View{
		Name:        "frontend/latency",
		Measure:     FeGrpcLatencySecs,
		Description: "The distribution of frontend latencies",
		Aggregation: latencyDistribution,
		TagKeys:     []tag.Key{KeyMethod},
	}

	FeRequestCountView = &view.View{
		Name:        "frontend/grpc/requests",
		Measure:     FeGrpcRequests,
		Description: "The number of successful frontend gRPC requests",
		Aggregation: view.Count(),
		TagKeys:     []tag.Key{KeyMethod},
	}

	FeStreamedResponseCountView = &view.View{
		Name:        "frontend/grpc/streamed_responses",
		Measure:     FeGrpcRequests,
		Description: "The number of successful streamed gRPC responses",
		Aggregation: view.Count(),
		TagKeys:     []tag.Key{KeyMethod},
	}

	FeErrorCountView = &view.View{
		Name:        "frontend/grpc/errors",
		Measure:     FeGrpcErrors,
		Description: "The number of gRPC errors",
		Aggregation: view.Count(),
		TagKeys:     []tag.Key{KeyMethod},
	}

	FeLogCountView = &view.View{
		Name:        "log_lines/total",
		Measure:     FeLogLines,
		Description: "The number of lines logged",
		Aggregation: view.Count(),
		TagKeys:     []tag.Key{KeySeverity},
	}

	FeFailureCountView = &view.View{
		Name:        "failures",
		Measure:     FeFailures,
		Description: "The number of failures",
		Aggregation: view.Count(),
	}
)

Package metrics provides some convience views. You need to register the views for the data to actually be collected. Note: The OpenCensus View 'Description' is exported to Prometheus as the HELP string. Note: If you get a "Failed to export to Prometheus: inconsistent label cardinality" error, chances are you forgot to set the tags specified in the view for a given measure when you tried to do a stats.Record()

var DefaultFrontendAPIViews = []*view.View{
	FeLatencyView,
	FeRequestCountView,
	FeStreamedResponseCountView,
	FeErrorCountView,
	FeLogCountView,
	FeFailureCountView,
}

DefaultFrontendAPIViews are the default frontend API OpenCensus measure views.

Types

type FrontendAPI

type FrontendAPI struct {
	// contains filtered or unexported fields
}

FrontendAPI implements frontend.ApiServer, the server generated by compiling the protobuf, by fulfilling the frontend.APIClient interface.

func New

func New(cfg *viper.Viper, pool *redis.Pool) *FrontendAPI

New returns an instantiated srvice

func (*FrontendAPI) Open

func (s *FrontendAPI) Open() error

Open starts the api grpc service listening on the configured port.