pkg/metrics/providers: add AWS CloudWatch metrics provider

This commit is contained in:
mathetake
2020-03-02 21:58:58 +09:00
parent 62f2851dfd
commit 4ff28d7bd5
8 changed files with 310 additions and 7 deletions
+112
View File
@@ -0,0 +1,112 @@
package providers
import (
"encoding/json"
"fmt"
"net/http"
"strings"
"time"
"github.com/aws/aws-sdk-go/aws"
"github.com/aws/aws-sdk-go/aws/awserr"
"github.com/aws/aws-sdk-go/aws/session"
"github.com/aws/aws-sdk-go/service/cloudwatch"
flaggerv1 "github.com/weaveworks/flagger/pkg/apis/flagger/v1beta1"
)
const (
cloudWatchMaxRetries = 3
cloudWatchStartDeltaMultiplierOnMetricInterval = 10
)
type CloudWatchProvider struct {
client cloudWatchClient
startDelta time.Duration
}
// for the testing purpose
type cloudWatchClient interface {
GetMetricData(input *cloudwatch.GetMetricDataInput) (*cloudwatch.GetMetricDataOutput, error)
}
// NewCloudWatchProvider takes a metricInterval, a provider spec and the credentials map, and
// returns a cloudWatchProvider ready to execute queries against the AWS CloudWatch metrics
func NewCloudWatchProvider(metricInterval string, provider flaggerv1.MetricTemplateProvider) (*CloudWatchProvider, error) {
region := strings.TrimLeft(provider.Address, "monitoring.")
region = strings.TrimRight(region, ".amazonaws.com")
sess, err := session.NewSession(
aws.NewConfig().
WithRegion(region).
WithMaxRetries(cloudWatchMaxRetries).
WithEndpoint(provider.Address),
)
md, err := time.ParseDuration(metricInterval)
if err != nil {
return nil, fmt.Errorf("error parsing metric interval: %s", err.Error())
}
return &CloudWatchProvider{
client: cloudwatch.New(sess),
startDelta: cloudWatchStartDeltaMultiplierOnMetricInterval * md,
}, err
}
// RunQuery executes the aws cloud watch metrics query against GetMetricsData endpoint
// and returns the the first result as float64
func (p *CloudWatchProvider) RunQuery(query string) (float64, error) {
var cq []*cloudwatch.MetricDataQuery
if err := json.Unmarshal([]byte(query), &cq); err != nil {
return 0, fmt.Errorf("error unmarshaling query: %s", err.Error())
}
end := time.Now()
start := end.Add(-p.startDelta)
res, err := p.client.GetMetricData(&cloudwatch.GetMetricDataInput{
EndTime: aws.Time(end),
MaxDatapoints: aws.Int64(20),
StartTime: aws.Time(start),
MetricDataQueries: cq,
})
if err != nil {
return 0, fmt.Errorf("error requesting cloudwatch: %s", err.Error())
}
mr := res.MetricDataResults
if len(mr) < 1 {
return 0, fmt.Errorf("no values found in response: %s", res.String())
}
vs := res.MetricDataResults[0].Values
if len(vs) < 1 {
return 0, fmt.Errorf("no values found in response: %s", res.String())
}
return aws.Float64Value(vs[0]), nil
}
// IsOnline calls GetMetricsData endpoint with the empty query
// and returns an error if the returned status code is NOT http.StatusBadRequests.
// For example, if the flagger does not have permission to perform `cloudwatch:GetMetricsData`,
// the returned status code would be http.StatusForbidden
func (p *CloudWatchProvider) IsOnline() (bool, error) {
_, err := p.client.GetMetricData(&cloudwatch.GetMetricDataInput{
EndTime: aws.Time(time.Time{}),
MetricDataQueries: []*cloudwatch.MetricDataQuery{},
StartTime: aws.Time(time.Time{}),
})
if err == nil {
return true, nil
}
ae, ok := err.(awserr.RequestFailure)
if !ok {
return false, fmt.Errorf("unexpected error: %v", err)
} else if ae.StatusCode() != http.StatusBadRequest {
return false, fmt.Errorf("unexpected status code: %v", ae)
}
return true, nil
}
+177
View File
@@ -0,0 +1,177 @@
package providers
import (
"net/http"
"strings"
"testing"
"time"
"github.com/aws/aws-sdk-go/aws"
"github.com/aws/aws-sdk-go/aws/awserr"
"github.com/aws/aws-sdk-go/service/cloudwatch"
flaggerv1 "github.com/weaveworks/flagger/pkg/apis/flagger/v1beta1"
)
type cloudWatchClientMock struct {
o *cloudwatch.GetMetricDataOutput
err error
}
func (c cloudWatchClientMock) GetMetricData(_ *cloudwatch.GetMetricDataInput) (*cloudwatch.GetMetricDataOutput, error) {
return c.o, c.err
}
func TestNewCloudWatchProvider(t *testing.T) {
p, err := NewCloudWatchProvider(
"5m",
flaggerv1.MetricTemplateProvider{
Address: "monitoring.ap-northeast-1.amazonaws.com",
})
if err != nil {
t.Fatal(err)
}
if exp := 5 * 60 * time.Second * cloudWatchStartDeltaMultiplierOnMetricInterval; p.startDelta != exp {
t.Fatalf("expected %d but got %d", exp, p.startDelta)
}
}
func TestCloudWatchProvider_IsOnline(t *testing.T) {
t.Run("forbidden", func(t *testing.T) {
p := CloudWatchProvider{client: cloudWatchClientMock{
o: nil,
err: awserr.NewRequestFailure(nil, http.StatusForbidden, "request-id"),
}}
actual, err := p.IsOnline()
if err == nil {
t.Error("error expected")
}
if actual {
t.Error("false expected")
}
})
t.Run("ok", func(t *testing.T) {
// no error
p := CloudWatchProvider{client: cloudWatchClientMock{}}
actual, err := p.IsOnline()
if err != nil {
t.Errorf("no error expected but got %v", err)
}
if !actual {
t.Error("true expected")
}
// with error but bad request
p = CloudWatchProvider{client: cloudWatchClientMock{
err: awserr.NewRequestFailure(nil, http.StatusBadRequest, "request-id"),
}}
if err != nil {
t.Errorf("no error expected but got %v", err)
}
if !actual {
t.Error("true expected")
}
})
}
func TestCloudWatchProvider_RunQuery(t *testing.T) {
// ref: https://aws.amazon.com/premiumsupport/knowledge-center/cloudwatch-getmetricdata-api/
query := `
[
{
"Id": "e1",
"Expression": "m1 / m2",
"Label": "ErrorRate"
},
{
"Id": "m1",
"MetricStat": {
"Metric": {
"Namespace": "MyApplication",
"MetricName": "Errors",
"Dimensions": [
{
"Name": "FunctionName",
"Value": "MyFunc"
}
]
},
"Period": 300,
"Stat": "Sum",
"Unit": "Count"
},
"ReturnData": false
},
{
"Id": "m2",
"MetricStat": {
"Metric": {
"Namespace": "MyApplication",
"MetricName": "Invocations",
"Dimensions": [
{
"Name": "FunctionName",
"Value": "MyFunc"
}
]
},
"Period": 300,
"Stat": "Sum",
"Unit": "Count"
},
"ReturnData": false
}
]`
t.Run("ok", func(t *testing.T) {
var exp float64 = 100
p := CloudWatchProvider{client: cloudWatchClientMock{
o: &cloudwatch.GetMetricDataOutput{
MetricDataResults: []*cloudwatch.MetricDataResult{
{Values: []*float64{aws.Float64(exp)}},
},
},
}}
actual, err := p.RunQuery(query)
if err != nil {
t.Fatalf("no error expected but got %v", err)
}
if actual != exp {
t.Errorf("expected %f but got %f", exp, actual)
}
})
t.Run("no values", func(t *testing.T) {
assert := func(err error) {
if err == nil {
t.Fatal("error expected")
}
if !strings.Contains(err.Error(), "no values") {
t.Fatalf("no values expected in %v", err)
}
}
p := CloudWatchProvider{client: cloudWatchClientMock{
o: &cloudwatch.GetMetricDataOutput{
MetricDataResults: []*cloudwatch.MetricDataResult{
{Values: []*float64{}},
},
},
}}
_, err := p.RunQuery(query)
assert(err)
p = CloudWatchProvider{client: cloudWatchClientMock{
o: &cloudwatch.GetMetricDataOutput{}}}
_, err = p.RunQuery(query)
assert(err)
})
}
+5 -4
View File
@@ -11,12 +11,13 @@ func (factory Factory) Provider(
provider flaggerv1.MetricTemplateProvider,
credentials map[string][]byte,
) (Interface, error) {
switch {
case provider.Type == "prometheus":
switch provider.Type {
case "prometheus":
return NewPrometheusProvider(provider, credentials)
case provider.Type == "datadog":
case "datadog":
return NewDatadogProvider(metricInterval, provider, credentials)
case "cloudwatch":
return NewCloudWatchProvider(metricInterval, provider)
default:
return NewPrometheusProvider(provider, credentials)
}