Merge pull request #89 from weaveworks/dynamo-monitoring

Metrics for Dynamo capacity

Author: tomwilkie

chunk/dynamo_client.go

@@ -0,0 +1,132 @@
+package chunk
+
+import (
+	"net/url"
+	"strings"
+	"sync"
+	"time"
+
+	"github.com/aws/aws-sdk-go/aws/session"
+	"github.com/aws/aws-sdk-go/service/dynamodb"
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/common/log"
+)
+
+const (
+	readLabel  = "read"
+	writeLabel = "write"
+)
+
+type dynamoWatcher struct {
+	accountMaxCapacity *prometheus.GaugeVec
+	tableCapacity      *prometheus.GaugeVec
+
+	dynamoDB  *dynamodb.DynamoDB
+	tableName string
+
+	updateInterval time.Duration
+	quit           chan struct{}
+	wait           sync.WaitGroup
+}
+
+// Watcher watches something and reports to Prometheus.
+type Watcher interface {
+	Stop()
+	prometheus.Collector
+}
+
+// WatchDynamo watches Dynamo and reports on resource limits.
+func WatchDynamo(dynamoDBURL string, interval time.Duration) (Watcher, error) {
+	url, err := url.Parse(dynamoDBURL)
+	if err != nil {
+		return nil, err
+	}
+	dynamoDBConfig, err := awsConfigFromURL(url)
+	if err != nil {
+		return nil, err
+	}
+	client := dynamodb.New(session.New(dynamoDBConfig))
+
+	tableName := strings.TrimPrefix(url.Path, "/")
+	w := &dynamoWatcher{
+		accountMaxCapacity: prometheus.NewGaugeVec(prometheus.GaugeOpts{
+			Namespace: "cortex",
+			Name:      "dynamo_account_max_capacity_units",
+			Help:      "Account-wide DynamoDB capacity, measured in DynamoDB capacity units.",
+		}, []string{"op"}),
+		tableCapacity: prometheus.NewGaugeVec(prometheus.GaugeOpts{
+			Namespace: "cortex",
+			Name:      "dynamo_table_capacity_units",
+			Help:      "Per-table DynamoDB capacity, measured in DynamoDB capacity units.",
+		}, []string{"op", "table"}),
+		dynamoDB:       client,
+		tableName:      tableName,
+		updateInterval: interval,
+		quit:           make(chan struct{}),
+	}
+	go w.updateLoop()
+	return w, nil
+}
+
+// Stop stops the dynamo watcher.
+func (w *dynamoWatcher) Stop() {
+	close(w.quit)
+	w.wait.Wait()
+}
+
+func (w *dynamoWatcher) updateLoop() {
+	defer w.wait.Done()
+	ticker := time.NewTicker(w.updateInterval)
+	for {
+		select {
+		case <-ticker.C:
+			log.Debugf("Updating limits from dynamo")
+			err := w.updateAccountLimits()
+			if err != nil {
+				// TODO: Back off if err is throttling related.
+				log.Warnf("Could not fetch account limits from dynamo: %v", err)
+			}
+			err = w.updateTableLimits()
+			if err != nil {
+				log.Warnf("Could not fetch table limits from dynamo: %v", err)
+			}
+		case <-w.quit:
+			ticker.Stop()
+		}
+	}
+}
+
+func (w *dynamoWatcher) updateAccountLimits() error {
+	limits, err := w.dynamoDB.DescribeLimits(&dynamodb.DescribeLimitsInput{})
+	if err != nil {
+		return err
+	}
+	w.accountMaxCapacity.WithLabelValues(readLabel).Set(float64(*limits.AccountMaxReadCapacityUnits))
+	w.accountMaxCapacity.WithLabelValues(writeLabel).Set(float64(*limits.AccountMaxWriteCapacityUnits))
+	return nil
+}
+
+func (w *dynamoWatcher) updateTableLimits() error {
+	output, err := w.dynamoDB.DescribeTable(&dynamodb.DescribeTableInput{
+		TableName: &w.tableName,
+	})
+	if err != nil {
+		return err
+	}
+	throughput := output.Table.ProvisionedThroughput
+	w.tableCapacity.WithLabelValues(readLabel, w.tableName).Set(float64(*throughput.ReadCapacityUnits))
+	w.tableCapacity.WithLabelValues(writeLabel, w.tableName).Set(float64(*throughput.WriteCapacityUnits))
+	return nil
+}
+
+// Describe implements prometheus.Collector.
+func (w *dynamoWatcher) Describe(ch chan<- *prometheus.Desc) {
+	w.accountMaxCapacity.Describe(ch)
+	w.tableCapacity.Describe(ch)
+}
+
+// Collect implements prometheus.Collector.
+func (w *dynamoWatcher) Collect(ch chan<- prometheus.Metric) {
+	w.accountMaxCapacity.Collect(ch)
+	w.tableCapacity.Collect(ch)
+}

cmd/cortex/main.go

@@ -62,6 +62,7 @@ type cfg struct {
 	s3URL                string
 	dynamodbURL          string
 	dynamodbCreateTables bool
+	dynamodbPollInterval time.Duration
 	memcachedHostname    string
 	memcachedTimeout     time.Duration
 	memcachedExpiration  time.Duration
@@ -83,6 +84,7 @@ func main() {
 	flag.StringVar(&cfg.s3URL, "s3.url", "localhost:4569", "S3 endpoint URL.")
 	flag.StringVar(&cfg.dynamodbURL, "dynamodb.url", "localhost:8000", "DynamoDB endpoint URL.")
 	flag.BoolVar(&cfg.dynamodbCreateTables, "dynamodb.create-tables", false, "Create required DynamoDB tables on startup.")
+	flag.DurationVar(&cfg.dynamodbPollInterval, "dynamodb.poll-interval", 2*time.Minute, "How frequently to poll DynamoDB to learn our capacity.")
 	flag.StringVar(&cfg.memcachedHostname, "memcached.hostname", "", "Hostname for memcached service to use when caching chunks. If empty, no memcached will be used.")
 	flag.DurationVar(&cfg.memcachedTimeout, "memcached.timeout", 100*time.Millisecond, "Maximum time to wait before giving up on memcached requests.")
 	flag.DurationVar(&cfg.memcachedExpiration, "memcached.expiration", 0, "How long chunks stay in the memcache.")
@@ -102,6 +104,15 @@ func main() {
 	if err != nil {
 		log.Fatalf("Error initializing chunk store: %v", err)
 	}
+	if cfg.dynamodbPollInterval < 1*time.Minute {
+		log.Warnf("Polling DynamoDB more than once a minute. Likely to get throttled: %v", cfg.dynamodbPollInterval)
+	}
+	resourceWatcher, err := chunk.WatchDynamo(cfg.dynamodbURL, cfg.dynamodbPollInterval)
+	if err != nil {
+		log.Fatalf("Error initializing DynamoDB watcher: %v", err)
+	}
+	defer resourceWatcher.Stop()
+	prometheus.MustRegister(resourceWatcher)
 
 	consul, err := ring.NewConsulClient(cfg.consulHost)
 	if err != nil {
@@ -234,7 +245,7 @@ func setupQuerier(
 
 	engine := promql.NewEngine(queryable, nil)
 
-	api := v1.NewAPI(engine, querier.DummyStorage{queryable})
+	api := v1.NewAPI(engine, querier.DummyStorage{Queryable: queryable})
 	router := route.New(func(r *http.Request) (context.Context, error) {
 		userID := r.Header.Get(userIDHeaderName)
 		if r.Method != "OPTIONS" && userID == "" {