Skip to content

Commit

Permalink
Fix kafka plugin and rename to kafka_consumer
Browse files Browse the repository at this point in the history
fixes #371
  • Loading branch information
sparrc committed Nov 18, 2015
1 parent f85bc6e commit a3ed5e5
Show file tree
Hide file tree
Showing 10 changed files with 451 additions and 431 deletions.
2 changes: 1 addition & 1 deletion README.md
Original file line number Diff line number Diff line change
Expand Up @@ -164,7 +164,6 @@ Telegraf currently has support for collecting metrics from:
* haproxy
* httpjson (generic JSON-emitting http service plugin)
* jolokia (remote JMX with JSON over HTTP)
* kafka_consumer
* leofs
* lustre2
* memcached
Expand Down Expand Up @@ -197,6 +196,7 @@ Telegraf currently has support for collecting metrics from:
Telegraf can collect metrics via the following services:

* statsd
* kafka_consumer

We'll be adding support for many more over the coming months. Read on if you
want to add support for another service or third-party API.
Expand Down
44 changes: 19 additions & 25 deletions config_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@ import (

"github.com/influxdb/telegraf/plugins"
"github.com/influxdb/telegraf/plugins/exec"
"github.com/influxdb/telegraf/plugins/kafka_consumer"
"github.com/influxdb/telegraf/plugins/memcached"
"github.com/influxdb/telegraf/plugins/procstat"
"github.com/naoina/toml"
"github.com/naoina/toml/ast"
Expand Down Expand Up @@ -205,17 +205,14 @@ func TestConfig_parsePlugin(t *testing.T) {
pluginConfigurationFieldsSet: make(map[string][]string),
}

subtbl := tbl.Fields["kafka"].(*ast.Table)
err = c.parsePlugin("kafka", subtbl)
subtbl := tbl.Fields["memcached"].(*ast.Table)
err = c.parsePlugin("memcached", subtbl)

kafka := plugins.Plugins["kafka"]().(*kafka_consumer.Kafka)
kafka.ConsumerGroupName = "telegraf_metrics_consumers"
kafka.Topic = "topic_with_metrics"
kafka.ZookeeperPeers = []string{"test.example.com:2181"}
kafka.BatchSize = 1000
memcached := plugins.Plugins["memcached"]().(*memcached.Memcached)
memcached.Servers = []string{"localhost"}

kConfig := &ConfiguredPlugin{
Name: "kafka",
mConfig := &ConfiguredPlugin{
Name: "memcached",
Drop: []string{"other", "stuff"},
Pass: []string{"some", "strings"},
TagDrop: []TagFilter{
Expand All @@ -233,10 +230,10 @@ func TestConfig_parsePlugin(t *testing.T) {
Interval: 5 * time.Second,
}

assert.Equal(t, kafka, c.plugins["kafka"],
"Testdata did not produce a correct kafka struct.")
assert.Equal(t, kConfig, c.pluginConfigurations["kafka"],
"Testdata did not produce correct kafka metadata.")
assert.Equal(t, memcached, c.plugins["memcached"],
"Testdata did not produce a correct memcached struct.")
assert.Equal(t, mConfig, c.pluginConfigurations["memcached"],
"Testdata did not produce correct memcached metadata.")
}

func TestConfig_LoadDirectory(t *testing.T) {
Expand All @@ -249,14 +246,11 @@ func TestConfig_LoadDirectory(t *testing.T) {
t.Error(err)
}

kafka := plugins.Plugins["kafka"]().(*kafka_consumer.Kafka)
kafka.ConsumerGroupName = "telegraf_metrics_consumers"
kafka.Topic = "topic_with_metrics"
kafka.ZookeeperPeers = []string{"test.example.com:2181"}
kafka.BatchSize = 10000
memcached := plugins.Plugins["memcached"]().(*memcached.Memcached)
memcached.Servers = []string{"192.168.1.1"}

kConfig := &ConfiguredPlugin{
Name: "kafka",
mConfig := &ConfiguredPlugin{
Name: "memcached",
Drop: []string{"other", "stuff"},
Pass: []string{"some", "strings"},
TagDrop: []TagFilter{
Expand Down Expand Up @@ -296,10 +290,10 @@ func TestConfig_LoadDirectory(t *testing.T) {

pConfig := &ConfiguredPlugin{Name: "procstat"}

assert.Equal(t, kafka, c.plugins["kafka"],
"Merged Testdata did not produce a correct kafka struct.")
assert.Equal(t, kConfig, c.pluginConfigurations["kafka"],
"Merged Testdata did not produce correct kafka metadata.")
assert.Equal(t, memcached, c.plugins["memcached"],
"Merged Testdata did not produce a correct memcached struct.")
assert.Equal(t, mConfig, c.pluginConfigurations["memcached"],
"Merged Testdata did not produce correct memcached metadata.")

assert.Equal(t, ex, c.plugins["exec"],
"Merged Testdata did not produce a correct exec struct.")
Expand Down
5 changes: 3 additions & 2 deletions outputs/influxdb/influxdb.go
Original file line number Diff line number Diff line change
Expand Up @@ -29,8 +29,9 @@ type InfluxDB struct {
}

var sampleConfig = `
# The full HTTP or UDP endpoint URL for your InfluxDB instance
# Multiple urls can be specified for InfluxDB cluster support.
# The full HTTP or UDP endpoint URL for your InfluxDB instance.
# Multiple urls can be specified but it is assumed that they are part of the same
# cluster, this means that only ONE of the urls will be written to each interval.
# urls = ["udp://localhost:8089"] # UDP endpoint example
urls = ["http://localhost:8086"] # required
# The target database for metrics (telegraf will create it if not exists)
Expand Down
218 changes: 108 additions & 110 deletions plugins/kafka_consumer/kafka_consumer.go
Original file line number Diff line number Diff line change
@@ -1,164 +1,162 @@
package kafka_consumer

import (
"os"
"os/signal"
"time"
"log"
"strings"
"sync"

"github.com/Shopify/sarama"
"github.com/influxdb/influxdb/models"
"github.com/influxdb/telegraf/plugins"

"github.com/Shopify/sarama"
"github.com/wvanbergen/kafka/consumergroup"
)

type Kafka struct {
ConsumerGroupName string
Topic string
ZookeeperPeers []string
Consumer *consumergroup.ConsumerGroup
BatchSize int
ConsumerGroup string
Topics []string
ZookeeperPeers []string
Consumer *consumergroup.ConsumerGroup
PointBuffer int
Offset string

sync.Mutex

// channel for all incoming kafka messages
in <-chan *sarama.ConsumerMessage
// channel for all kafka consumer errors
errs <-chan *sarama.ConsumerError
// channel for all incoming parsed kafka points
pointChan chan models.Point
done chan struct{}

// doNotCommitMsgs tells the parser not to call CommitUpTo on the consumer
// this is mostly for test purposes, but there may be a use-case for it later.
doNotCommitMsgs bool
}

var sampleConfig = `
# topic to consume
topic = "topic_with_metrics"
# the name of the consumer group
consumerGroupName = "telegraf_metrics_consumers"
# topic(s) to consume
topics = ["telegraf"]
# an array of Zookeeper connection strings
zookeeperPeers = ["localhost:2181"]
# Batch size of points sent to InfluxDB
batchSize = 1000
zookeeper_peers = ["localhost:2181"]
# the name of the consumer group
consumer_group = "telegraf_metrics_consumers"
# Maximum number of points to buffer between collection intervals
point_buffer = 100000
# Offset (must be either "oldest" or "newest")
offset = "oldest"
`

func (k *Kafka) SampleConfig() string {
return sampleConfig
}

func (k *Kafka) Description() string {
return "read metrics from a Kafka topic"
}

type Metric struct {
Measurement string `json:"measurement"`
Values map[string]interface{} `json:"values"`
Tags map[string]string `json:"tags"`
Time time.Time `json:"time"`
return "Read line-protocol metrics from Kafka topic(s)"
}

func (k *Kafka) Gather(acc plugins.Accumulator) error {
func (k *Kafka) Start() error {
k.Lock()
defer k.Unlock()
var consumerErr error
metricQueue := make(chan []byte, 200)

if k.Consumer == nil {
config := consumergroup.NewConfig()
switch strings.ToLower(k.Offset) {
case "oldest", "":
config.Offsets.Initial = sarama.OffsetOldest
case "newest":
config.Offsets.Initial = sarama.OffsetNewest
default:
log.Printf("WARNING: Kafka consumer invalid offset '%s', using 'oldest'\n",
k.Offset)
config.Offsets.Initial = sarama.OffsetOldest
}

if k.Consumer == nil || k.Consumer.Closed() {
k.Consumer, consumerErr = consumergroup.JoinConsumerGroup(
k.ConsumerGroupName,
[]string{k.Topic},
k.ConsumerGroup,
k.Topics,
k.ZookeeperPeers,
nil,
config,
)

if consumerErr != nil {
return consumerErr
}

c := make(chan os.Signal, 1)
halt := make(chan bool, 1)
signal.Notify(c, os.Interrupt)
go func() {
<-c
halt <- true
emitMetrics(k, acc, metricQueue)
k.Consumer.Close()
}()

go readFromKafka(k.Consumer.Messages(),
metricQueue,
k.BatchSize,
k.Consumer.CommitUpto,
halt)
// Setup message and error channels
k.in = k.Consumer.Messages()
k.errs = k.Consumer.Errors()
}

return emitMetrics(k, acc, metricQueue)
}
k.done = make(chan struct{})
if k.PointBuffer == 0 {
k.PointBuffer = 100000
}
k.pointChan = make(chan models.Point, k.PointBuffer)

func emitMetrics(k *Kafka, acc plugins.Accumulator, metricConsumer <-chan []byte) error {
timeout := time.After(1 * time.Second)
// Start the kafka message reader
go k.parser()
log.Printf("Started the kafka consumer service, peers: %v, topics: %v\n",
k.ZookeeperPeers, k.Topics)
return nil
}

// parser() reads all incoming messages from the consumer, and parses them into
// influxdb metric points.
func (k *Kafka) parser() {
for {
select {
case batch := <-metricConsumer:
var points []models.Point
var err error
if points, err = models.ParsePoints(batch); err != nil {
return err
case <-k.done:
return
case err := <-k.errs:
log.Printf("Kafka Consumer Error: %s\n", err.Error())
case msg := <-k.in:
points, err := models.ParsePoints(msg.Value)
if err != nil {
log.Printf("Could not parse kafka message: %s, error: %s",
string(msg.Value), err.Error())
}

for _, point := range points {
acc.AddFields(point.Name(), point.Fields(), point.Tags(), point.Time())
select {
case k.pointChan <- point:
continue
default:
log.Printf("Kafka Consumer buffer is full, dropping a point." +
" You may want to increase the point_buffer setting")
}
}

if !k.doNotCommitMsgs {
k.Consumer.CommitUpto(msg)
}
case <-timeout:
return nil
}
}
}

const millisecond = 1000000 * time.Nanosecond

type ack func(*sarama.ConsumerMessage) error

func readFromKafka(
kafkaMsgs <-chan *sarama.ConsumerMessage,
metricProducer chan<- []byte,
maxBatchSize int,
ackMsg ack,
halt <-chan bool,
) {
batch := make([]byte, 0)
currentBatchSize := 0
timeout := time.After(500 * millisecond)
var msg *sarama.ConsumerMessage

for {
select {
case msg = <-kafkaMsgs:
if currentBatchSize != 0 {
batch = append(batch, '\n')
}

batch = append(batch, msg.Value...)
currentBatchSize++

if currentBatchSize == maxBatchSize {
metricProducer <- batch
currentBatchSize = 0
batch = make([]byte, 0)
ackMsg(msg)
}
case <-timeout:
if currentBatchSize != 0 {
metricProducer <- batch
currentBatchSize = 0
batch = make([]byte, 0)
ackMsg(msg)
}

timeout = time.After(500 * millisecond)
case <-halt:
if currentBatchSize != 0 {
metricProducer <- batch
ackMsg(msg)
}
func (k *Kafka) Stop() {
k.Lock()
defer k.Unlock()
close(k.done)
if err := k.Consumer.Close(); err != nil {
log.Printf("Error closing kafka consumer: %s\n", err.Error())
}
}

return
}
func (k *Kafka) Gather(acc plugins.Accumulator) error {
k.Lock()
defer k.Unlock()
npoints := len(k.pointChan)
for i := 0; i < npoints; i++ {
point := <-k.pointChan
acc.AddFields(point.Name(), point.Fields(), point.Tags(), point.Time())
}
return nil
}

func init() {
plugins.Add("kafka", func() plugins.Plugin {
plugins.Add("kafka_consumer", func() plugins.Plugin {
return &Kafka{}
})
}
Loading

0 comments on commit a3ed5e5

Please sign in to comment.