jessedobbelaere / datadog-agent-burrow

Datadog check to pull metrics out of Burrow

Purpose

This datadog agent plugin will connect to instances of Burrow (https://github.com/linkedin/Burrow) and automatically get the offsets for every consumer and topic and push them into Datadog.

Metrics and Checks

This plugin will push the offsets for all topics (except the offsets_topic) and consumers for every kafka cluster it finds into Datadog as a metric.

The reported lags for each consumer group are also pushed up, along with the burrow consumer group status.

It will also push a check burrow.can_connect into Datadog to monitor the health of Burrow

Metrics

Topic Offsets

• kafka.topic.offsets - Per partition offsets of each topic
• kafka.topic.offsets.total - Total offset of each topic

Consumer Offsets

• kafka.consumer.offsets - Per partition offsets of each consumer groups
• kafka.consumer.offsets.total - Total offset of each consumer group

Lag

• kafka.consumer.totallag - Total lag of each consumer group
• kafka.consumer.partition_lag - Lag of each partition of a consumer group

The lag statuses are defined in burrow here: https://github.com/linkedin/Burrow/wiki/Consumer-Lag-Evaluation-Rules

Consumer group lag status:

• kafka.consumer.lag_status.unknown
• kafka.consumer.lag_status.ok
• kafka.consumer.lag_status.warn
• kafka.consumer.lag_status.err
• kafka.consumer.lag_status.stop
• kafka.consumer.lag_status.stall
• kafka.consumer.lag_status.rewind

Per Partition lag status of a consumer group:

• kafka.consumer.partition_lag_status.unknown
• kafka.consumer.partition_lag_status.ok
• kafka.consumer.partition_lag_status.warn
• kafka.consumer.partition_lag_status.err
• kafka.consumer.partition_lag_status.stop
• kafka.consumer.partition_lag_status.stall
• kafka.consumer.partition_lag_status.rewind

All metrics are tagged with cluster, topic, consumer and partition (where applicable) to make it easy to filter out in Datadog.

Development

1. Start zookeeper and kafka up before Burrow

   docker-compose up -d kafka zookeeper
   

2. Start up burrow

   docker-compose up burrow
   

3. Populate some data into the kafka topic. Replace the IP address with whatever IP your kafka docker container is running as.

   cat ./sample_messages | kafka-console-producer --broker-list 192.168.99.100:9092 --topic test-topic
   

4. Consume some messages from the topic to get consumer offsets

   docker-compose run datadog-agent /opt/datadog-agent/embedded/bin/python2.7 /test.py
   

5. Execute the burrow check

   docker-compose run datadog-agent /opt/datadog-agent/agent/agent.py check burrow