Quick start for docker-cluster

The metrictank repository holds various stack configurations. See docker for more info. If you're not familiar with Grafana dashboarding and datasources, you should first follow our standard quickstart guide which goes further in depth on that.

The docker-cluster image is similar to docker-standard, but with the following changes and features:

• Cluster of 4 metrictanks (with 2x replication)
• Separate Graphite monitoring server
• The metrictank binary is mounted from metrictank/build, so make bin must be run before spinning up the stack (benefits: speeds up development and testing)
• Loads custom configurations and scripts
• Supports tags
• More dashboards are available
• Kafka input and clustering backend
• Short chunkspan & numchunks (benefits: easy to trigger mem and mem_and_cass requests, frequent cass saves, notifier messages etc)
• Prometheus monitoring
• Jaeger tracing
• caddy-proxied endpoints/datasources to check with different org-id auth headers
• tsdb-gw

The following programs are required to build and run the docker-cluster image:

• Docker
• docker-compose >= version 1.6
• Go (needed to build metrictank)
  • Ensure that $GOPATH/bin is added to your $PATH

Getting the repository

Pull down the repository and its dependencies:

go get github.com/grafana/metrictank/...
cd $GOPATH/src/github.com/grafana/metrictank

Build metrictank

Build metrictank and the docker images:

make

make uses/creates the directory $GOPATH/src/github.com/grafana/metrictank/build. If you tried to spin up the stack prior to building metrictank then docker already created this directory, as root, and it will need to be deleted (then run make again).

Bring up the stack

The stack will listen on the following ports:

• 80 tcp (carbon)
• 2003 tcp (metrictank's carbon input)
• 2181 tcp (kafka)
• 3000 tcp (grafana's http port)
• 6060 tcp (metrictank's internal endpoint)
• 6061 tcp (caddy)
• 6062 tcp (caddy)
• 6063 tcp (caddy)
• 6831 udp (jaeger)
• 8080 tcp (the graphite query endpoint)
• 8081 tcp (caddy)
• 8082 tcp (caddy)
• 8125 udp (statsd endpoint)
• 9042 tcp (cassandra)
• 9090 tcp (prometheus)
• 9092 tcp (kafka)
• 9100 tcp (prometheus node exporter)
• 9999 tcp (kafka)
• 16686 tcp (jaeger)

If you already have something else listening on those ports (such as a carbon or grafana server), shut it down, as it will conflict.

Inside your copy of the repository, you can bring up the stack like so:

cd docker/docker-cluster
docker-compose up

If this gives you the error service 'version' doesn't have any configuration options, your version of docker-compose is too old and you need to update to >=1.6.

Wait until the stack is up.

Working with Grafana and metrictank

In your browser, open Grafana at http://localhost:3000 (or your docker-machine address) and log in as admin:admin.
If Grafana prompts you to change the password, you can skip it, since it doesn't matter for a local test setup.

Working with the cluster

This stack has some unique properties not found in the standard stack, such as multiple data sources to choose from including, but not limited to:

• monitoring and prometheus for data about the stack
• metrictank to see data in metrictank
• graphite to see data in metrictank via graphite

When using fakemetrics to generate data you should now use kafka

• fakemetrics feed --kafka-mdm-addr localhost:9092
• stats will still show up under some.id.of.a.metric.*

If anything doesn't work, please let us know via a ticket on github or reach out on slack. See Community

Shut down the stack

You can shut down the stack by just hitting Ctrl-C on your running docker-compose up command.

You can remove the containers like so:

docker-compose stop

To clean up all data so you can start fresh, run this after you stopped the stack:

docker rm -v $(docker ps -a -q -f status=exited)

This will remove the stopped containers and their data volumes.