Grafana stack

Tiny docker images for grafana, graphite and statsdly (statsd implementation). For example, grafana-4.6.3 ~ 90M, graphite-1.1.1 ~ 125M, statsdly-0.4.1 ~ 33M. All images are build from official sources with sane (read very opinionated :) configuration defaults.

Docker network

Grafana needs access to graphite so you should create a docker network, use links (deprecated) or use docker-compose. All examples below assume grafana-stack network.

docker network create grafana-stack

Graphite

Dockerhub: baverman/graphite.

Start container:

mkdir -p data/carbon
export DOCKER_USER=$(id -u):$(id -g)
docker run -d --name graphite -p 2003:2003 --restart unless-stopped --network grafana-stack \
           -v $PWD/data/carbon:/data -u $DOCKER_USER baverman/graphite

This command will start graphite-web and carbon-cache services under current user. Only 2003 TCP port (carbon text protocol) will be exposed and all data will be written to data/carbon directory.

Ports:

• 2003: carbon text protocol.
• 8080: graphite-web application, needed for grafana to fetch metrics, it's better to not expose it to public until you configure authorization.
• 7002: carbon cache interface to get metrics data.

Volumes:

• /data: carbon and graphite-web storage dir.

• /conf: directory with all config files. Image includes default configs but one can override any by mounting own /conf volume. Obtaining default /conf/carbon.conf:

  docker run --rm baverman/graphite /template.py /tpl/carbon.conf > /tmp/carbon.conf
  

Environment variables:

• CC_*: Overrides settings from carbon.conf [cache] section. For example you can set CC_MAX_UPDATES_PER_SECOND=100 to limit IOPS.

• CARBON_STORAGE_SCHEMA_*: control retention policies in storage-schemas.conf. Format is pattern|retentions. You can define any number of variables. Default retentions are:

  CARBON_STORAGE_SCHEMA_CARBON=^carbon\.|60s:1d,5m:7d,10m:30d
  CARBON_STORAGE_SCHEMA_ZDEFAULT=.*|60s:7d,5m:30d,30m:90d,1h:1y
  

  IMPORTANT! Default minimal retention is 60s, you MUST configure carbon clients to flush metrics not more often than once in every 60s or you will loose data.

• CARBON_STORAGE_AGG_*: control aggregation policies in storage-aggregation.conf. Format is pattern|xFilesFactor|aggregationMethod. You can define any number of variables. Default aggregations are:

  CARBON_STORAGE_AGG_MIN=\.min$|0.1|min
  CARBON_STORAGE_AGG_MAX=\.max$|0.1|max
  CARBON_STORAGE_AGG_SUM=\.count$|0|sum
  CARBON_STORAGE_AGG_ZDEFAULT=.*|0.2|average
  

  Metrics ending with .count can be used to precisely store counters.

Build image:

./graphite/build.sh [name] [graphite-version] [tag]

Grafana

Dockerhub: baverman/grafana.

Start container:

mkdir -p data/grafana
export DOCKER_USER=$(id -u):$(id -g)
docker run -d --name grafana -p 3000:3000 --restart unless-stopped --network grafana-stack \
           -v $PWD/data/grafana:/data -u $DOCKER_USER baverman/grafana

This command will start grafana service under current user. Only 3000 TCP port (HTTP) will be exposed and all data will be written to data/grafana directory.

And you can add http://graphite:8080/ as graphite datasource with 1.1.x version.

Ports:

• 3000: Grafana HTTP port.

Volumes:

• /data: grafana storage dir.

Environment variables:

You can override any grafana settings via envvars.

• GF_SERVER_DOMAIN: domain.
• GF_SERVER_ROOT_URL: root url.
• GF_SECURITY_SECRET_KEY: secret key to sign up cookies.

Build image:

./grafana/build.sh [name] [grafana-version] [sha256-src-checksum] [tag]

Statsdly

Dockerhub: baverman/statsdly.

Start container:

export DOCKER_USER=$(id -u):$(id -g)
docker run -d --name statsdly -p 8125:8125/udp --restart unless-stopped --network grafana-stack \
           -u $DOCKER_USER baverman/statsdly -l 0.0.0.0 -g graphite -f 60

This command will start statsdly service (a StatsD implementation) on 8125 UDP port with 60s flush interval and service will forward metrics to graphite host.

Ports:

• 8125: StatsD UDP port.

Build image:

./statsdly/build.sh [name] [statsdly-version] [tag]