Skip to content

Latest commit

 

History

History
59 lines (40 loc) · 2.08 KB

README.md

File metadata and controls

59 lines (40 loc) · 2.08 KB
Raw

Local Development

This docker-compose file will setup the following environment:

  1. 1 M3DB nodes with a single node acting as an ETCD seed
  2. 1 M3Coordinator node
  3. 1 Grafana node (with a pre-configured Prometheus source)
  4. 1 Prometheus node that scrapes the M3DB/M3Coordinator nodes and writes the metrics to M3Coordinator

The environment variables that let's you configure this setup are:

  • USE_MULTI_DB_NODES=true: uses 3 database nodes instead of 1 for cluster.
  • USE_JAEGER=true: look at traces emitted by M3 services.
  • USE_PROMETHEUS_HA=true: send data to M3 from two HA Prometheus instances to replicate deployments of HA Prometheus sending data to M3.
  • USE_AGGREGATOR=true: use dedicate aggregator to aggregate metrics.
  • USE_AGGREGATOR_HA=true: use two dedicated aggregators for HA aggregated metrics.
  • USE_MULTIPROCESS_COORDINATOR=true: use multi-process coordinator, with default number of processes configured.

Usage

Use the start_m3.sh and stop_m3.sh scripts. Requires successful run of make m3dbnode from project root first.

Grafana

Use Grafana by navigating to http://localhost:3000 and using admin for both the username and password. The M3DB dashboard should already be populated and working.

Jaeger

To start Jaeger, you need to set the environment variable USE_JAEGER to true when you run start_m3.sh.

USE_JAEGER=true ./start_m3.sh

To modify the sampling rate, etc. you can modify the following in your m3dbnode.yml file under db:

tracing:
  backend: jaeger
  jaeger:
    reporter:
      localAgentHostPort: jaeger:6831
    sampler:
      type: const
      param: 1

Use Jaeger by navigating to http://localhost:16686.

Prometheus

Use Prometheus by navigating to http://localhost:9090.

Increasing Load

Load can easily be increased by modifying the prometheus.yml file to reduce the scrape interval to 1s.

Containers Hanging / Unresponsive

Running the entire stack can be resource intensive. If the containers are unresponsive try increasing the amount of cores and memory that the docker daemon is allowed to use.