nimbus-eth1/waku/README.md

3.6 KiB

Introduction

wakunode is a cli application that allows you to run a Waku enabled node.

The application and Waku specification are still experimental and fully in flux.

Additionally the original Whisper (EIP-627) protocol can also be enabled as can an experimental Whisper - Waku bridging option.

How to Build & Run

Prerequisites

  • GNU Make, Bash and the usual POSIX utilities. Git 2.9.4 or newer.
  • PCRE

More information on the installation of these can be found here.

Build & Run

make # The first `make` invocation will update all Git submodules and prompt you to run `make` again.
     # It's only required once per Git clone. You'll run `make update` after each `git pull`, in the future,
     # to keep those submodules up to date.
make wakunode
./build/wakunode --help

Using Metrics

Metrics are available for valid envelopes and dropped envelopes.

To compile in an HTTP endpoint for accessing the metrics we need to provide the insecure flag:

make NIMFLAGS="-d:insecure" wakunode
./build/wakunode --metrics-server

Ensure your Prometheus config prometheus.yml contains the targets you care about, e.g.:

scrape_configs:
  - job_name: "waku"
    static_configs:
      - targets: ['localhost:8008', 'localhost:8009', 'localhost:8010']

For visualisation, similar steps can be used as is written down for Nimbus here.

There is a similar example dashboard that includes visualisation of the envelopes available at waku/examples/waku-grafana-dashboard.json.

Testing Waku Protocol

One can set up several nodes, get them connected and then instruct them via the JSON-RPC interface. This can be done via e.g. web3.js, nim-web3 (needs to be updated) or simply curl your way out.

The JSON-RPC interface is currently the same as the one of Whisper. The only difference is the addition of broadcasting the topics interest when a filter with a certain set of topics is subcribed.

Example of a quick simulation using this approach:

# Build wakunode + quicksim
make NIMFLAGS="-d:insecure" wakusim

# Start the simulation nodes, this currently requires multitail to be installed
./build/start_network --topology:FullMesh --amount:6 --test-node-peers:2
# In another shell run
./build/quicksim

The start_network tool will also provide a prometheus.yml with targets set to all simulation nodes that are started. This way you can easily start prometheus with this config, e.g.:

cd waku/metrics/prometheus
prometheus

A Grafana dashboard containing the example dashboard for each simulation node is also generated and can be imported in case you have Grafana running. This dashboard can be found at ./waku/metrics/waku-sim-all-nodes-grafana-dashboard.json

Spec support

This section last updated April 7, 2020

This client of Waku is spec compliant with Waku spec v0.4.

It doesn't yet implement the following recommended features:

  • No support for rate limiting
  • No support for DNS discovery to find Waku nodes
  • It doesn't disconnect a peer if it receives a message before a Status message
  • No support for negotiation with peer supporting multiple versions via Devp2p capabilities in Hello packet

Additionally it makes the following choices:

  • It doesn't send message confirmations
  • It has partial support for accounting:
    • Accounting of total resource usage and total circulated envelopes is done through metrics But no accounting is done for individual peers.