Meter-gear

Proxy Meter's RESTful API to Eth JSON-RPC, to support Remix, Truffle and more (You should give priority to using Meter's RESTful API).

Quick Start

Installation

On OS X

• Python 3.8+ support

1. Install the system-dependecies

   brew install openssl
   export CFLAGS="-I$(brew --prefix openssl)/include $CFLAGS"
   export LDFLAGS="-L$(brew --prefix openssl)/lib $LDFLAGS"
   

2. Installation of Meter-Gear and it's dependent Python packages via PyPI

   pip3 install meter-gear
   

On Ubuntu

• Python 3.8+ support

1. Install the system-dependecies

   sudo apt-get install build-essential libssl-dev python-dev
   

2. Use a virtual environment to isolate your meter-gear project

   python3 -m venv meter-venv
   source meter-venv/bin/activate
   

3. Installation of Meter-Gear and it's dependent Python packages via PyPI

   pip3 install meter-gear
   

Note that you can activate and deactivate your virtual environment now

source meter-venv/bin/activate
deactivate

See https://docs.python.org/3/library/venv.html for more information

On Windows

• Python 3.8+ support

1. Install Visual C++ Build Tools.

2. Installation of Meter-Gear and it's dependent Python packages via PyPI

   pip3 install meter-gear
   

Run

Installing through pip will make the meter-gear command available on your machine (must run meter client first.)

meter-gear

This will run meter-gear on 127.0.0.1:8545.

You can change its default behavior with the following parameters:

• host: rpc service host, eg: --host 127.0.0.1
• port: rpc service port, eg: --port 8545
• endpoint: meter restful service endpoint, eg: --endpoint http://127.0.0.1:8669
• keystore: keystore file path, eg: --keystore /Users/(username)/keystore), default=meter stand-alone(solo) built-in accounts
• passcode: passcode of keystore, eg: --passcode xxxxxxxx
• debug: bool default=false, whether to display debug logs, eg: --debug true
• log: bool default=false, whether to display rpc logs, eg: --log false

Resource limits and auto-tuning

Meter-Gear detects the CPU quota and memory limit from Linux cgroups and the process file-descriptor limit from RLIMIT_NOFILE. It uses those resources to set a safe Meter REST connection ceiling. The auto-tuner starts at 128 backend calls (or the lower resource ceiling), then uses backend p95 latency, error rate, saturation, and event-loop lag to increase or decrease concurrency.

The startup log includes the detected resources and every calculated limit. Explicit request or operation limits remain fixed; unset limits follow the backend auto-tuner.

Variable	Default	Purpose
GEAR_AUTOTUNE	1	Enable container-aware sizing and runtime adaptation
GEAR_BACKEND_CONNECTION_LIMIT	128	Initial backend concurrency; fixed value when auto-tuning is disabled
GEAR_BACKEND_CONNECTION_MIN	64	Lower adaptive backend bound, capped by available resources
GEAR_BACKEND_CONNECTION_MAX	512	Upper adaptive/connector bound, capped by available resources
GEAR_RPC_CONCURRENCY	automatic	Concurrent HTTP requests and WebSocket messages; setting it disables automatic resizing for this limit
GEAR_RPC_CONCURRENCY_MAX	1024	Maximum automatically resized request limit
GEAR_RPC_OPERATION_LIMIT	automatic	Concurrent JSON-RPC operations; defaults to twice backend concurrency
GEAR_RPC_OPERATION_LIMIT_MAX	1024	Maximum automatically resized operation limit
GEAR_RPC_PER_IP_CONCURRENCY	8	Concurrent request envelopes allowed from one client IP
GEAR_RPC_PER_IP_OPERATION_LIMIT	64	Concurrent JSON-RPC operations allowed from one client IP
GEAR_RPC_QUEUE_TIMEOUT	0.05	Seconds to wait for request/operation admission
GEAR_BACKEND_QUEUE_TIMEOUT	0.25	Seconds a Meter REST call waits for adaptive backend capacity
GEAR_MAX_BATCH_SIZE	50	Maximum calls in one JSON-RPC batch
GEAR_MAX_REQUEST_SIZE	1048576	Maximum HTTP request body in bytes

Each JSON-RPC batch consumes operation capacity equal to its element count. A batch of 50 therefore consumes 50 slots instead of bypassing the limiter as one HTTP request. Every actual Meter REST call is also bounded before it can enter the aiohttp connection queue.

Auto-tuning behavior can be adjusted with these advanced settings:

Variable	Default	Purpose
GEAR_AUTOTUNE_INTERVAL	10	Seconds between adaptive evaluations
GEAR_AUTOTUNE_WINDOW	60	Rolling backend sample window in seconds
GEAR_AUTOTUNE_MIN_SAMPLES	20	Minimum completed calls before changing a limit
GEAR_BACKEND_TARGET_P95	0.5	Healthy backend p95 latency in seconds
GEAR_BACKEND_ERROR_THRESHOLD	0.02	Error fraction that triggers a decrease
GEAR_EVENT_LOOP_LAG_THRESHOLD	0.2	Event-loop lag in seconds that triggers a decrease
GEAR_BACKEND_TOTAL_TIMEOUT	10	Total Meter REST timeout in seconds
GEAR_BACKEND_CONNECT_TIMEOUT	2	Meter REST connection timeout in seconds
GEAR_BACKEND_READ_TIMEOUT	8	Meter REST read timeout in seconds
GEAR_BACKEND_KEEPALIVE_TIMEOUT	15	Idle pooled connection lifetime in seconds

Oversized or per-IP-limited requests return HTTP 429. Globally saturated requests return HTTP 503 instead of creating unbounded backend work.

For Docker deployments, set the file-descriptor limit on the container rather than changing it after startup. The included Compose configuration uses:

ulimits:
  nofile:
    soft: 65536
    hard: 524288

Use Remix

Change the Remix environment to Web3 provide.

Use Truffle

• Truffle 4.0.6+ support

Modify the configuration of truffle first(truffle.js):

module.exports = {
    networks: {
        development: {
            host: "localhost",
            port: 8545,
            network_id: "*" // Match any network id
        }
    }
};

Then you can use truffle's command line tool.