• Homepage •
Supported protocols •
Chainstack blog •
Chainstack docs •
Blockchain API reference •
• Start for free •
This project allows you to benchmark your blockchain infrastructure. It uses Locust under the hood.
Chainbench lets you to easily define profiles for any EVM-compatible chain. You can use not only hard-coded values but also real chain data to generate dynamic call parameters.
Main features:
- Built-in profiles for Ethereum, Binance Smart Chain, Polygon, Oasis, Avalanche and Solana
- Support for custom profiles
- Dynamic call params generation using real chain data
- Headless and web UI modes
Check out the docs for more information about the profile creation.
- Python 3.10+
- Poetry 1.7+ (installation instructions here)
pip install chainbenchAfter installation, you can run the tool using the following command:
chainbench start --helpClone the repository:
git clone git@github.com:chainstacklabs/chainbench.gitInstall dependencies:
cd chainbench && poetry install --without devWhen installing using Poetry, you can run the tool using the following command:
poetry run chainbenchAll the examples below assume that you have installed the tool using pip. If you installed it using poetry, replace chainbench with poetry run chainbench.
To list all available commands and options, run:
chainbench --helpstart command is used to start a load test. To learn about the parameters and flags, run the following command:
chainbench start --helpYou may specify a method to test, for example:
chainbench start eth_blockNumber --users 50 --workers 2 --test-time 12h --target https://node-url --headless --autoquit This will run a load test with eth_blockNumber method calls with 2 workers, 50 users and 12 hours test time in headless mode with default test data size S.
After the test is finished, the tool will automatically quit.
If you run the command above without '--headless' flag, it will start a web UI where you may select other methods to run tests with.
To see which methods are supported, run:
chainbench list methodsYou may also specify a profile with multiple methods and weights to use for the test instead of testing a single method. For example:
chainbench start --profile bsc.general --users 50 --workers 2 --test-time 12h --target https://node-url --headless --autoquitThis will run a load test with a general BSC profile.
-p, --profile: Specifies the profile to use for the benchmark. Available profiles can be found in the profile directory. Sample usage-p bsc.general-s, --shape: Specifies the shape of the load pattern. List available shapes withchainbench list shapes.-u, --users: Sets the number of simulated users to use for the benchmark.-r, --spawn-rate: Sets the spawn rate of users per second.-w, --workers: Sets the number of worker threads to use for the benchmark.-t, --test-time: Sets the duration of the test to run.--target: Specifies the target blockchain node URL that the benchmark will connect to.--headless: Runs the benchmark in headless mode, meaning that no graphical user interface (GUI) will be displayed during the test. This is useful for running the test on a remote server or when the GUI is not needed.--autoquit: Tells the Chainbench tool to automatically quit after the test has finished. This is useful for running the benchmark in an automated environment where manual intervention is not desired.--help: Displays the help message.--debug-trace-methods: Enables tasks tagged with debug or trace to be executed-E, --exclude-tags: Exclude tasks tagged with custom tags from the test. You may specify this option multiple times.--use-latest-blocks: Use latest blocks for test data generation and runs a background process to update the test data with latest blocks.--size: Specifies the test data size. Available values are XS, S, M, L, XL. Default is S.--batch: Runs the test using batch requests. This will send multiple requests in a single batch request. The number of requests in a batch can be specified using the--batch-sizeflag. Default batch size is 10.
You may also run chainbench start --help for the full list of parameters and flags.
Default profiles are located in the profile directory. For a tutorial on how to create custom profiles, please refer to this document.
You may use the following command to list all profiles available out of the box:
chainbench list profilesThe -d or --profile-dir flag can be used to specify a custom directory with profiles. For example:
chainbench start --profile-dir /path/to/profiles --profile my-profile --users 50 --workers 2 --test-time 12h --target https://node-url --headless --autoquitThis will run a load test using /path/to/profiles/my-profile.py profile.
It's possible to group profiles into directories. For example, you can create a directory called bsc and put all the BSC profiles there. Then you can run a load test using the following command:
chainbench start --profile-dir /path/to/profiles --profile bsc.my-profile --users 50 --workers 2 --test-time 12h --target https://node-url --headless --autoquitChainbench will look for the profile in /path/to/profiles/bsc/my-profile.py. Currently, only one level of nesting is supported.
There are built-in evm.light and evm.heavy profiles for EVM-compatible chains.
Here's an example of how to run a load test for Ethereum using the evm.light profile:
chainbench start --profile evm.light --users 50 --workers 2 --test-time 12h --target https://node-url --headless --autoquitLoad pattern shapes are used to define how the load will be distributed over time. You may specify the shape of the load pattern using the -s or --shape flag.
This is an optional flag and if not specified, the default shape will be used. The default shape is ramp-up which means the load will increase linearly over time at
the spawn-rate until the specified number of users is reached, after that it will maintain the number of users until test duration is over.
Other available shapes are:
step- The load will increase in steps.--spawn-rateflag is required to specify the step size. The number of steps will be calculated based on--usersdivided by--spawn-rate. The duration of each step will be calculated based on--test-timedivided by the number of steps.spike- The load will run in a spike pattern. The load will ramp up to 10% of the total users for 40% of the test duration and then spike to 100% of the total users as specified by--usersfor 20% of test duration and then reduce back to 10% of total users until the test duration is over.
You may specify the test data size using the --size flag. This will determine how much data is used in the test.
Take note that larger data size will result in longer test data generation time before the test starts.
We plan to add an option to reuse previously generated test data in the future to avoid regenerating the same
test data for multiple test runs.
| Size | Blocks |
|---|---|
| XS | 10 |
| S | 100 |
| M | 1,000 |
| L | 10,000 |
| XL | 100,000 |
When running tests on nodes that are running in sync modes that do not have access to the full blockchain history, it is recommended to use the --use-latest-blocks flag.
Example: If the node runs in a sync mode that only keeps the last 128 blocks in history, you may run a test with the following command:
chainbench start --profile evm.light --users 50 --workers 2 --test-time 1h --target https://node-url --headless --autoquit --use-latest-blocks --size SWith test data size S and --use-latest-blocks flag, the tool will generate test data using the latest 100 blocks.
Chainbench will keep the test data up to date by running a background process that will update the test data with latest blocks.
Monitors are separate processes that run during the test to collect or process some additional data and metrics relevant to the test.
For example, sync-lag-monitor will collect the latest block information from the node under test, check the timestamp and compare it to current time to calculate how much the node lags behind.
You may include monitors in your test by using the -m or --monitor option and specifying the name of the monitor. At the moment, monitors only work in headless mode.
Here's an example:
chainbench start --profile evm.light --users 50 --workers 2 --test-time 12h --target https://node-url --headless --autoquit -m sync-lag-monitorRun the following command to run a load test for BSC in UI mode. It will start a web server on port 8089. Target is required to initialize the test data, however you may change the target endpoint later in the UI, along with the number of users, spawn rate and test time. Do take note when changing the target option that test data initialized with an Ethereum node endpoint, for example, can only be used to test other Ethereum node endpoints.
chainbench start --profile bsc.general --workers 1 --target https://any-working-node-endpoint.comIf you'd like to select which profiles or more specifically which user classes to use for the test, you may start a test in Web UI Mode without specifying a profile or passing in the method argument.
chainbench start --workers 1 --target https://any-working-node-endpoint.com --profile-dir chainbench/profile/evmThis will start a test with all the EVM profiles and user classes available in the UI for you to select.
If you want to run a load test for BSC in headless mode, run the following command:
chainbench start --profile bsc.general --workers 4 --users 100 --test-time 1h --target https://node-url --headless --autoquitIt will run a load test for BSC with 4 workers, 100 users and 1 hour test time.
In practice, you will probably want to run the benchmark on a remote server. Here's the example utilizing nohup:
nohup chainbench start --profile bsc.general --workers 4 --users 100 --test-time 1h --target https://node-url --headless --autoquit &This command will discover all available rpc methods on the specified endpoint and print them to the console. List of methods that are tested are based on the --clients option.
chainbench discover https://node-url --clients geth,erigonTo list valid arguments for --clients option and the reference client version, run:
chainbench list clientsIf you don't specify the --clients option, the tool will default to Ethereum JSON-RPC Specification (eth).
This project is licensed under the Apache 2.0 License.