Cymometer
Cymometer is a decaying counter for rate limiting using Redis. It provides a simple and efficient mechanism for counting events over a sliding time window, which is useful in applications such as rate limiting.
Cymometer uses Redis sorted sets and Lua scripts to maintain atomic increment and decrement operations, ensuring accurate and thread-safe counting in distributed systems.
Cymometer is designed for simplicity and geared towards the distributed systems use case, such as applying rate-limiting to an application's business logic, background job processing, consumption of external APIs, and so on. Cymometer can also be used for request-based rate limiting, but gems like rack-attack or Rails 8's built-in rate limiting API suit these applications well.
Features
- Atomic Operations: Uses Redis Lua scripts for atomic increment and decrement.
- Configurable Limits: Set custom limits and time windows for your counters.
- Transaction Support: Execute blocks of code only if the counter can be incremented.
-
No Redis Client Dependency: Works with any Redis client, as long as you assign it to
Cymometer.redis
. - Helper DSL: Optional, makes it easy to add per-class, named rate limiters to jobs, service objects, or any Ruby class.
Installation
gem 'cymometer'
Usage
Configuration
First, assign a Redis client to Cymometer.redis
before using the gem:
require 'redis'
require 'cymometer'
# Initialize Redis client
redis = Redis.new(host: 'localhost', port: 6379, db: 0)
# Assign Redis client to Cymometer
Cymometer.redis = redis
The Cymometer gem does not specify a dependency on a particular Redis client. You can configure a default Redis client to be used by all Cymometer counters, or supply one to the Cymometer::Counter
initializer.
Global Configuration
custom_redis = Redis.new(host: 'localhost', port: 6379, db: 1)
Cymometer.redis = custom_redis
Per-Counter Configuration
custom_redis = Redis.new(host: 'localhost', port: 6379, db: 1)
counter = Cymometer::Counter.new(
key_namespace: 'api',
key: 'user_456',
limit: 500,
window: 1800, # 30 minutes
redis: custom_redis
)
Create a Counter
Create a new counter with a namespace, key, limit, and window size:
counter = Cymometer::Counter.new(
key_namespace: 'api',
key: 'user_123',
limit: 1000, # Maximum allowed actions in the time window
window: 3600 # Time window in seconds (e.g., 1 hour)
)
Options
-
key_namespace
: A namespace to group related counters. -
key
: A unique identifier for the counter (e.g., user ID). -
limit
: The maximum number of allowed actions within the time window. -
window
: The time window in seconds over which the actions are counted.
Increment a Counter
To increment the counter and check if the limit has been exceeded:
begin
count = counter.increment!
puts "Counter incremented. Current count: #{count}"
rescue Cymometer::Counter::LimitExceeded => e
puts "Rate limit exceeded: #{e.message}"
end
If the limit is not exceeded, increment! returns the current count. Otherwise, a Cymometer::Counter::LimitExceeded
exception is raised.
Decrement a Counter
If you need to undo an action (e.g., due to an error), you can decrement the counter:
count = counter.decrement!
puts "Counter decremented. Current count: #{count}"
The decrement!
method safely decrements the counter by removing the oldest entry. If the counter is already at zero, it remains unchanged.
Getting the Current Count
To retrieve the current count without modifying the counter:
puts "Current count: #{counter.count}"
Using Transactions
Transactions allow you to execute a block of code only if the counter can be incremented:
begin
result = counter.transaction do
# Your rate-limited code goes here
perform_api_call
"Success"
end
puts result
rescue Cymometer::Counter::LimitExceeded => e
puts "Rate limit exceeded: #{e.message}"
rescue => e
puts "An error occurred: #{e.message}"
end
If the counter can be incremented, the block is executed. If the block raises an exception, the counter is decremented automatically, and the exception is re-raised. If the limit is exceeded, the block is not executed, and Cymometer::Counter::LimitExceeded
is raised.
Rollback Behavior
If you want the counter to remain incremented even when an exception occurs (i.e., not roll back the increment), set rollback
to false:
begin
result = counter.transaction(rollback: false) do
# Your rate-limited code here
perform_api_call
"Success"
end
puts result
rescue Cymometer::Counter::LimitExceeded => e
puts "Rate limit exceeded: #{e.message}"
rescue => e
puts "An error occurred: #{e.message}"
# The counter remains incremented despite the exception
end
Rollback Enabled (default)
Useful when you want to ensure that only successful actions count against the rate limit (e.g., API endpoints where you don’t want failed requests (due to server errors) to penalize the user).
Rollback Disabled (with rollback: false
)
Useful when you want all attempts, successful or not, to count against the rate limit to prevent abuse (e.g., login attempts where you need to prevent brute-force attacks by limiting the number of attempts regardless of success).
Cymometer::Helper DSL
The Cymometer::Helper
DSL makes it easy to add per-class, named rate limiters to jobs, service objects, or any Ruby class. Counters are lazily instantiated and memoized per instance, so repeat calls are efficient. You declare your counters and options up front with a simple block, then access them with the counter method. No boilerplate required!
When to Use
- Background jobs: Limit how often certain work can be performed per job type or account.
- Custom service classes: Rate limit expensive or risky operations across different classes or namespaces.
- Dynamic keys: Build counter keys from instance data (e.g., a user ID, IP address, etc) at runtime.
How It Works
- Include the module in your class.
- Declare counters and options in a configure_cymometer block using a simple DSL.
- Access counters by name with counter(:counter_name) anywhere in your class.
DSL Options
-
namespace "my_app"
: Sets a prefix for all counter keys. -
counter :name do ... end
: Defines a new counter. Inside the block, you can specify:-
limit
: Max actions per window. -
window
: Time window in seconds. -
key { ... }
: (Optional) Block to dynamically build the counter's unique key using instance data.
-
Here's an example:
class MyJob
include Cymometer::Helper
configure_cymometer do
namespace "my_app"
counter :fast_calls do
limit 10 # max 10 per window
window 30 # 30 seconds
end
counter :slow_calls do
limit 3
window 300 # 5 minutes
key { some_dynamic_method } # key is built at runtime
end
end
def perform
# Increment and check the :fast_calls rate limit
counter(:fast_calls).increment!
# Use a transaction for the :slow_calls counter
counter(:slow_calls).transaction do
do_something_slow
end
end
private
def some_dynamic_method
"user_#{user_id}"
end
end
Note
Centralizing your rate limits into configuration (e.g, Rails.application.config_for(:rate_limits)[:my_job][:fast_calls]
in Rails) is a great way to keep them clear, maintainable, and easy to test.
Testing with Cymometer::Helper
Cymometer::Helper
makes it very straightforward to test rate limiting configuration and behavior in your test suite. We'll use RSpec for these examples, but you could achieve the same result with Minitest or your preferred test suite.
To start, create a Cymometer::Counter
test double:
let(:test_counter) { instance_double("Cymometer::Counter") }
Then, add some stubs. In the example below, we configure spy methods to return our test double from Cymometer::Counter#new
, and configure the transaction
and increment!
methods to increment an instance variable counter for our tests.
allow(Cymometer::Counter).to receive(:new).and_return(test_counter)
@transactions_counter = 0
allow(test_counter).to receive(:transaction) do |&block|
@transactions_counter += 1
block&.call
end
allow(test_counter).to receive(:increment!) { @transactions_counter += 1 }
You can then wire up tests for rate limiting behavior. Here's an example:
describe "rate limiting behavior" do
before do
@transactions_counter = 0
end
# Tests for configuration
it "configures and uses Cymometer counters for rate limiting" do
counters = described_class.cymometer_counters
expect(counters[:hour_api_calls]).not_to be_nil
expect(counters[:hour_api_calls][:limit]).to equal(Rails.application.config_for(:rate_limits)[:api_calls_job][:per_hour])
expect(counters[:hour_api_calls][:window]).to equal(1.hour.to_i)
end
# Tests for behavior
context "for a project in trial mode" do
let(:job) { described_class.new }
before do
allow(project).to receive(:trial_mode?).and_return(true)
end
it "increments all counters" do
@transactions_counter = 0
described_class.perform_now(backlog_entry)
expect(@transactions_counter).to eq(2)
end
end
end
Patterns
Retry with ActiveJob
If you're using Cymometer to rate limit background jobs (perhaps ones that consume an external API), you can use ActiveJob's retry_on
mechanism to re-enqueue the job automatically if a rate limit exception is raised:
retry_on Cymometer::Counter::LimitExceeded, wait: :polynomially_longer, attempts: :unlimited, jitter: 0.5
Depending on your setup, you might also combine Cymometer with concurrency limiting and job uniqueness mechanisms for finer-grained control.
Interlocking Rate Limits
Sometimes, you might be working with an external API or system that imposes multi-level rate limits, e.g.:
- No more than X requests per hour, and no more than Y requests per minute.
Thankfully, this is easy to address with Cymometer: Just nest transactions and increment operations against multiple counters alongside or within one another.
You have a few options to consider depending on rollback behavior, exception handling, and so on.
counter(:per_hour).increment!
counter(:per_minute).increment!
# Or:
counter(:per_hour).transaction do
counter(:per_minute).increment!
end
# Or:
counter(:per_hour).transaction do
counter(:per_minute).transaction do
# Do your thing...
end
end
Development
After checking out the repo, run bin/setup
to install dependencies. Then, run rake test
to run the tests. You can also run bin/console
for an interactive prompt that will allow you to experiment.
To install this gem onto your local machine, run bundle exec rake install
. To release a new version, update the version number in version.rb
, and then run bundle exec rake release
, which will create a git tag for the version, push git commits and the created tag, and push the .gem
file to rubygems.org.
Contributing
Bug reports and pull requests are welcome on GitHub at https://github.com/chtzvt/cymometer.
License
The gem is available as open source under the terms of the MIT License.