The Event Gateway combines both API Gateway and Pub/Sub functionality into a single event-driven experience. It's dataflow for event-driven, serverless architectures. It routes Events (data) to Functions (serverless compute). Everything it cares about is an event! Even calling a function. It makes it easy to share events across different systems, teams and organizations!
Use the Event Gateway right now, by running the Event Gateway Getting Started Application with the Serverless Framework.
Features:
- Platform agnostic - All your cloud services are now compatible with one another: share cross-cloud functions and events with AWS Lambda, Microsoft Azure, IBM OpenWhisk and Google Cloud Platform.
- Send events from any cloud - Data streams in your application become events. Centralize events from any cloud provider to get a bird’s eye view of all the data flowing through your cloud.
- React to cross-cloud events - You aren’t locked in to events and functions being on the same provider: Any event, on any cloud, can trigger any function. Set events and functions up like dominoes and watch them fall.
- Expose events to your team - Share events and functions to other parts of the application. Your teammates can find them and utilize them in their own services.
- Extendable through middleware - Perform data transforms, authorizations, serializations, and other custom computes straight from the Event Gateway.
The Event Gateway is a L7 proxy and realtime dataflow engine, intended for use with Functions-as-a-Service on AWS, Azure, Google & IBM.
The project is under heavy development. The APIs will continue to change until we release a 1.0.0 version. It's not yet ready for production applications.
Website • Slack • Newsletter • Forum • Meetups • Twitter
- Quick Start
- Running the Event Gateway
- Motivation
- Components
- System Events
- APIs
- Client SDKs
- Versioning
- Comparison
- Architecture
- Background
Looking for an example to get started? The easiest way to use the Event Gateway is with the serverless-event-gateway-plugin with the Serverless Framework. Check out the Getting Started Example to deploy your first service to the Event Gateway.
If you don't want to run the Event Gateway yourself, you can use the hosted version provided by the Serverless team. Please contact us for an invitation to the beta.
There is a official Docker image.
docker run -p 4000:4000 -p 4001:4001 serverless/event-gateway -devOn macOS or Linux run the following to download the binary:
curl -sfL https://raw.githubusercontent.com/serverless/event-gateway/master/install.sh | sh
On Windows download binary.
Then run the binary in development mode with:
$ event-gateway -devIf you want more detailed information on running and developing with the Event Gateway, please check Running Locally and Developing guides.
- It is cumbersome to plug things into each other. This should be easy! Why do I need to set up a queue system to keep track of new user registrations or failed logins?
- Introspection is terrible. There is no performant way to emit logs and metrics from a function. How do I know a new piece of code is actually working? How do I feed metrics to my existing monitoring system? How do I plug this function into to my existing analytics system?
- Using new functions is risky without the ability to incrementally deploy them.
- The AWS API Gateway is frequently cited as a performance and cost-prohibitive factor for using AWS Lambda.
Discover and call serverless functions from anything that can reach the Event Gateway. Function Discovery supports the following function types:
- FaaS functions (AWS Lambda, Google Cloud Functions, Azure Functions, OpenWhisk Actions)
- Connectors (AWS Kinesis, AWS Kinesis Firehose, AWS SQS)
- HTTP endpoints/Webhook (e.g. POST http://example.com/function)
Function Discovery stores information about functions allowing the Event Gateway to call them as a reaction to received event.
curl --request POST \
--url http://localhost:4001/v1/spaces/default/functions \
--header 'content-type: application/json' \
--data '{
"functionId": "hello",
"type": "awslambda",
"provider":{
"arn": "arn:aws:lambda:us-east-1:377024778620:function:bluegreen-dev-hello",
"region": "us-east-1"
}
}'const eventGateway = new EventGateway({ url: 'http://localhost' })
eventGateway.registerFunction({
functionId: 'sendEmail',
type: 'awslambda',
provider: {
arn: 'xxx',
region: 'us-west-2'
}
})curl --request POST \
--url http://localhost:4000/ \
--header 'content-type: application/json' \
--header 'event: invoke' \
--header 'function-id: createUser' \
--data '{ "name": "Max" }'const eventGateway = new EventGateway({ url: 'http://localhost' })
eventGateway.invoke({
functionId: 'createUser',
data: { name: 'Max' }
})Lightweight pub/sub system. Allows functions to asynchronously receive custom events. Instead of rewriting your functions every time you want to send data to another place, this can be handled entirely in configuration using the Event Gateway. This completely decouples functions from one another, reducing communication costs across teams, eliminates effort spent redeploying functions, and allows you to easily share events across functions, HTTP services, even different cloud providers. Functions may be registered as subscribers to a custom event. When an event occurs, all subscribers are called asynchronously with the event as its argument.
Creating a subscription requires providing ID of registered function, an event type and a path (/ by default). The
path property indicated URL path which Events API will be listening on.
curl --request POST \
--url http://localhost:4001/v1/spaces/default/subscriptions \
--header 'content-type: application/json' \
--data '{
"functionId": "sendEmail",
"event": "user.created",
"path": "/myteam"
}'const eventGateway = new EventGateway({ url: 'http://localhost' })
eventGateway.subscribe({
event: 'user.created',
functionId: 'sendEmail',
path: '/myteam'
})sendEmail function will be invoked for every user.created event to <Events API>/myteam endpoint.
curl --request POST \
--url http://localhost:4000/ \
--header 'content-type: application/json' \
--header 'event: user.created' \
--data '{ "name": "Max" }'const eventGateway = new EventGateway({ url: 'http://localhost' })
eventGateway.emit({
event: 'user.created',
data: { name: 'Max' }
})Custom event subscriptions are asynchronous. There is a special http event type for creating synchronous
subscriptions. http event is an HTTP request received to specified path and for specified HTTP method. There can be
only one http subscription for the same method and path pair.
curl --request POST \
--url http://localhost:4001/v1/spaces/default/subscriptions \
--header 'content-type: application/json' \
--data '{
"functionId": "listUsers",
"event": "http",
"method": "GET",
"path": "/users"
}'const eventGateway = new EventGateway({ url: 'http://localhost' })
eventGateway.subscribe({
functionId: 'listUsers',
event: 'http',
method: 'GET',
path: '/users'
})listUsers function will be invoked for every HTTP GET request to <Events API>/users endpoint.
One additional concept in the Event Gateway are Spaces. Spaces provide isolation between resources. Space is a coarse-grained sandbox in which entities (Functions and Subscriptions) can interact freely. All actions are possible within a space: publishing, subscribing and invoking.
Space is not about access control/authentication/authorization. It's only about isolation. It doesn't enforce any specific subscription path.
This is how Spaces fit different needs depending on use-case:
- single user - single user uses default space for registering function and creating subscriptions.
- multiple teams/departments - different teams/departments use different spaces for isolation and for hiding internal implementation and architecture.
Technically speaking Space is a mandatory field ("default" by default) on Function or Subscription object that user has to provide during function registration or subscription creation. Space is a first class concept in Config API. Config API can register function in specific space or list all functions or subscriptions from a space.
System Events are special type of events emitted by the Event Gateway instance. They are emitted on each stage of event processing flow starting from receiving event to function invocation end. Those events are:
gateway.event.received- the event is emitted when an event was received by Events API. Data fields:event- event payloadpath- Events API pathheaders- HTTP request headers
gateway.function.invoking- the event emitted before invoking a function. Data fields:event- event payloadfunctionId- registered function ID
gateway.function.invoked- the event emitted after successful function invocation. Data fields:event- event payloadfunctionId- registered function IDresult- function response
gateway.function.invocationFailed- the event emitted after failed function invocation. Data fields:event- event payloadfunctionId- registered function IDerror- invocation error
The Event Gateway has two APIs: the Configuration API for registering functions and subscriptions, and the runtime Events API for sending events into the Event Gateway.
This project uses Semantic Versioning 2.0.0. We are in initial development phase right now (v0.X.Y). The public APIs should not be considered stable. Every breaking change will be listed in the release changelog.
- it's not a replacement for message queues (no message ordering, currently weak durability guarantees only)
- it's not a replacement for streaming platforms (no processing capability and consumers group)
- it's not a replacement for existing service discovery solutions from the microservices world
The Event Gateway is NOT a FaaS platform. It integrates with existing FaaS providers (AWS Lambda, Google Cloud Functions, Azure Functions, OpenWhisk Actions). The Event Gateway enables building large serverless architectures in a unified way across different providers.
┌──────────────┐
│ │
│ Client │
│ │
└──────────────┘
▲
│
Event
│
▼
┌───────────────────────────────────────────────────────────┐
│ │
│ Event Gateway Cluster │
│ │
└───────────────────────────────────────────────────────────┘
▲
│
│
┌─────────────────────────────┼─────────────────────────────┐
│ │ │
│ │ │
▼ ▼ ▼
┌───────────────┐ ┌───────────────┐ ┌───────────────┐
│ AWS Lambda │ │ Google Cloud │ │Azure Function │
│ Function │ │ Function │ │ │
│ │ │ │ │ Region: │
│ Region: │ │ Region: │ │ West US │
│ us-east-1 │ │ us-central1 │ │ │
└───────────────┘ └───────────────┘ └───────────────┘
The Event Gateway instances use a strongly consistent, subscribable DB (initially etcd, with support for Consul, and Zookeeper planned) to store and broadcast configuration. The instances locally cache configuration used to drive low-latency event routing. The instance local cache is built asynchronously based on events from backing DB.
The Event Gateway is a horizontally scalable system. It can be scaled by adding instances to the cluster. A cluster is a group of instances sharing the same database. A cluster can be created in one cloud region, across multiple regions, across multiple cloud provider or even in both cloud and on-premise data centers.
The Event Gateway is a stateless service and there is no direct communication between different instances. All configuration data is shared using backing DB. If the instance from region 1 needs to call a function from region 2 the invocation is not routed through the instance in region 2. The instance from region 1 invokes the function from region 2 directly.
┌─────────────────────────────────────────────Event Gateway Cluster──────────────────────────────────────────────┐
│ │
│ │
│ Cloud Region 1───────┐ │
│ │ │ │
│ │ ┌─────────────┐ │ │
│ │ │ │ │ │
│ ┌ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─│─ ▶│etcd cluster │◀ ┼ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ │
│ │ │ │ │ │ │
│ │ │ └─────────────┘ │ │
│ │ ▲ │ │ │
│ │ │ │ │
│ Cloud Region 2───────┐ │ │ │ Cloud Regio│ 3───────┐ │
│ │ │ │ │ │ │ │ │
│ │ ▼ │ │ ▼ │ │ ▼ │ │
│ │ ┌───────────────┐ │ │ ┌──────────────┐ │ │ ┌──────────────┐ │ │
│ │ │ │ │ │ │ │ │ │ │ │ │ │
│ │ │ Event Gateway │ │ │ │Event Gateway │ │ │ │Event Gateway │ │ │
│ │ │ instance │◀┼──────────┐ │ │ instance │◀─┼──────────┐ │ │ instance │ │ │
│ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │
│ │ └───────────────┘ │ │ │ └──────────────┘ │ │ │ └──────────────┘ │ │
│ │ ▲ │ │ │ ▲ │ │ │ ▲ │ │
│ │ │ │ │ │ │ │ │ │ │ │ │
│ │ │ │ │ │ │ │ │ │ │ │ │
│ │ ▼ │ │ │ ▼ │ │ │ ▼ │ │
│ │ ┌───┐ │ │ │ ┌───┐ │ │ │ ┌───┐ │ │
│ │ │ λ ├┐ │ └───┼───────▶│ λ ├┐ │ └────┼───────▶│ λ ├┐ │ │
│ │ └┬──┘│ │ │ └┬──┘│ │ │ └┬──┘│ │ │
│ │ └───┘ │ │ └───┘ │ │ └───┘ │ │
│ └────────────────────┘ └────────────────────┘ └────────────────────┘ │
│ │
│ │
└────────────────────────────────────────────────────────────────────────────────────────────────────────────────┘
The event received by Event Gateway is stored only in memory, it's not persisted to disk before processing. This means that in case of hardware failure or software crash the event may not be delivered to the subscriber. For a synchronous subscription (http or invoke event) it can manifest as error message returned to the requester. For asynchronous custom event with multiple subscribers it means that the event may not be delivered to all of the subscribers.
Event Gateway attempts delivery fulfillment for an event only once and consequently any event received successfully by the Event Gateway is guaranteed to be received by the subscriber at most once. That said, the nature of Event Gateway provider implementation could result in retries under specific circumstances, but these should not cause delivering the same event multiple times. For example, Providers for AWS Services that use the AWS SDK are subject to auto retry logic that's built into the SDK (AWS documentation on API retries).
AWS Lambda provider uses RequestResponse invocation type which means that retry logic for asynchronous AWS events doesn't apply here. Among others it means, that failed deliveries of custom events are not sent to DLQ. Please find more information in Understanding Retry Behavior, "Synchronous invocation" section.
SOA came along with a new set of challenges. In monolithic architectures, it was simple to call a built-in library or rarely-changing external service. In SOA it involves much more network communication which is not reliable. The main problems to solve include:
- Where is the service deployed? How many instances are there? Which instance is the closest to me? (service discovery)
- Requests to the service should be balanced between all service instances (load balancing)
- If a remote service call failed I want to retry it (retries)
- If the service instance failed I want to stop sending requests there (circuit breaking)
- Services are written in multiple languages, I want to communicate between them using the best language for the particular task (sidecar)
- Calling remote service should not require setting up new connection every time as it increases request time (persistent connections)
The following systems are solutions those problems:
The main goal of those tools is to manage the inconveniences of network communication.
The greatest benefit of serverless/FaaS is that it solves almost all of above problems:
- service discovery: I don't care! I have a function name, that's all I need.
- load balancing: I don't care! I know that there will be a function to handle my request (blue/green deployments still an issue though)
- retries: It's highly unusual that my request will not proceed as function instances are ephemeral and failing function is immediately replaced with a new instance. If it happens I can easily send another request. In case of failure, it's easy to understand what is the cause.
- circuit breaking: Functions are ephemeral and auto-scaled, low possibility of flooding/DoS & cascading failures.
- sidecar: calling function is as simple as calling method from cloud provider SDK.
- in FaaS setting up persistent connection between two functions defeats the purpose as functions instances are ephemeral.
Tools like Envoy/Linkerd solve different domain of technical problems that doesn't occur in serverless space. They have a lot of features that are unnecessary in the context of serverless computing.
Service discovery problems may be relevant to serverless architectures, especially when we have a multi-cloud setup or we want to call a serverless function from a legacy system (microservices, etc...). There is a need for some proxy that will know where the function is actually deployed and have retry logic built-in. Mapping from function name to serverless function calling metadata is a different problem from tracking the availability of a changing number of service instances. That's why there is a room for new tools that solves function discovery problem rather than the service discovery problem. Those problems are fundamentally different.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent