I'm building a graphql gateway service, which merge multiple services into one graph, using Apollo/Node/Express and following the Apollo Federation model. Initially, most of the services I'll be connecting to are REST services.
In all of the examples I find (e.g. here), I see that the gateway project runs multiple instances of ApolloServer, one for every REST service plus one more for the gateway itself, and runs them all using a package like concurrently. Basically the gateway project runs n+1 ApolloServers. Having all of these servers running seems strange to me, but I'm pretty new to this whole ecosystem.
I'm not clear if this is just for demonstration purposes, or is this also how it's implemented and deployed in the real world?
I hope that those were just examples, and are not the expected pattern.
If you need multiple GraphQL Services, each one of those would be served as a separate Domain Graph application, as its own project. Then an additional service (the gateway) would consume all of those applications and expose a single unified GraphQL API.
Related
My company has been using AWS Lambda for many years to run our Spring Boot REST API. We are migrating to GCP and they want me to deploy our code to GCF the same way we were with AWS Lambda, but I am not sure that GCF works that way.
According to Google Cloud Functions are only good for Single Endpoints and can only work as a web server using the functions framework.
Spring has a document that uses the GcfJarLauncher, but that is still in alpha and I can only get it to work for a single endpoint. Any additional functions I put into the code are ignored and every endpoint triggers the same function.
There were some posts here on SO that talked about using Functional Beans to map to multiple functions, but I couldn't fully get it working and my boss isn't interested in that.
I've also read of people putting the endpoint in the request payload and then mapping to the proper function, but we are not interested in doing that either.
TLDR/Conclusion:
Is it even possible to deploy our app to GCF or do we need to use Cloud Run (as Google suggests in my first link)?
I've implemented two lambda's (let's call A and B) behind api gateway. Assume A is called from "outside" and B is being called from outside and also from A.
I've also implemented lambda Authorizer (token-based; cognito) as auth layer. Everything is working as expected.
Is there a way to bypass authorizer process for B, for calls coming from A only?
Tnx
There are multiple possibilities I have explored myself in the past for the exact same issue.
Change the calls to lambda:Invoke
Assuming you're generating some client code for your micro-services, you can create two versions of these clients:
external to call your service via HTTP API
internal to use lambda:Invoke operation straight to your micro-service.
Create a mirrored VPC-private API
This is probably feasible if you're deploying your infrastructure using CDK (or a similar alternative). Essentially, you keep your existing API where it is, and you create another internal version of it that does not have the authorizer. (Note that you may still want some sort of authorization process happening depending on the nature of your project.)
From this point on, you can pass the endpoint of your internal HTTP API to the Lambdas as environment variables and have them call that.
You can find more info about this, here. As a perk you should probably get lower latencies when talking to API Gateway as traffic through the VPC endpoints will only flow through AWS network, instead of going out on the internet and back in.
Move your workloads to ECS
This is perhaps a major change to your project, but one worth mentioning.
You can create true micro-services using ECS. You can run these services in private subnets of your VPC. In order not to have to deal with IP addresses yourself, you can explore multiple options:
have a VPC-internal Route53 Hosted Zone (more on this here). See more on ECS Service Discovery here
create Network Load Balancers in the private subnets of your VPCs and pass their endpoints to your services.
Planning to migrate my PHP APIs to Graphql using Apollo Federation. After a bit of research, I see it is done using the following way:
My questions are:
Is there any better way to create the federated services so it is not a separate layer (1 for each REST API)? Maybe something close to the previous schema stitching approach where all can sit in one place and be stitched together at the end (instead of a specific federated layer for each service).
If this is the recommended way, how do I deploy this infrastructure? From the diagram, does it mean I have 5 instances running to cover all of the services?
Is it recommended to run Gateway and Federated services all inside one instance (from diagram - 3 servers running in one instance)?
Let me know if it helps.
Federated services are great when you want to break the monolithic structure of the non federated implementation of apollo server. It can be designed by incorporating the micro-service best practices. Instead of blindly having one federated service per rest endpoint, you can have federated services based on the functionality the service is suppose to take care. One service can call multiple rest endpoit. This would provide you better control on scaling, securing and managing services at infrastructure level. An example can be as simple as amazon where item browsing hits will be way more than buying transactions. In this case you can have one federated service which provides browsing data where as another one can for managing transactions. Then you can scale one to multiple instance to handle user load and have additional security in place for the one hadnling transactions.
2 & 3. Yes you would need to have deploy all the components separately. I would recomend to have all the services in the same VPC cluster so that you don't have to worry about network layer security. If the services are deployed across multiple clusters, it will be adding handling firewall and https/tls for every request, which would cause unnecssaery delay becuase of network call. Although it would be in milliseconds but can be easily avoided.
I'm using the GraphQL server which is most-often a gateway that bundles all my different microservices created using gqlgen.
I want to make a switch to the TwirpQL server as I was following the official documentation of TwirpQL. I believe it wants me to wrap every microservice inside it so there will be multiple GraphQL layers one for every microservice.
I don't want to expose the other microservices as Graphql, but rather "just" create one entry point for all the requests and later send them to the respected microservice accordingly.
Can anyone please enlighten me on the correct usage?
Thanks.
I have been studying microservices and serverless solutions and am playing with an angular frontend hosted on S3 and Lambda functions that talk to various DynamoDb tables via the API gateway on AWS.
Every example and video I read/watch uses a simple CRUD microservices as part of a simple 'todo' application or similar. My problem is where does the business logic sit? If I'm building a complex application I don't want all my business logic in my frontend Angular application. Or do I? I could build an Application API which in turn calls CRUD microservices but that feels like a monolithic approach.
I appreciate there may not be a definitive answer but can anybody advise a novice on best practice?
There are several best practices I follow in designing Serverless Microservices
Start with only few Microservices (Less the better up front, unless you know exactly how the service separation should be, delaying the decision to split)
Separate your business logic that goes to the API, and use the handler as a controller in MVC to invoke the business logic. (This will also helps to unit test logic without depending on Lambda).
Its not necessary to write only simple CRUD in your API. It depends on your domain and Business Logic required. (But don't build another monolith without separating the code in to different services. Several AWS Service limits will also give you some guide on how much endpoints should be there in a service & etc.)
Apply the design patterns available for Microservices (e.g If you want to sync data bases between each Microservice, use Pub-Sub pattern using SNS, DynamoDB Streams and Lambda)
Use the Angular App to put most of the presentation logic.
Use CloudFront as a proxy and a CDN to avoid CORs.
If you need more information you can refer the following articles I have written on this.
Deploying Angular/React Apps in AWS
Full Stack Serverless Web Apps with AWS
Note: You can use the CloudFormation in Deploying Angular/React Apps in AWS to automate the creation of S3 and CloudFront with best practices.