Attempt at a .NET Api Gateway
This project is aimed at people using .NET running a micro services / service orientated architecture that need a unified point of entry into their system.
In particular I want easy integration with IdentityServer reference and bearer tokens.
We have been unable to find this in my current workplace without having to write our own Javascript middlewares to handle the IdentityServer reference tokens. We would rather use the IdentityServer code that already exists to do this.
Ocelot is a bunch of middlewares in a specific order.
Ocelot manipulates the HttpRequest object into a state specified by its configuration until it reaches a request builder middleware where it creates a HttpRequestMessage object which is used to make a request to a downstream service. The middleware that makes the request is the last thing in the Ocelot pipeline. It does not call the next middleware. The response from the downstream service is stored in a per request scoped repository and retrived as the requests goes back up the Ocelot pipeline. There is a piece of middleware that maps the HttpResponseMessage onto the HttpResponse object and that is returned to the client. That is basically it with a bunch of other features.
Pull requests, issues and commentary welcome! No special process just create a request and get in touch either via gitter or create an issue.
Ocelot is designed to work with ASP.NET core only and is currently built to netcoreapp1.1 this documentation may prove helpful when working out if Ocelot would be suitable for you.
Install Ocelot and it's dependecies using nuget. At the moment all we have is the pre version. Once we have something working in a half decent way we will drop a version.
Install-Package Ocelot
All versions can be found here
An example configuration can be found here and an explained configuration can be found here. More detailed instructions to come on how to configure this.
There are two sections to the configuration. An array of ReRoutes and a GlobalConfiguration. The ReRoutes are the objects that tell Ocelot how to treat an upstream request. The Global configuration is a bit hacky and allows overrides of ReRoute specific settings. It's useful if you don't want to manage lots of ReRoute specific settings.
{
"ReRoutes": [],
"GlobalConfiguration": {}
}More information on how to use these options is below..
An example startup using a json file for configuration can be seen below. Currently this is the only way to get configuration into Ocelot.
public class Startup
{
public Startup(IHostingEnvironment env)
{
var builder = new ConfigurationBuilder()
.SetBasePath(env.ContentRootPath)
.AddJsonFile("appsettings.json", optional: true, reloadOnChange: true)
.AddJsonFile($"appsettings.{env.EnvironmentName}.json", optional: true)
.AddJsonFile("configuration.json")
.AddEnvironmentVariables();
Configuration = builder.Build();
}
public IConfigurationRoot Configuration { get; }
public void ConfigureServices(IServiceCollection services)
{
Action<ConfigurationBuilderCachePart> settings = (x) =>
{
x.WithMicrosoftLogging(log =>
{
log.AddConsole(LogLevel.Debug);
})
.WithDictionaryHandle();
};
services.AddOcelotOutputCaching(settings);
services.AddOcelotFileConfiguration(Configuration);
services.AddOcelot();
}
public void Configure(IApplicationBuilder app, IHostingEnvironment env, ILoggerFactory loggerFactory)
{
loggerFactory.AddConsole(Configuration.GetSection("Logging"));
app.UseOcelot();
}
}This is pretty much all you need to get going.......more to come!
Ocelot's primary functionality is to take incomeing http requests and forward them on to a downstream service. At the moment in the form of another http request (in the future this could be any transport mechanism.).
Ocelot always adds a trailing slash to an UpstreamPathTemplate.
Ocelot's describes the routing of one request to another as a ReRoute. In order to get anything working in Ocelot you need to set up a ReRoute in the configuration.
{
"ReRoutes": [
]
}In order to set up a ReRoute you need to add one to the json array called ReRoutes like the following.
{
"DownstreamPathTemplate": "/api/posts/{postId}",
"DownstreamScheme": "https",
"DownstreamPort": 80,
"DownstreamHost" "localhost"
"UpstreamPathTemplate": "/posts/{postId}",
"UpstreamHttpMethod": "Put"
}The DownstreamPathTemplate,Scheme, Port and Host make the URL that this request will be forwarded to. The UpstreamPathTemplate is the URL that Ocelot will use to identity which DownstreamPathTemplate to use for a given request. Finally the UpstreamHttpMethod is used so Ocelot can distinguish between requests to the same URL and is obviously needed to work :) In Ocelot you can add placeholders for variables to your Templates in the form of {something}. The placeholder needs to be in both the DownstreamPathTemplate and UpstreamPathTemplate. If it is Ocelot will attempt to replace the placeholder with the correct variable value from the Upstream URL when the request comes in.
At the moment without any configuration Ocelot will default to all ReRoutes being case insensitive. In order to change this you can specify on a per ReRoute basis the following setting.
"ReRouteIsCaseSensitive": trueThis means that when Ocelot tries to match the incoming upstream url with an upstream template the evaluation will be case sensitive. This setting defaults to false so only set it if you want the ReRoute to be case sensitive is my advice!
Ocelot allows you to specify a service discovery provider and will use this to find the host and port for the downstream service Ocelot is forwarding a request to. At the moment this is only supported in the GlobalConfiguration section which means the same service discovery provider will be used for all ReRoutes you specify a ServiceName for at ReRoute level.
In the future we can add a feature that allows ReRoute specfic configuration.
At the moment the only supported service discovery provider is Consul. The following is required in the GlobalConfiguration. The Provider is required and if you do not specify a host and port the Consul default will be used.
"ServiceDiscoveryProvider": {
"Provider":"Consul",
"Host":"localhost",
"Port":8500
}In order to tell Ocelot a ReRoute is to use the service discovery provider for its host and port you must add the ServiceName and load balancer you wish to use when making requests downstream. At the moment Ocelot has a RoundRobin and LeastConnection algorithm you can use. If no load balancer is specified Ocelot will not load balance requests.
{
"DownstreamPathTemplate": "/api/posts/{postId}",
"DownstreamScheme": "https",
"UpstreamPathTemplate": "/posts/{postId}",
"UpstreamHttpMethod": "Put",
"ServiceName": "product",
"LoadBalancer": "LeastConnection"
}When this is set up Ocelot will lookup the downstream host and port from the service discover provider and load balancer requests across any available services.
Ocelot currently supports the use of bearer tokens with Identity Server (more providers to come if required). In order to identity a ReRoute as authenticated it needs the following configuration added.
"AuthenticationOptions": {
"Provider": "IdentityServer",
"ProviderRootUrl": "http://localhost:52888",
"ScopeName": "api",
"AdditionalScopes": [
"openid",
"offline_access"
],
"ScopeSecret": "secret"
}In this example the Provider is specified as IdentityServer. This string is important because it is used to identity the authentication provider (as previously mentioned in the future there might be more providers). Identity server requires that the client talk to it so we need to provide the root url of the IdentityServer as ProviderRootUrl. IdentityServer requires at least one scope and you can also provider additional scopes. Finally if you are using IdentityServer reference tokens you need to provide the scope secret.
Ocelot will use this configuration to build an authentication handler and if authentication is succefull the next middleware will be called else the response is 401 unauthorised.
Ocelot supports claims based authorisation which is run post authentication. This means if you have a route you want to authorise you can add the following to you ReRoute configuration.
"RouteClaimsRequirement": {
"UserType": "registered"
},In this example when the authorisation middleware is called Ocelot will check to see if the user has the claim type UserType and if the value of that claim is registered. If it isn't then the user will not be authorised and the response will be 403 forbidden.
Ocelot allows the user to access claims and transform them into headers, query string parameters and other claims. This is only available once a user has been authenticated.
After the user is authenticated we run the claims to claims transformation middleware. This allows the user to transform claims before the authorisation middleware is called. After the user is authorised first we call the claims to headers middleware and Finally the claims to query strig parameters middleware.
The syntax for performing the transforms is the same for each proces. In the ReRoute configuration a json dictionary is added with a specific name either AddClaimsToRequest, AddHeadersToRequest, AddQueriesToRequest.
Note I'm not a hotshot programmer so have no idea if this syntax is good..
Within this dictionary the entries specify how Ocelot should transform things! The key to the dictionary is going to become the key of either a claim, header or query parameter.
The value of the entry is parsed to logic that will perform the transform. First of all a dictionary accessor is specified e.g. Claims[CustomerId]. This means we want to access the claims and get the CustomerId claim type. Next is a greater than (>) symbol which is just used to split the string. The next entry is either value or value with and indexer. If value is specifed Ocelot will just take the value and add it to the transform. If the value has an indexer Ocelot will look for a delimiter which is provided after another greater than symbol. Ocelot will then split the value on the delimiter and add whatever was at the index requested to the transform.
Below is an example configuration that will transforms claims to claims
"AddClaimsToRequest": {
"UserType": "Claims[sub] > value[0] > |",
"UserId": "Claims[sub] > value[1] > |"
},This shows a transforms where Ocelot looks at the users sub claim and transforms it into UserType and UserId claims. Assuming the sub looks like this "usertypevalue|useridvalue".
Below is an example configuration that will transforms claims to headers
"AddHeadersToRequest": {
"CustomerId": "Claims[sub] > value[1] > |"
},This shows a transform where Ocelot looks at the users sub claim and trasnforms it into a CustomerId header. Assuming the sub looks like this "usertypevalue|useridvalue".
Below is an example configuration that will transforms claims to query string parameters
"AddQueriesToRequest": {
"LocationId": "Claims[LocationId] > value",
},This shows a transform where Ocelot looks at the users LocationId claim and add its as a query string parameter to be forwarded onto the downstream service.
Ocelot supports one QoS capability at the current time. You can set on a per ReRoute basis if you want to use a circuit breaker when making requests to a downstream service. This uses the an awesome .NET library called Polly check them out here.
Add the following section to a ReRoute configuration.
"QoSOptions": {
"ExceptionsAllowedBeforeBreaking":3,
"DurationOfBreak":5,
"TimeoutValue":5000
}You must set a number greater than 0 against ExceptionsAllowedBeforeBreaking for this rule to be implemented. Duration of break is how long the circuit breaker will stay open for after it is tripped. TimeoutValue means ff a request takes more than 5 seconds it will automatically be timed out.
If you do not add a QoS section QoS will not be used.
Ocelot supports a client sending a request id in the form of a header. If set Ocelot will use the requestid for logging as soon as it becomes available in the middleware pipeline. Ocelot will also forward the request id with the specified header to the downstream service. I'm not sure if have this spot on yet in terms of the pipeline order becasue there are a few logs that don't get the users request id at the moment and ocelot just logs not set for request id which sucks. You can still get the framework request id in the logs if you set IncludeScopes true in your logging config. This can then be used to match up later logs that do have an OcelotRequestId.
In order to use the requestid feature in your ReRoute configuration add this setting
"RequestIdKey": "OcRequestId"In this example OcRequestId is the request header that contains the clients request id.
There is also a setting in the GlobalConfiguration section which will override whatever has been set at ReRoute level for the request id. The setting is as fllows.
"RequestIdKey": "OcRequestId",It behaves in exactly the same way as the ReRoute level RequestIdKey settings.
Ocelot supports some very rudimentary caching at the moment provider by the CacheManager project. This is an amazing project that is solving a lot of caching problems. I would reccomend using this package to cache with Ocelot. If you look at the example here you can see how the cache manager is setup and then passed into the Ocelot AddOcelotOutputCaching configuration method. You can use any settings supported by the CacheManager package and just pass them in.
Anyway Ocelot currently supports caching on the URL of the downstream service and setting a TTL in seconds to expire the cache. More to come!
In orde to use caching on a route in your ReRoute configuration add this setting.
"FileCacheOptions": { "TtlSeconds": 15 }In this example ttl seconds is set to 15 which means the cache will expire after 15 seconds.
Warning use with caution. If you are seeing any exceptions or strange behavior in your middleware pipeline and you are using any of the following. Remove them and try again!
When setting up Ocelot in your Startup.cs you can provide some additonal middleware and override middleware. This is done as follos.
var configuration = new OcelotMiddlewareConfiguration
{
PreErrorResponderMiddleware = async (ctx, next) =>
{
await next.Invoke();
}
};
app.UseOcelot(configuration);In the example above the provided function will run before the first piece of Ocelot middleware. This allows a user to supply any behaviours they want before and after the Ocelot pipeline has run. This means you can break everything so use at your own pleasure!
The user can set functions against the following.
-
PreErrorResponderMiddleware - Already explained above.
-
PreAuthenticationMiddleware - This allows the user to run pre authentication logic and then call Ocelot's authentication middleware.
-
AuthenticationMiddleware - This overrides Ocelots authentication middleware.
-
PreAuthorisationMiddleware - This allows the user to run pre authorisation logic and then call Ocelot's authorisation middleware.
-
AuthorisationMiddleware - This overrides Ocelots authorisation middleware.
-
PreQueryStringBuilderMiddleware - This alows the user to manipulate the query string on the http request before it is passed to Ocelots request creator.
Obviously you can just add middleware as normal before the call to app.UseOcelot() It cannot be added after as Ocelot does not call the next middleware.
Ocelot uses the standard logging interfaces ILoggerFactory / ILogger at the moment. This is encapsulated in IOcelotLogger / IOcelotLoggerFactory with an implementation for the standard asp.net core logging stuff at the moment.
There are a bunch of debugging logs in the ocelot middlewares however I think the system probably needs more logging in the code it calls into. Other than the debugging there is a global error handler that should catch any errors thrown and log them as errors.
The reason for not just using bog standard framework logging is that I could not work out how to override the request id that get's logged when setting IncludeScopes to true for logging settings. Nicely onto the next feature.
Ocelot does not support...
-
Chunked Encoding - Ocelot will always get the body size and return Content-Length header. Sorry if this doesn't work for your use case!
-
Fowarding a host header - The host header that you send to Ocelot will not be forwarded to the downstream service. Obviously this would break everything :(
-
The ReRoute configuration object is too large.
-
The base OcelotMiddleware lets you access things that are going to be null and doesnt check the response is OK. I think the fact you can even call stuff that isnt available is annoying. Let alone it be null.
You can see what we are working on here