Why To Use HttpClientFactory In .Net Core

HttpClient instance was used to invoke or consume the external rest API by clients like console applications or web applications.

What Is Socket?:

A Socket is a system communication protocol providing a communication channel over TCP connection.

In programming, terminology socket can be defined as the combination of Url and Port Number to make communication channels.

HttpClient Working Flow:

Let's assume that we have .Net Core application, which will consume an external API using HttpClient. The steps involved in communication are as follows:

1. The user requests our application, in which we need to call an external API to serve results to the user.
2. On receiving user request our application creates an instance of HttpClient.
3. Next HttpClient looks for HttpRequestHandler object in the pool of HttpRequestHandlers. Then HttpClient picks up one of the HttpRequestHandler objects and gives all input information like Url, payload, tokens, etc to make an external API call.
4. Then HttpRequestHandler looks for Port Number on the server system. Then HttpRequestHandler grabs the Port Number and opens a Web Socket to open a communication channel.
5. On successfully establishing the communication channel, the external API response will be delivered to the user who requested our application.

Never Destroy HttpClient Object With C# Using Statement:

One of the most common mistakes or bad way of implementing HttpClient is destroying it by using C# Using statement. The reason behind this bad practice is the HttpClient implement 'IDisposible' interface. Because of 'IDisposible' interface, every developer believed it is good to destroy the HttpClient object to free up the resource but it is an incorrect assumption. One of the main problem caused by HttpClient object destruction is orphan web sockets that leads Port Number issues.

Example of destroying HttpClient object with Using Statement:

Using(var httpClient = new HttpClient())
{
  // code sample
}

Let's understand what happens if we destroy the HttpClient object in upcoming steps. To understand let's assume we have .Net Core application hosted on a server, let's assume we have only 3 port numbers available on the hosting server to create WebSocket(note:- in real server port numbers will be in 1000's).

1. Now in our application, let's assume we have implemented HttpClient to be destroyed once it's usage is finished.
2. Now the first user requested our application, on receiving request application creates a new HttpClient object inside the c# Using Statement(which means HttpClient object destroyed once code executes out of Using statement)
3. HttpClient object picks up a HttpRequestMessageHandler object from its pool.
4. Now HttpRequestMessageHandler object picks '1' port number available in the pool of port number(we assumed total '3' port numbers in the pool).
5. Now using port number and information in HttpRequestMessageHandler opens WebSocket to create a communication channel.
6. WebSocket once creates will not be closed immediately once its job is done. Because sockets are costly resources so they will wait for some time to serve if any new request comes from the application. So that the existing socket will be effectively used for the subsequent requests. But domain should be the same for subsequent requests to serve by the existing socket if the domain different new Websocket will be created.
7. Now let's think our WebSocket created communication channel, then receiving output it will return a response to the application code.
8. Now we got our response once our code execution comes out of C# Using statement, it will call the IDisposible interface which will destroy the HttpClient object.
9. Destroying the HttpClient object means it will also break the communication with the WebSocket and HttpRequestMessageHandler. Now Websocket lost its communication with HttpRequestMessage, but WebSocket still alive as Orphan Websocket this means port number still be used by this Orphan Websocket and this Orphan WebSocket can't be used by any application. So here we can observe Server resources like the port number in a working mode without any purpose or we can say all those resources in useless mode. This leads to resource unavailability for subsequent requests to our application.
10. Similarly, let's assume that same time we got 2 more user requests, now the same process we discussed in the above steps will execute for these 2 requests. So now we have 3 requests from the user to our application means our application creates 3 WebSockets to build a communication channel. This means all these WebSocket will be  Orphan WebSockets. So from this, we can understand we are out resource to create new WebSocket because now we don't have another port number to create 4th WebSocket.
11. Now let's assume we got 4th user request now our application will search for the port number to create a WebSocket, but there no port numbers available at currently for the application because all ports are in use, so application waits for some time for the port number, finally it will throw an error like 'WebSockets Exhausted'(example exception), 'time out exception' etc.
12. This is the main reason we need to avoid the destroying of the HttpClient object.

Solution Provided By HttpClientFactory:

HttpClienFactory will create a static instance of a HttpClient per Domain of the third-party API. So HttpRequestMessageHandler object will be lives a long time by connecting with WebSocket by setting the expiration time. By default expiration, time will 2 minutes for HttpRequestMessageHandler but we can increase that value based on our application requirements. Using this approach we can avoid orphan WebSocket and consumes fewer resources on the server.

Let's understand the steps of HttpClientFactory flow as follows:

1. Let's assume User Requested (R1) to our application.
2. Then in HttpClientFactory create a static instance of HttpClient. The static instance is for single time creation. HttpClientFactory creates a static instance per domain of third party API. All the requests for specific third party API will be handled by that HttpClient instance.
3. Then HttpClient instance looks for the HttpRequestMessageHandler(HRMH1) object in the HttpRequestMessageHandler pool.
4. Now HttpRequestMessageHandler(HRMH1) object contains all information to make API calls like Url, Payload, Token, Header's, etc.
5. HttpRequestMessageHandler(HRMH1) object looks for the port number on the server. Then creates a WebSocket(WS1) that establishes a communication channel.
6. Now let's assume another User Requested(R2) to our application. If (R1) and (R2) aim the same external API then HttpClientFactory provides the HttpClient instance which is created in previous steps. If (R1) and (R2) are requested for different third party API then the HttpClientFactory Provides the new static HttpClient instance. 
7. So in our case let's assume both (R1) and (R2) user requests want the same external API. Now HttpClient instance looks HttpRequestMessageHandler object. For (R1) we know the (HRMH1) object is used. So now HttpClient checks if (HRMH1) is free to serve (R2). If (HRMH1) is free and it has a connection with active WebSocket(WS1) to serve (R2). Here we can observe (WS1) serve both (R1) and (R2) instead of creating new WebSocket(WS2). So in this process, we are efficiently utilizing our server resource. Let's check another case also like (HRMH1) is busy serving (R1) request which means (R1) request still in progress, then HttpClient instance picks up another HttpRequestMessageHander(HRMH2) object. Then (HRMH2) picks up a new port number on the server and creates a new WebSocket(WS2).
8. The similar way all the user requests will be effectively controlled by the HttpClientFactory.

Support Me!
Buy Me A Coffee PayPal Me

Wrapping Up:

Hopefully, I think this article delivered some useful information about the HttpClientFactory in .Net Core. I love to have your feedback, suggestions, and better techniques in the comment section below.

Follow Me:

Facebook

Twitter

Pinterest