👁 Image
Grpc.StatusProto 2.80.0

gRPC C# API for error handling with status.proto

This is a protoype NuGet package providing C# and .NET client and server side support for the gRPC richer error model.

This feature is already available in many other implementations including C++, Go, Java and Python.

This package has dependencies on these NuGet packages:

• Google.Api.CommonProtos - to provide the proto implementations used by the richer error model
• Grpc.Core.Api - for API classes such as RpcException

Error handling in gRPC

The standard way for gRPC to report the success or failure of a gRPC call is for a status code to be returned. If a call completes successfully the server returns an OK status to the client, otherwise an error status code is returned with an optional string error message that provides further details about what happened. This is known as the standard error model and is the official gRPC error model supported by all gRPC implementations.

There is another error model known as the richer error model that allows additional error details to be included by the server. These are expressed in protocol buffers messages, and a set of standard error message types is defined to cover most needs. The protobuf binary encoding of this extra error information is provided as trailing metadata in the response.

For more information on the richer error model see the gRPC documentation on error handling, and the Google APIs overview of the error model.

.NET implementation of the richer error model

The error model is defined by the protocol buffers files status.proto and error_details.proto, and the Google.Api.CommonProtos NuGet package that provides the generated .NET classes from these proto files.

The error is encapsulated by an instance of Google.Rpc.Status and returned in the trailing response metadata with well-known key grpc-status-details-bin. Setting and reading this metadata is handled for you when using the methods provided in this package.

Server Side

The server side uses C#'s Object and Collection initializer syntax.

The server returns the additional error information by throwing an RpcException that is created from a Google.Rpc.Status which contains the details of the error.

To add messages to the Details repeated field in Google.Rpc.Status, wrap each one in Any.Pack() - see example below.

The Google.Rpc.Status extension method ToRpcException creates the appropriate RpcException from the status.

Example - creating and throwing a RpcException:

public override Task<HelloReply> SayHello(HelloRequest request, ServerCallContext context)
{
 ArgumentNotNullOrEmpty(request.Name);

 return Task.FromResult(new HelloReply { Message = "Hello " + request.Name });
}

private static void ArgumentNotNullOrEmpty(string value, [CallerArgumentExpression(nameof(value))] string? paramName = null)
{
 if (string.IsNullOrEmpty(value))
 {
 throw new Google.Rpc.Status
 {
 Code = (int)Code.InvalidArgument,
 Message = "Bad request",
 Details =
 {
 Any.Pack(new BadRequest
 {
 FieldViolations =
 {
 new BadRequest.Types.FieldViolation
 {
 Field = paramName,
 Description = "Value is null or empty"
 }
 }
 })
 }
 }.ToRpcException();
 }
}

A note on error codes

Both Grpc.Core.StatusCode and Google.Rpc.Code define enums for a common set of status codes such as NotFound, PermissionDenied, etc. They have the same values and are based on the codes defined in grpc/status.h.

The recommendation is to use the values in Google.Rpc.Code as a convention. This is a must for Google APIs and strongly recommended for third party services. But users can use a different domain of values if they want to and as long as their services are mutually compatible, things will work fine.

In the richer error model the RpcException will contain both a Grpc.Core.Status (for the standard error model) and a Google.Rpc.Status (for the richer error model), each with their own status code. While an application is free to set these to different values we recommend that they are set to the same value to avoid ambiguity.

Passing stack traces from the server to the client

The richer error model defines a standard way of passing stack traces from the server to the client. The DebugInfo message can be populated with stack traces and then it can be included in the Details of the Google.Rpc.Status.

This package includes the extension method ToRpcDebugInfo for System.Exception to help create the DebugInfo message with the details from the exception.

Example:

try
{
 // ...
}
catch (Exception e)
{
 throw new Google.Rpc.Status
 {
 Code = (int)Google.Rpc.Code.Internal,
 Message = "Internal error",
 Details =
 {
 // populate debugInfo from the exception
 Any.Pack(e.ToRpcDebugInfo()),
 // Add any other messages to the details ...
 }
 }.ToRpcException();
}

Client Side

There is an extension method to retrieve a Google.Rpc.Status from the metadata in an RpcException.

Once the Google.Rpc.Status has been retrieved the messages in the Details can be unpacked. There are two ways of doing this:

• calling GetDetail<T>() with one of the expected message types
• iterating over all the messages in the Details using UnpackDetailMessage()

Example - calling GetDetail<T>():

void PrintError(RpcException ex)
{
 // Get the status from the RpcException
 Google.Rpc.Status? rpcStatus = ex.GetRpcStatus(); // Extension method

 if (rpcStatus != null)
 {
 Console.WriteLine($"Google.Rpc Status: Code: {rpcStatus.Code}, Message: {rpcStatus.Message}");

 // Try and get the ErrorInfo from the details
 ErrorInfo? errorInfo = rpcStatus.GetDetail<ErrorInfo>();
 if (errorInfo != null)
 {
 Console.WriteLine($"\tErrorInfo: Reason: {errorInfo.Reason}, Domain: {errorInfo.Domain}");
 foreach (var md in errorInfo.Metadata)
 {
 Console.WriteLine($"\tKey: {md.Key}, Value: {md.Value}");
 }
 }
 // etc, for any other messages expected in the Details ...
 }
}

Example - iterating over all the messages in the Details:

void PrintStatusDetails(RpcException ex)
{
 // Get the status from the RpcException
 Google.Rpc.Status? rpcStatus = ex.GetRpcStatus(); // Extension method

 if (rpcStatus != null)
 {
 // Decode each message item in the details in turn
 foreach (var msg in rpcStatus.UnpackDetailMessages())
 {
 switch (msg)
 {
 case ErrorInfo errorInfo:
 Console.WriteLine($"ErrorInfo: Reason: {errorInfo.Reason}, Domain: {errorInfo.Domain}");
 foreach (var md in errorInfo.Metadata)
 {
 Console.WriteLine($"\tKey: {md.Key}, Value: {md.Value}");
 }
 break;

 case BadRequest badRequest:
 Console.WriteLine("BadRequest:");
 foreach (BadRequest.Types.FieldViolation fv in badRequest.FieldViolations)
 {
 Console.WriteLine($"\tField: {fv.Field}, Description: {fv.Description}");
 }
 break;

 // Other cases handled here ...
 }
 }
 }

Returning errors within gRPC streams

The model described above allows you to return an error status when the gRPC call finishes.

As an extension to the richer error model you may want to allow servers to send back multiple statuses when streaming responses without terminating the call.

One way of doing this is to include a google.rpc.Status message in the definition of the response messages returned by the server. The client should also be aware that it may receive a status in the response.

For example:

service WidgetLookupProvider {
 rpc streamingLookup(stream WidgetReq) returns (stream WidgetRsp) {}
}

message WidgetReq {
 string widget_name = 1;
}

message WidgetRsp {
 oneof message{
 // details when ok
 string widget_details = 1;
 // or error details
 google.rpc.Status status = 2;
 } 
}

Note: the status.proto and error_details.proto files are provided in the Google.Api.CommonProtos NuGet package.

Example server code fragment:

await foreach (var request in requestStream.ReadAllAsync())
{
 var response = new WidgetRsp();

 // ... process the request ...

 // to return an error
 if (error)
 {
 response.Status = new Google.Rpc.Status { /* ... */ };
 }
 else
 {
 response.WidgetDetails = "the details";
 }
}

Example client code fragment:

// reading the responses
var responseReaderTask = Task.Run(async () =>
{
 await foreach (var rsp in call.ResponseStream.ReadAllAsync())
 {
 switch (rsp.MessageCase)
 {
 case WidgetRsp.MessageOneofCase.WidgetDetails:
 // ... processes the details ...
 break;
 case WidgetRsp.MessageOneofCase.Status:
 // ... handle the error ...
 break;
 }
 }
});

// sending the requests
foreach (var request in requests)
{
 await call.RequestStream.WriteAsync(request);
}

See also

• gRPC richer error model
• Google.Api.CommonProtos