Error handling

General Overview

The Hyperledger Fabric error handling framework can be found in the source repository under common/errors. It defines a new type of error, CallStackError, to use in place of the standard error type provided by Go.

A CallStackError consists of the following:

  • Component code - a name for the general area of the code that is generating the error. Component codes should consist of three uppercase letters. Numerics and special characters are not allowed. A set of component codes is defined in common/errors/codes.go
  • Reason code - a short code to help identify the reason the error occurred. Reason codes should consist of three numeric values. Letters and special characters are not allowed. A set of reason codes is defined in common/error/codes.go
  • Error code - the component code and reason code separated by a colon, e.g. MSP:404
  • Error message - the text that describes the error. This is the same as the input provided to fmt.Errorf() and Errors.New(). If an error has been wrapped into the current error, its message will be appended.
  • Callstack - the callstack at the time the error is created. If an error has been wrapped into the current error, its error message and callstack will be appended to retain the context of the wrapped error.

The CallStackError interface exposes the following functions:

  • Error() - returns the error message with callstack appended
  • Message() - returns the error message (without callstack appended)
  • GetComponentCode() - returns the 3-character component code
  • GetReasonCode() - returns the 3-digit reason code
  • GetErrorCode() - returns the error code, which is “component:reason”
  • GetStack() - returns just the callstack
  • WrapError(error) - wraps the provided error into the CallStackError

Usage Instructions

The new error handling framework should be used in place of all calls to fmt.Errorf() or Errors.new(). Using this framework will provide error codes to check against as well as the option to generate a callstack that will be appended to the error message.

Using the framework is simple and will only require an easy tweak to your code.

First, you’ll need to import github.com/hyperledger/fabric/common/errors into any file that uses this framework.

Let’s take the following as an example from core/chaincode/chaincode_support.go:

err = fmt.Errorf("Error starting container: %s", err)

For this error, we will simply call the constructor for Error and pass a component code, reason code, followed by the error message. At the end, we then call the WrapError() function, passing along the error itself.

fmt.Errorf("Error starting container: %s", err)

becomes

errors.ErrorWithCallstack("CHA", "505", "Error starting container").WrapError(err)

You could also just leave the message as is without any problems:

errors.ErrorWithCallstack("CHA", "505", "Error starting container: %s", err)

With this usage you will be able to format the error message from the previous error into the new error, but will lose the ability to print the callstack (if the wrapped error is a CallStackError).

A second example to highlight a scenario that involves formatting directives for parameters other than errors, while still wrapping an error, is as follows:

fmt.Errorf("failed to get deployment payload %s - %s", canName, err)

becomes

errors.ErrorWithCallstack("CHA", "506", "Failed to get deployment payload %s", canName).WrapError(err)

Displaying error messages

Once the error has been created using the framework, displaying the error message is as simple as:

logger.Errorf(err)

or

fmt.Println(err)

or

fmt.Printf("%s\n", err)

An example from peer/common/common.go:

errors.ErrorWithCallstack("PER", "404", "Error trying to connect to local peer").WrapError(err)

would display the error message:

PER:404 - Error trying to connect to local peer
Caused by: grpc: timed out when dialing

Note

The callstacks have not been displayed for this example for the sake of brevity.

General guidelines for error handling in Hyperledger Fabric

  • If it is some sort of best effort thing you are doing, you should log the error and ignore it.
  • If you are servicing a user request, you should log the error and return it.
  • If the error comes from elsewhere, you have the choice to wrap the error or not. Typically, it’s best to not wrap the error and simply return it as is. However, for certain cases where a utility function is called, wrapping the error with a new component and reason code can help an end user understand where the error is really occurring without inspecting the callstack.
  • A panic should be handled within the same layer by throwing an internal error code/start a recovery process and should not be allowed to propagate to other packages.