Report runtime function errors (1st gen)

You should handle and report runtime errors that occur in Cloud Run functions. Uncaught exceptions or executions that crash the process can result in cold starts, which you should generally try to minimize.

The recommended way for a function to signal an error depends on the function type:

  • HTTP functions should return appropriate HTTP status codes which denote an error. See HTTP Functions for more information.

  • Event-driven functions should log and return an error message. See Write event-driven functions for more information.

If errors are appropriately handled, then function instances that encounter errors can remain active and available to serve requests.

Emit errors to Error Reporting

You can emit an error from a Cloud Run function to Error Reporting as shown in the following:

Node.js 

// These WILL be reported to Error Reporting
throw new Error('I failed you'); // Will cause a cold start if not caught

Python 

@functions_framework.http
def hello_error_1(request):
    # This WILL be reported to Error Reporting,
    # and WILL NOT show up in logs or
    # terminate the function.
    from google.cloud import error_reporting

    client = error_reporting.Client()

    try:
        raise RuntimeError("I failed you")
    except RuntimeError:
        client.report_exception()

    # This WILL be reported to Error Reporting,
    # and WILL terminate the function
    raise RuntimeError("I failed you")


@functions_framework.http
def hello_error_2(request):
    # These errors WILL NOT be reported to Error
    # Reporting, but will show up in logs.
    import logging
    import sys

    print(RuntimeError("I failed you (print to stdout)"))
    logging.warning(RuntimeError("I failed you (logging.warning)"))
    logging.error(RuntimeError("I failed you (logging.error)"))
    sys.stderr.write("I failed you (sys.stderr.write)\n")

    # This is considered a successful execution and WILL NOT be reported
    # to Error Reporting, but the status code (500) WILL be logged.
    from flask import abort

    return abort(500)

Go 


package tips

import (
	"fmt"
	"net/http"
	"os"

	"github.com/GoogleCloudPlatform/functions-framework-go/functions"
)

func init() {
	functions.HTTP("HTTPError", HTTPError)
}

// HTTPError describes how errors are handled in an HTTP function.
func HTTPError(w http.ResponseWriter, r *http.Request) {
	// An error response code is NOT reported to Error Reporting.
	// http.Error(w, "An error occurred", http.StatusInternalServerError)

	// Printing to stdout and stderr is NOT reported to Error Reporting.
	fmt.Println("An error occurred (stdout)")
	fmt.Fprintln(os.Stderr, "An error occurred (stderr)")

	// Calling log.Fatal sets a non-zero exit code and is NOT reported to Error
	// Reporting.
	// log.Fatal("An error occurred (log.Fatal)")

	// Panics are reported to Error Reporting.
	panic("An error occurred (panic)")
}

Java 


import com.google.cloud.functions.HttpFunction;
import com.google.cloud.functions.HttpRequest;
import com.google.cloud.functions.HttpResponse;
import java.io.IOException;
import java.util.logging.Logger;

public class HelloError implements HttpFunction {

  private static final Logger logger = Logger.getLogger(HelloError.class.getName());

  @Override
  public void service(HttpRequest request, HttpResponse response)
      throws IOException {
    // These will NOT be reported to Error Reporting
    System.err.println("I failed you");
    logger.severe("I failed you");

    // This WILL be reported to Error Reporting
    throw new RuntimeException("I failed you");
  }
}

If you would like more fine-grained error reporting, you can use the Error Reporting client libraries.

You can view the reported errors in Error Reporting in the Google Cloud console. You can also see the errors reported from a particular function when you select it from the list of functions in the Google Cloud console.

Uncaught exceptions produced by your function will appear in Error Reporting. Note that some types of uncaught exceptions (such as those thrown asynchronously) will cause a cold start to occur upon a future function invocation. This increases the amount of time your function will take to run.