Download profile data

This document describes how you can download your profile data to your local system, and how you can programmatically retrieve profile data by using a Go application.

Download profiles by using the Google Cloud console

To download the profile displayed in the flame graph, click Download .

Profiler uses the following naming convention for the downloaded file:

profiler_[SERVICE_NAME]_[PROFILE_TYPE]_[FROM_DATE]_[TO_DATE]_[ZONE]_[VERSION].pb.gz

In this expression:

• SERVICE_NAME contains your Service selection
• PROFILE_TYPE contains your Profile type selection
• FROM_DATE and TO_DATE contains your time-range specifications
• ZONE contains your Zone selection
• VERSION contains your Version selection

Example: profiler_docdemo-service_HEAP_2018-04-22T20_25_31Z_2018-05-22T20_25_31Z_us-east1-c.pb.gz

Download profiles programmatically

Preview

This product or feature is subject to the "Pre-GA Offerings Terms" in the General Service Terms section of the Service Specific Terms. Pre-GA products and features are available "as is" and might have limited support. For more information, see the launch stage descriptions.

To retrieve profile data, use the ListProfiles API method. The following sample Go program demonstrates the use of this API.

The sample program creates a folder in the directory from where it is run, and generates a set of numbered pprof files. Each file has a naming convention similar to profile000042.pb.gz. Each directory contains profile data and a metadata file - metadata.csv, which contains information about the downloaded files.

// Sample export shows how ListProfiles API can be used to download
// existing pprof profiles for a given project from GCP.
package main

import (
	"bytes"
	"context"
	"encoding/csv"
	"encoding/json"
	"flag"
	"fmt"
	"io"
	"log"
	"os"
	"time"

	cloudprofiler "cloud.google.com/go/cloudprofiler/apiv2"
	pb "cloud.google.com/go/cloudprofiler/apiv2/cloudprofilerpb"
	"google.golang.org/api/iterator"
)

var project = flag.String("project", "", "GCP project ID from which profiles should be fetched")
var pageSize = flag.Int("page_size", 100, "Number of profiles fetched per page. Maximum 1000.")
var pageToken = flag.String("page_token", "", "PageToken from a previous ListProfiles call. If empty, the listing will start from the begnning. Invalid page tokens result in error.")
var maxProfiles = flag.Int("max_profiles", 1000, "Maximum number of profiles to fetch across all pages. If this is <= 0, will fetch all available profiles")

const ProfilesDownloadedSuccessfully = "Read max allowed profiles"

// This function reads profiles for a given project and stores them into locally created files.
// The profile metadata gets stored into a 'metdata.csv' file, while the individual pprof files
// are created per profile.
func downloadProfiles(ctx context.Context, w io.Writer, project, pageToken string, pageSize, maxProfiles int) error {
	client, err := cloudprofiler.NewExportClient(ctx)
	if err != nil {
		return err
	}
	defer client.Close()
	log.Printf("Attempting to fetch %v profiles with a pageSize of %v for %v\n", maxProfiles, pageSize, project)

	// Initial request for the ListProfiles API
	request := &pb.ListProfilesRequest{
		Parent:    fmt.Sprintf("projects/%s", project),
		PageSize:  int32(pageSize),
		PageToken: pageToken,
	}

	// create a folder for storing profiles & metadata
	profilesDirName := fmt.Sprintf("profiles_%v", time.Now().Unix())
	if err := os.Mkdir(profilesDirName, 0750); err != nil {
		log.Fatal(err)
	}
	// create a file for storing profile metadata
	metadata, err := os.Create(fmt.Sprintf("%s/metadata.csv", profilesDirName))
	if err != nil {
		return err
	}
	defer metadata.Close()

	writer := csv.NewWriter(metadata)
	defer writer.Flush()

	writer.Write([]string{"File", "Name", "ProfileType", "Target", "Duration", "Labels"})

	profileCount := 0
	// Keep calling ListProfiles API till all profile pages are fetched or max pages reached
	profilesIterator := client.ListProfiles(ctx, request)
	for {
		// Read individual profile - the client will automatically make API calls to fetch next pages
		profile, err := profilesIterator.Next()

		if err == iterator.Done {
			log.Println("Read all available profiles")
			break
		}
		if err != nil {
			return fmt.Errorf("error reading profile from response: %w", err)
		}
		profileCount++

		filename := fmt.Sprintf("%s/profile%06d.pb.gz", profilesDirName, profileCount)
		err = os.WriteFile(filename, profile.ProfileBytes, 0640)

		if err != nil {
			return fmt.Errorf("unable to write file %s: %w", filename, err)
		}
		fmt.Fprintf(w, "deployment target: %v\n", profile.Deployment.Labels)

		labelBytes, err := json.Marshal(profile.Labels)
		if err != nil {
			return err
		}

		err = writer.Write([]string{filename, profile.Name, profile.Deployment.Target, profile.Duration.String(), string(labelBytes)})
		if err != nil {
			return err
		}

		if maxProfiles > 0 && profileCount >= maxProfiles {
			fmt.Fprintf(w, "result: %v", ProfilesDownloadedSuccessfully)
			break
		}

		if profilesIterator.PageInfo().Remaining() == 0 {
			// This signifies that the client will make a new API call internally
			log.Printf("next page token: %v\n", profilesIterator.PageInfo().Token)
		}
	}
	return nil
}

func main() {
	flag.Parse()
	// validate project ID
	if *project == "" {
		log.Fatalf("No project ID provided, please provide the GCP project ID via '-project' flag")
	}
	var writer bytes.Buffer
	if err := downloadProfiles(context.Background(), &writer, *project, *pageToken, *pageSize, *maxProfiles); err != nil {
		log.Fatal(err)
	}
	log.Println("Finished reading all profiles")
}

The sample program accepts the following command line arguments:

• project: The project from which the profiles are retrieved. Required.
• page_size: The maximum number of profiles retrieved per API call. The maximum value of page_size is 1000. When not specified, this field is set to 100.
• page_token: A string token generated by a previous run of the program to resume downloads. Optional.
• max_profiles: The maximum number of profiles to retrieve. If a non-positive integer is provided, then the program attempts to retrieve all profiles.
  Optional.

Run the sample application

Note: To successfully run the program, it must be authenticated to Profiler. Authentication can be set up using Application Default Credentials (ADC). For more information, see Set up ADC for a local development environment.

To run the sample application, do the following:

1. Clone the repository:

   git clone https://github.com/GoogleCloudPlatform/golang-samples.git
   

2. Change to the directory that contains the sample program:

   cd golang-samples/profiler/export
   

3. Run the program after you replace YOUR_GCP_PROJECT with the ID of your Google Cloud project:

   go run main.go -project YOUR_GCP_PROJECT -page_size 1000 -max_profiles 10000
   

The program might take considerable time to complete. The program outputs a token for the next page after retrieving the current page. You can use the token to resume the process if the program is interrupted.

View the downloaded profiles

To read a downloaded file, which is written in the serialized protocol buffer format, use the open source pprof tool.