Profiling Java applications

This page describes how to modify your Java application to capture profiling data and have that data sent to your Google Cloud project. For general information about profiling, see Profiling concepts.

Profile types for Java:

  • CPU time
  • Heap (requires Java 11 or App Engine standard environment, disabled by default)
  • Wall time (not available for Java 8 App Engine standard environment)

Supported Java language versions:

  • HotSpot-based JVMs (including Oracle JDK and some OpenJDK builds) for Java 8, 11 or later.

Supported profiling agent versions:

  • The most recent release of the agent is supported. In general, releases older than one year aren't supported. We recommend that you use the most recently released version of the agent.

Supported operating systems:

  • Linux. Profiling Java applications is supported for Linux kernels whose standard C library is implemented with glibc or with musl. For configuration information specific to Linux Alpine kernels, see Running on Linux Alpine.

Supported environments:

Enabling the Profiler API

Before you use the profiling agent, ensure that the underlying Profiler API is enabled. You can check the status of the API and enable it if necessary by using either the Google Cloud CLI or the Google Cloud console:

gcloud CLI

  1. If you have not already installed the Google Cloud CLI on your workstation, see the Google Cloud CLI documentation.

  2. Run the following command:

    gcloud services enable cloudprofiler.googleapis.com
    

For more information, see gcloud services.

Google Cloud console

  1. In the navigation panel of the Google Cloud console, select APIs & Services, click Enable APIs and Services, and then enable the Cloud Profiler API:

    Go to Profiler API settings

  2. If API enabled is displayed, then the API is already enabled. If not, click the Enable button.

Installing the Profiler agent

Compute Engine

  1. Create an installation directory, for example, /opt/cprof, for the Profiler agent:

     sudo mkdir -p /opt/cprof

  2. Download the agent archive from the storage.googleapis.com repository and extract it into the installation directory:

    wget -q -O- https://storage.googleapis.com/cloud-profiler/java/latest/profiler_java_agent.tar.gz \
    | sudo tar xzv -C /opt/cprof

GKE

Modify the Dockerfile to create an installation directory for the Profiler agent, download the agent archive, and then extract the archive into the installation directory.

Linux (glibc-based C library):

Use the following installation command:

RUN mkdir -p /opt/cprof && \
  wget -q -O- https://storage.googleapis.com/cloud-profiler/java/latest/profiler_java_agent.tar.gz \
  | tar xzv -C /opt/cprof

Linux Alpine (musl-based C library):

Use the following installation command:

wget -q -O- https://storage.googleapis.com/cloud-profiler/java/latest/profiler_java_agent_alpine.tar.gz \
| tar xzv -C /opt/cprof

Flexible Environment

When you use the Google Java 8 runtime base image or the Java 9 / Jetty 9 runtime base image, the Profiler agent is pre-installed, so there are no additional steps that you need to take to install the agent.

For all other base images, you need to install the agent. For example, the following Dockerfile contains the instructions to use the openjdk:11-slim image, to install the Profiler agent, and it defines the default parameters to be used when starting the application:

FROM openjdk:11-slim

COPY . .
RUN  apt-get update \
     && apt-get install wget \
     && rm -rf /var/lib/apt/lists/*

RUN mkdir -p /opt/cprof && \
    wget -q -O- https://storage.googleapis.com/cloud-profiler/java/latest/profiler_java_agent.tar.gz \
    | tar xzv -C /opt/cprof

CMD ["java", "-agentpath:/opt/cprof/profiler_java_agent.so=OPTION1,OPTION2", "-jar", "PATH_TO_YOUR_JAR_FILE"]

To use this Dockerfile with App Engine flexible environment, you need to do the following:

  • Replace OPTION1 and OPTION2 with the agent configuration values needed for your application, and replace PATH_TO_YOUR_JAR_FILE with the path to your jar file.
  • Place the Dockerfile in the same directory as your app.yaml file.
  • Modify your app.yaml file to specify a custom runtime. For more information, see Building Custom Runtimes.

Standard Environment

When you use the Java runtime environment, the Profiler agent is pre-installed, so there are no additional steps that you need to take to install the agent. For Java version 11 and later, it is pre-installed in /opt/cprof.

Outside Google Cloud

  1. Create an installation directory, for example, /opt/cprof, for the Profiler agent:

     sudo mkdir -p /opt/cprof

  2. Download the agent archive from the storage.googleapis.com repository and extract it into the installation directory:

    wget -q -O- https://storage.googleapis.com/cloud-profiler/java/latest/profiler_java_agent.tar.gz \
    | sudo tar xzv -C /opt/cprof

To list all versions of the agent available for downloading, run the following command:

gsutil ls gs://cloud-profiler/java/cloud-profiler-*

The command response is similar to the following:

gs://cloud-profiler/java/cloud-profiler-java-agent_20191014_RC00.tar.gz
gs://cloud-profiler/java/cloud-profiler-java-agent_20191021_RC00.tar.gz
gs://cloud-profiler/java/cloud-profiler-java-agent_20191028_RC00.tar.gz

To download a specific version of the agent, pass its URL to the download command. For example, to download the agent built on 28 October 2019, you would use the following statement:

wget -q -O- https://storage.googleapis.com/cloud-profiler/java/cloud-profiler-java-agent_20191028_RC00.tar.gz \
  | sudo tar xzv -C /opt/cprof

The version of the agent is logged during its initialization.

Loading the Profiler agent

To profile your application, start Java as you normally would to run your program, but specify the agent-configuration options. You specify the path to the agent library, and you can pass options to the library.

For the App Engine standard environment, the agent is automatically loaded and configured. Skip ahead to Starting your program, for details on configuring, and starting, your program.

Agent configuration

To configure the profiling agent, include the -agentpath flag when starting your application:

 -agentpath:INSTALL_DIR/profiler_java_agent.so=OPTION1,OPTION2,OPTION3

In this expression, INSTALL_DIR is the path to the profiling agent, while OPTION1, OPTION2, and OPTION3 are agent configuration options. For example, if you replace OPTION1 with -cprof_service=myapp in the previous expression, then you set the service name to myapp. There is no restriction on the number of options or their ordering. Supported configuration options are listed in the following table:

Agent option Description
-cprof_service If your application isn't running on App Engine, then you must use this configuration option to set the service name. For service name restrictions, see Service name and version arguments.
-cprof_service_version When you want the ability to analyze profiling data using the Profiler UI by the version of the service, use this option to set the version. For version restrictions, see Service name and version arguments.
-cprof_project_id When you are running outside of Google Cloud, use this option to specify your Google Cloud project ID. For more information, see Profiling applications running outside of Google Cloud.
-cprof_zone_name When your application is running on Google Cloud, the profiling agent determines the zone by communicating with the Compute Engine metadata service. If the profiling agent can't communicate with the metadata service, then you need to use this option.
-cprof_gce_metadata_server_retry_count
-cprof_gce_metadata_server_retry_sleep_sec
Together, these two options define the retry policy that the profiler agent uses when it communicates with the Compute Engine metadata service. to gather your Google Cloud project ID and zone information.

Default policy is to retry up to 3 times waiting 1 second between attempts. This policy is sufficient for most configurations.
-cprof_cpu_use_per_thread_timers For the most accurate CPU time profiles, set this option to true. Use of this option results in increased per-thread overhead.

Default value is false.
-cprof_force_debug_non_safepoints By default, the profiling agent forces JVM to generate debugging information for all just in time (JIT) generated code, in addition to generating debug information for all safepoints. This results in the most accurate function and line-level location information for CPU time and heap profiles at the expense of additional agent overhead. You can disable the generation of debugging information for JIT code by setting this option to false.

Default value is true.
-cprof_wall_num_threads_cutoff By default, wall profiles aren't collected if the total number of threads in the application exceeds 4096. The limit ensures that during profile collection, the cost of traversing the stack of threads is minimal. If your service normally has more than 4096 threads and if you want to collect profiling data at the expense of additional overhead, use this flag to increase limit.

Default limit is 4096 threads.
-cprof_enable_heap_sampling To enable heap profiling for Java 11 and higher, set
-cprof_enable_heap_sampling=true. Heap profiling isn't supported for Java 10 and lower.

Heap profiling is disabled by default.

When you enable heap profiling, the sampling interval is set to 512 KiB by default. This interval is sufficient for most applications and incurs less than 0.5% overhead for the application. Sampling intervals from 256 KiB (262144) to 1024 KiB (1048576) are supported. For example, to set the sampling interval to 256 KiB, which doubles the sampling rate, add the agent option:
-cprof_heap_sampling_interval=262144
Similarly, to set the sampling interval to 1024 KiB, which halves the sampling rate, add the agent option:
-cprof_heap_sampling_interval=1048576
If you enable this profile type, then specify a new service version when you deploy your application. For more information, see Why don't I have data for a specific profile type?

Service name and version arguments

When you load the Profiler agent, you specify a service-name argument and an optional service-version argument to configure it.

The service name lets Profiler collect profiling data for all replicas of that service. The profiler service ensures a collection rate of one profile per minute, on average, for each service name across each combination service versions and zones.

For example, if you have a service with two versions running across replicas in three zones, the profiler will create an average of 6 profiles per minute for that service.

If you use different service names for your replicas, then your service will be profiled more often than necessary, with a correspondingly higher overhead.

When selecting a service name:

  • Choose a name that clearly represents the service in your application architecture. The choice of service name is less important if you only run a single service or application. It is more important if your application runs as a set of micro-services, for example.

  • Make sure to not use any process-specific values, like a process ID, in the service-name string.

  • The service-name string must match this regular expression:

    ^[a-z0-9]([-a-z0-9_.]{0,253}[a-z0-9])?$

A good guideline is to use a static string like imageproc-service as the service name.

The service version is optional. If you specify the service version, Profiler can aggregate profiling information from multiple instances and display it correctly. It can be used to mark different versions of your services as they get deployed. The Profiler UI lets you filter the data by service version; this way, you can compare the performance of older and newer versions of the code.

The value of the service-version argument is a free-form string, but values for this argument typically look like version numbers, for example, 1.0.0 or 2.1.2.

Starting your program

Compute Engine

Start Java as you normally would to run your program, and add the the agent-configuration options:

java \
    -agentpath:/opt/cprof/profiler_java_agent.so=-cprof_service=myapp,-cprof_service_version=1.0.0 \
    JAVA_OPTIONS -jar PATH_TO_YOUR_JAR_FILE PROGRAM_OPTIONS

GKE

Modify the service container Dockerfile to start Java as you normally would to run your program, and add the agent-configuration options:

CMD ["java", \
    "-agentpath:/opt/cprof/profiler_java_agent.so=-cprof_service=myapp,-cprof_service_version=1.0.0", \
     "-jar", "PATH_TO_YOUR_JAR_FILE" ]
    

Flexible Environment

Modify the app.yaml configuration file to set the PROFILER_ENABLE environment variable. Then start your program as usual:

env_variables:
   PROFILER_ENABLE: true

See Defining environment variables for more information.

Standard Environment

Java 21 runtime environment

If you don't use legacy bundled services, enable the profiler collection by modifying the app.yaml file to specify the agentpath flag using either of the following methods:

  • Set the JAVA_TOOL_OPTIONS environment variable:

    runtime: java21
    env_variables:
      JAVA_TOOL_OPTIONS: "-agentpath:/opt/cprof/profiler_java_agent.so=-logtostderr,-cprof_enable_heap_sampling=true"
    
  • Specify the agentpath using the entrypoint element:

    runtime: java21
    entrypoint: java \
      -agentpath:/opt/cprof/profiler_java_agent.so=-logtostderr,-cprof_enable_heap_sampling=true \
      Main.java
    

If you use legacy bundled services, enable the profiler collection by modifying the appengine-web.xml file to specify the agentpath flag using either of the following methods:

  • Set the JAVA_USER_OPTS environment variable:

    <?xml version="1.0" encoding="utf-8"?>
    <appengine-web-app xmlns="http://appengine.google.com/ns/1.0">
    <env-variables>
    <env-var name="JAVA_USER_OPTS" value="-agentpath:/opt/cprof/profiler_java_agent.so=-logtostderr,-cprof_enable_heap_sampling=true" />
    </env-variables>
    </appengine-web-app>
    
  • Set the CPROF_ENABLE environment variable:

    <?xml version="1.0" encoding="utf-8"?>
    <appengine-web-app xmlns="http://appengine.google.com/ns/1.0">
    <env-variables>
    <env-var name="CPROF_ENABLE" value="-agentpath:/opt/cprof/profiler_java_agent.so=-logtostderr,-cprof_enable_heap_sampling=true" />
    </env-variables>
    </appengine-web-app>
    
  • Specify the agentpath using the entrypoint element:

    <?xml version="1.0" encoding="utf-8"?>
    <appengine-web-app xmlns="http://appengine.google.com/ns/1.0">
      <entrypoint>
       java
       -agentpath:/opt/cprof/profiler_java_agent.so=-logtostderr,-cprof_enable_heap_sampling=true
      </entrypoint>
    </appengine-web-app>
    

If a new profile type is configured for collection, ensure that you specify a new service version when you deploy your application. For more information, see Why don't I have data for a specific profile type?

Agent logging

The profiling agent can report logging information for App Engine flexible environment, Compute Engine, and GKE. The profiling agent supports the following logging levels:

  • 0: Log all messages. Default logging level.
  • 1: Log warning, error, and fatal messages.
  • 2: Log error and fatal messages.
  • 3: Log only fatal messages and stop the application.

To enable writing logs to standard error with the default logging level, append -logtostderr to the -agentpath configuration.

To set the logging level to log only error and fatal messages, append -minloglevel=2 to the -agentpath configuration.

For example, to enable logging of error and fatal messages to standard error, append -logtostderr and ‑minloglevel=2 to the -agentpath configuration:

 java -agentpath:/opt/cprof/profiler_java_agent.so=-cprof_service=myapp,-logtostderr,-minloglevel=2 \
   -jar myApp.jar

Troubleshooting

This section lists issues that are specific to profiling Java applications. See Troubleshooting for help with common problems.

Behavior Cause Solution
You've enabled multiple heap profilers and have no profile data. Simultaneous use of multiple heap profilers disables all heap profiling support for Java. This is a JVM limitation. Enable 1 profiler.

Running with Linux Alpine

The Java profiling agent for Linux Alpine is supported only for Google Kubernetes Engine configurations.

To install the latest Java profiling agent for Linux Alpine, see Installing the Profiler agent.

Authentication error

If you use Docker images that run with Linux Alpine (such as golang:alpine or just alpine), you might see the following authentication error:

connection error: desc = "transport: authentication handshake failed: x509: failed to load system roots and no roots provided"

Note that to see the error you must have agent logging enabled.

The error indicates that the Docker images with Linux Alpine don't have the root SSL certificates installed by default. Those certificates are necessary for the profiling agent to communicate with the profiler API. To resolve this error, add the following apk command to your Dockerfile:

FROM alpine
...
RUN apk add --no-cache ca-certificates

You then need to rebuild and redeploy your application.

What's next