Python

The topic provides information to instrument your application with Traceable's Python tracing agent.

Traceable's Python tracing agent is an in-app tracing agent. It collects information from your Python application and sends it to Traceable's platform agent. The platform agent sends the information to Traceable's SaaS platform for further analysis. The topic covers deployment, verification, troubleshooting, upgrade, and uninstall information. 

 High-level Python tracing agent deployment architecture

Before you begin

Make a note of the following points before you proceed with the deployment:

• Make sure that Traceable's platform agent is already installed. For more information, see Platform agent.
• Make a note of the address of the platform agent as it is configured as an endpoint in the deployment process.
• Choose a service name by which you would like your Python application to be identified in the Traceable platform. 
• Traceable supports Python 3.6+

Support for request blocking

Traceable provides request blocking support on the following operating system versions. You can install the Traceable Python agent on other operating systems and continue to collect telemetry data. However, request blocking is not supported on any other operating system.

• Debian 10+
• Ubuntu 18.04+
• Amazon Linux 2+
• CentOS 7+
• Alpine 3.9 and later - While using Traceable's Python agent on Alpine, make sure to install two additional system dependencies as mentioned below:
  ActionScriptActionScript

  apk add --no-cache build-base linux-headers

Support Matrix

Deployment process

The Python agent deployment process has three parts:

1. Installation: Install Python tracing agent using pip.
2. Configuration: Configuration can be done using one of the following options:
   • YAML
   • Environment variable
   • Code-based

3. Instrumentation: You can instrument the Python tracing agent using one of the two instrumentation options:
   • Auto instrumentation - Auto Instrumentation option requires zero code for installation.
   • Manual instrumentation - The manual instrumentation option requires minimal code for installation.

Traceable recommends using Auto instrumentation for most cases because it keeps the instrumentation logic and your application logic separate. However, if you want more granular control over how the Python agent gets initialized, you can choose the manual installation method.

Installation

You can install Traceable's Python tracing agent using pip. Enter the following command:

pip install traceable-agent

Configuration

Option 1: YAML

Create a file named config.yaml with a service_name. The python application is identified by the service_name in the Traceable Dashboard. Following is a sample config.yaml file. Change the endpoint configuration to point to your deployed Traceable platform agent. 

# Observability Configuration
service_name: "blocking-python-svc"
reporting:
  endpoint: http://agent.traceableai:4317 # update to point to your deployed traceable-agent
# Blocking Configuration
opa:
  endpoint: http://agent.traceableai:8181 # update to point to your deployed traceable-agent
blocking_config:
  enabled: false
  remote_config:
    endpoint: agent.traceableai:5441 # update to point to your deployed traceable-agent

Initialize an environment variable which points to the location of the configuration file. This environment variable will be automatically detected by the Traceable python agent.

export TA_CONFIG_FILE=./config.yaml

If you have a django app deployed in gunicorn, set the following environment variable:

export DJANGO_SETTINGS_MODULE="<project-name>.settings"

Option 2: Environment variables

Following is a list of environment variables that you can configure.

export TA_SERVICE_NAME="example-service-name"
export TA_REPORTING_ENDPOINT="<http://hostname:4317>"
export TA_REPORTING_TRACE_REPORTER_TYPE="OTLP"
export TA_BLOCKING_CONFIG_ENABLED=true
export TA_BLOCKING_CONFIG_EVALUATE_BODY=true
export TA_BLOCKING_CONFIG_MODSECURITY_ENABLED=TRUE
export TA_OPA_ENDPOINT=http://agent.traceableai:8181
export TA_BLOCKING_CONFIG_REMOTE_CONFIG_ENDPOINT=agent.traceableai:5441

If you have a django app deployed in gunicorn, set the following environment variable:

export DJANGO_SETTINGS_MODULE="<project-name>.settings"

Option 3: Code

Set the following to edit configurations using code. 

from traceableai.agent import Agent

agent = Agent()
with agent.edit_config() as config:
    config["service_name"] = "some new service name"
    config["reporting"]["endpoint"] = "http://localhost:4317"
    config["reporting"]["trace_reporter_type"] = "OTLP"
    config["propagation_formats"] = ["B3"]
    config["blocking_config"]["enabled"] = True

agent.instrument()

See manual instrumentation for further details.

Auto Instrumentation

To auto instrument your application with Traceable's tracing agent, prefix your startup command with traceableai-instrument. For example:

Python

For Python, enter the following command:

traceableai-instrument python app.py

flask CLI

If you are using flask CLI, enter the following command:

traceableai-instrument flask run

If you are using a config.yaml, you can set the config file as follows:

TA_CONFIG_FILE=./config.yaml traceableai-instrument flask run

gunicorn

If you are using gunicorn, you can prepend your gunicorn startup command. For example:

traceableai-instrument gunicorn --workers=2 test:app

Auto instrumentation with pre-fork Web servers

If you are using gunicorn or other pre-fork Web servers, then as per a known limitation of a deadlock, Python 3.6 spans are not exported. Make a note of the following:

• If you are using Python 3.7 or later, forked Web server is also instrumented automatically.
• If you are using Python 3.6, following is a gunicorn example. Add the following code snippet to gunicorn configuration file, or create one if you are not already using a configuration file.
  ActionScript

  # gunicorn_config.py
  import traceableai
  
  def post_fork(server, worker):
      traceableai.post_fork(server, worker)
  
  

  Next, reference the configuration file in your gunicorn launch command, for example:
  ActionScript

  traceableai-instrument gunicorn -w 5 -b 0.0.0.0:5000 wsgi:app -c gunicorn_config.py

Manual Instrumentation

To use manual instrumentation, you have to make some modifications to your application's entry file. For example:

from traceableai.agent import Agent

...

agent = Agent() # initialize the agent
# .instrument() will instrument all available libraries
agent.instrument()

# if you are using flask, you must pass your app instance to the instrument call
agent.instrument(your_flask_app)
agent.add_traceable_filter() # Adds the traceableai blocking filter
...

Startup

Start your application by entering the following command:

python app.py

 Or, if you are using the flask CLI, then enter the following command:

flask run

Troubleshooting

If you do not see any data in Traceable's platform after a few minutes:

1. Make sure that there is load on the instrumented Python application. 
2. Check whether Traceable's platform agent has the correct agent token configured. You can copy the agent token by logging into your Traceable platform and navigating to Administration() > Account > Agent token.
3. Check the Traceable platform agent logs for any connectivity issues or errors. You can view the logs by navigating to /var/traceable/log/.
4. Check the Python tracing agent logs for any connectivity issues or errors. You will have to change the log level to DEBUG by setting, ENV TA_LOG_LEVEL DEBUG, and restarting the application. For example, in the case of a K8s application, you can configure this in the dockerfile.

If you need further assistance, reach out to Traceable support.

 Upgrade

To upgrade to the latest version of Traceable's tracing agent, enter the following command. Make sure that the current configuration files and environment variables are not changed before you run the upgrade command:

pip install traceable-agent --upgrade

Uninstall

To uninstall the tracing agent, enter the following command:

pip uninstall traceable-agent