Configuration

To adapt the Elastic APM agent to your needs, configure it using environment variables or framework-specific configuration.

You can either configure the agent by setting environment variables:

ELASTIC_APM_SERVICE_NAME=foo python manage.py runserver

or with inline configuration:

apm_client = Client(service_name="foo")

or by using framework specific configuration e.g. in your Django settings.py file:

ELASTIC_APM = {
    "SERVICE_NAME": "foo",
}

The precedence is as follows:

• Central configuration (supported options are marked with )
• Environment variables
• Inline configuration
• Framework-specific configuration
• Default value

Configuration options marked with the badge can be changed at runtime when set from a supported source.

The Python Agent supports Central configuration, which allows you to fine-tune certain configurations from in the APM app. This feature is enabled in the Agent by default with central_config.

To configure Django, add an ELASTIC_APM dictionary to your settings.py:

ELASTIC_APM = {
    'SERVICE_NAME': 'my-app',
    'SECRET_TOKEN': 'changeme',
}

To configure Flask, add an ELASTIC_APM dictionary to your app.config:

app.config['ELASTIC_APM'] = {
    'SERVICE_NAME': 'my-app',
    'SECRET_TOKEN': 'changeme',
}

apm = ElasticAPM(app)

The name of your service. This is used to keep all the errors and transactions of your service together and is the primary filter in the Elastic APM user interface.

While a default is provided, it is essential that you override this default with something more descriptive and unique across your infrastructure.

The URL for your APM Server. The URL must be fully qualified, including protocol (http or https) and port. Note: Do not set this if you are using APM in an AWS lambda function. APM Agents are designed to proxy their calls to the APM Server through the lambda extension. Instead, set ELASTIC_APM_LAMBDA_APM_SERVER. For more info, see AWS Lambda.

Enable or disable the agent. When set to false, the agent will not collect any data or start any background threads.

Enable or disable recording of events. If set to false, then the Python agent does not send any events to the Elastic APM server, and instrumentation overhead is minimized. The agent will continue to poll the server for configuration changes.

The logging.logLevel at which the elasticapm logger will log. The available options are:

• "off" (sets logging.logLevel to 1000)
• "critical"
• "error"
• "warning"
• "info"
• "debug"
• "trace" (sets logging.log_level to 5)

Options are case-insensitive

Note that this option doesn’t do anything with logging handlers. In order for any logs to be visible, you must either configure a handler (logging.basicConfig will do this for you) or set log_file. This will also override any log level your app has set for the elasticapm logger.

This enables the agent to log to a file. This is disabled by default. The agent will log at the logging.logLevel configured with log_level. Use log_file_size to configure the maximum size of the log file. This log file will automatically rotate.

Note that setting log_level is required for this setting to do anything.

If ecs_logging is installed, the logs will automatically be formatted as ecs-compatible json.

The size of the log file if log_file is set.

The agent always keeps one backup file when rotating, so the maximum space that the log files will consume is twice the value of this setting.

Valid options:

• "off"
• "override"

If ecs_logging is installed, setting this to "override" will cause the agent to automatically attempt to enable ecs-formatted logging.

For base logging from the standard library, the agent will get the root logger, find any attached handlers, and for each, set the formatter to ecs_logging.StdlibFormatter().

If structlog is installed, the agent will override any configured processors with ecs_logging.StructlogFormatter().

Note that this is a very blunt instrument that could have unintended side effects. If problems arise, please apply these formatters manually and leave this setting as "off". See the ecs_logging docs for more information about using these formatters.

Also note that this setting does not facilitate shipping logs to Elasticsearch. We recommend Filebeat for that purpose.

The transport class to use when sending events to the APM Server.

The name of the given service node. This is optional and if omitted, the APM Server will fall back on system.container.id if available, and host.name if necessary.

This option allows you to set the node name manually to ensure it is unique and meaningful.

The name of the environment this service is deployed in, e.g. "production" or "staging".

Environments allow you to easily filter data on a global level in the APM app. It’s important to be consistent when naming environments across agents. See environment selector in the APM app for more information.

This config value allows you to specify which cloud provider should be assumed for metadata collection. By default, the agent will attempt to detect the cloud provider or, if that fails, will use trial and error to collect the metadata.

Valid options are "auto", "aws", "gcp", and "azure". If this config value is set to "none", then no cloud metadata will be collected.

This string is used to ensure that only your agents can send data to your APM Server. Both the agents and the APM Server have to be configured with the same secret token. An example to generate a secure secret token is:

python -c "import secrets; print(secrets.token_urlsafe(32))"

This base64-encoded string is used to ensure that only your agents can send data to your APM Server. The API key must be created using the APM server command-line tool.

A version string for the currently deployed version of the service. If youre deploys are not versioned, the recommended value for this field is the commit identifier of the deployed revision, e.g. the output of git rev-parse HEAD.

The name of the used framework. For Django and Flask, this defaults to django and flask respectively, otherwise, the default is None.

The version number of the used framework. For Django and Flask, this defaults to the used version of the framework, otherwise, the default is None.

A list of exception types to be filtered. Exceptions of these types will not be sent to the APM Server.

A list of URLs for which the agent should not capture any transaction data.

Optionally, * can be used to match multiple URLs at once.

A list of regular expressions. Transactions with a name that matches any of the configured patterns will be ignored and not sent to the APM Server.

A timeout for requests to the APM Server. The setting has to be provided in duration format. If a request to the APM Server takes longer than the configured timeout, the request is cancelled and the event (exception or transaction) is discarded. Set to None to disable timeouts.

The host name to use when sending error and transaction data to the APM Server.

If set to True (the default), the agent will add a stack trace to each log event, indicating where the log message has been issued.

This setting can be overridden on an individual basis by setting the extra-key stack:

logger.info('something happened', extra={'stack': False})

Possible values: errors, transactions, all, off

The Elastic APM Python agent can collect local variables for stack frames. By default, this is only done for errors.

When collecting local variables, they will be converted to strings. This setting allows you to limit the length of the resulting string.

This setting allows you to limit the length of lists in local variables.

This setting allows you to limit the length of dicts in local variables.

By default, the APM agent collects source code snippets for errors. This setting allows you to modify the number of lines of source code that are being collected.

We differ between errors and spans, as well as library frames and app frames.

For transactions that are HTTP requests, the Python agent can optionally capture the request body (e.g. POST variables).

Possible values: errors, transactions, all, off.

If the request has a body and this setting is disabled, the body will be shown as [REDACTED].

For requests with a content type of multipart/form-data, any uploaded files will be referenced in a special _files key. It contains the name of the field and the name of the uploaded file, if provided.

For transactions and errors that happen due to HTTP requests, the Python agent can optionally capture the request and response headers.

Possible values: true, false

This limits the amount of spans that are recorded per transaction. This is helpful in cases where a transaction creates a very high amount of spans (e.g. thousands of SQL queries). Setting an upper limit will prevent edge cases from overloading the agent and the APM Server.

This limits the number of frames captured for each stack trace.

Setting the limit to 0 will disable stack trace collection, while any positive integer value will be used as the maximum number of frames to collect. To disable the limit and always capture all frames, set the value to -1.

By default, the APM agent collects a stack trace with every recorded span that has a duration equal to or longer than this configured threshold. While stack traces are very helpful to find the exact place in your code from which a span originates, collecting this stack trace does have some overhead. Tune this threshold to ensure that you only collect stack traces for spans that could be problematic.

To collect traces for all spans, regardless of their length, set the value to 0.

To disable stack trace collection for spans completely, set the value to -1.

Except for the special values -1 and 0, this setting should be provided in duration format.

Enable/disable span compression.

If enabled, the agent will compress very short, repeated spans into a single span, which is beneficial for storage and processing requirements. Some information is lost in this process, e.g. exact durations of each compressed span.

Consecutive spans that are exact match and that are under this threshold will be compressed into a single composite span. This reduces the collection, processing, and storage overhead, and removes clutter from the UI. The tradeoff is that the DB statements of all the compressed spans will not be collected.

Two spans are considered exact matches if the following attributes are identical: * span name * span type * span subtype * destination resource (e.g. the Database name)

Consecutive spans to the same destination that are under this threshold will be compressed into a single composite span. This reduces the collection, processing, and storage overhead, and removes clutter from the UI. The tradeoff is that metadata such as database statements of all the compressed spans will not be collected.

Two spans are considered to be of the same kind if the following attributes are identical: * span type * span subtype * destination resource (e.g. the Database name)

Exit spans are spans that represent a call to an external service, like a database. If such calls are very short, they are usually not relevant and can be ignored.

This feature is disabled by default.

The maximum queue length of the request buffer before sending the request to the APM Server. A lower value will increase the load on your APM Server, while a higher value can increase the memory pressure of your app. A higher value also impacts the time until data is indexed and searchable in Elasticsearch.

This setting is useful to limit memory consumption if you experience a sudden spike of traffic. It has to be provided in size format.

The maximum queue time of the request buffer before sending the request to the APM Server. A lower value will increase the load on your APM Server, while a higher value can increase the memory pressure of your app. A higher value also impacts the time until data is indexed and searchable in Elasticsearch.

This setting is useful to limit memory consumption if you experience a sudden spike of traffic. It has to be provided in duration format.

A list of processors to process transactions and errors. For more information, see Sanitizing Data.

A list of glob-matched field names to match and mask when using processors. For more information, see Sanitizing Data.

By default, the agent samples every transaction (e.g. request to your service). To reduce overhead and storage requirements, set the sample rate to a value between 0.0 and 1.0. We still record overall time and the result for unsampled transactions, but no context information, labels, or spans.

A set of paths, optionally using shell globs (see fnmatch for a description of the syntax). These are matched against the absolute filename of every frame, and if a pattern matches, the frame is considered to be an "in-app frame".

include_paths takes precedence over exclude_paths.

A set of paths, optionally using shell globs (see fnmatch for a description of the syntax). These are matched against the absolute filename of every frame, and if a pattern matches, the frame is considered to be a "library frame".

include_paths takes precedence over exclude_paths.

The default value varies based on your Python version and implementation, e.g.:

• PyPy3: ['\*/lib-python/3/*', '\*/site-packages/*']
• CPython 2.7: ['\*/lib/python2.7/*', '\*/lib64/python2.7/*']

If your app is in debug mode (e.g. in Django with settings.DEBUG = True or in Flask with app.debug = True), the agent won’t send any data to the APM Server. You can override it by changing this setting to True.

If set to True, the agent won’t send any events to the APM Server, independent of any debug state.

If set to False, the agent won’t instrument any code. This disables most of the tracing functionality, but can be useful to debug possible instrumentation issues.

By default, the agent verifies the SSL certificate if an HTTPS connection to the APM Server is used. Verification can be disabled by changing this setting to False. This setting is ignored when server_cert is set.

If you have configured your APM Server with a self-signed TLS certificate, or you just wish to pin the server certificate, you can specify the path to the PEM-encoded certificate via the ELASTIC_APM_SERVER_CERT configuration.

By default, the agent will validate the TLS/SSL certificate of the APM Server using the well-known CAs curated by Mozilla, and provided by the certifi package.

You can set this option to the path of a file containing a CA certificate that will be used instead.

Specifying this option is required when using self-signed certificates, unless server certificate validation is disabled.

By default, the Python Agent uses the certifi certificate store. To use Python’s default mechanism for finding certificates, set this option to False.

The interval in which the agent collects metrics. A shorter interval increases the granularity of metrics, but also increases the overhead of the agent, as well as storage requirements.

It has to be provided in duration format.

A comma-separated list of dotted metrics names that should not be sent to the APM Server. You can use * to match multiple metrics; for example, to disable all CPU-related metrics, as well as the "total system memory" metric, set disable_metrics to:

"*.cpu.*,system.memory.total"

Enable or disable the tracking and collection of breakdown metrics. Setting this to False disables the tracking of breakdown metrics, which can reduce the overhead of the agent.

Enable/disable the tracking and collection of metrics from prometheus_client.

See Prometheus metric set (beta) for more information.

A prefix to prepend to Prometheus metrics names.

See Prometheus metric set (beta) for more information.

List of import paths for the MetricSets that should be used to collect metrics.

See Custom Metrics for more information.

When enabled, the agent will make periodic requests to the APM Server to fetch updated configuration.

See Dynamic configuration for more information.

Labels added to all events, with the format key=value[,key=value[,...]]. Any labels set by application via the API will override global labels with the same keys.

By default in python 3, the agent installs a LogRecord factory that automatically adds tracing fields to your log records. Disable this behavior by setting this to True.

To enable distributed tracing, the agent sets a number of HTTP headers to outgoing requests made with instrumented HTTP libraries. These headers (traceparent and tracestate) are defined in the W3C Trace Context specification.

Additionally, when this setting is set to True, the agent will set elasticapm-traceparent for backwards compatibility.

This option allows some control on how the APM agent handles W3C trace-context headers on incoming requests. By default, the traceparent and tracestate headers are used per W3C spec for distributed tracing. However, in certain cases it can be helpful to not use the incoming traceparent header. Some example use cases:

• An Elastic-monitored service is receiving requests with traceparent headers from unmonitored services.
• An Elastic-monitored service is publicly exposed, and does not want tracing data (trace-ids, sampling decisions) to possibly be spoofed by user requests.

Valid values are:

• 'continue': The default behavior. An incoming traceparent value is used to continue the trace and determine the sampling decision.
• 'restart': Always ignores the traceparent header of incoming requests. A new trace-id will be generated and the sampling decision will be made based on transaction_sample_rate. A span link will be made to the incoming traceparent.
• 'restart_external': If an incoming request includes the es vendor flag in tracestate, then any traceparent will be considered internal and will be handled as described for 'continue' above. Otherwise, any 'traceparent' is considered external and will be handled as described for 'restart' above.

Starting with Elastic Observability 8.2, span links will be visible in trace views.

If set to True, the agent will intercept the default sys.excepthook, which allows the agent to collect all uncaught exceptions.

Whether each transaction should have the process arguments attached. Disabled by default to save disk space.

By default, we use the function or class name of the view as the transaction name. Starting with Django 2.2, Django makes the route (e.g. users/<int:user_id>/) available on the request.resolver_match object. If you want to use the route instead of the view name as the transaction name, set this config option to true.

To trace Django requests, the agent uses a middleware, elasticapm.contrib.django.middleware.TracingMiddleware. By default, this middleware is inserted automatically as the first item in settings.MIDDLEWARES. To disable the automatic insertion of the middleware, change this setting to False.

Some environment variables that are not specific to the APM agent can be used to configure the agent.

By using HTTP_PROXY and HTTPS_PROXY, the agent can be instructed to use a proxy to connect to the APM Server. If both are set, HTTPS_PROXY takes precedence.

To instruct the agent to not use a proxy, you can use the NO_PROXY environment variable. You can either set it to a comma-separated list of hosts for which no proxy should be used (e.g. localhost,example.com) or use * to match any host.

This is useful if HTTP_PROXY / HTTPS_PROXY is set for other reasons than agent / APM Server communication.

To tell the agent to use a different SSL certificate, you can use these environment variables. See also OpenSSL docs.

Please note that these variables may apply to other SSL/TLS communication in your service, not just related to the APM agent.

Some options require a unit, either duration or size. These need to be provided in a specific format.

The duration format is used for options like timeouts. The unit is provided as a suffix directly after the number–without any separation by whitespace.

Example: 5ms

Supported units

• us (microseconds)
• ms (milliseconds)
• s (seconds)
• m (minutes)

The size format is used for options like maximum buffer sizes. The unit is provided as suffix directly after the number, without and separation by whitespace.

Example: 10kb

Supported units:

• b (bytes)
• kb (kilobytes)
• mb (megabytes)
• gb (gigabytes)