API reference
editAPI reference
editThe Elastic APM Python agent has several public APIs. Most of the public API functionality is not needed when using one of our supported frameworks, but they allow customized usage.
Client API
editThe public Client API consists of several methods on the Client
class.
This API can be used to track exceptions and log messages,
as well as to mark the beginning and end of transactions.
Instantiation
editAdded in v1.0.0.
To create a Client
instance, import it and call its constructor:
from elasticapm import Client client = Client({'SERVICE_NAME': 'example'}, **defaults)
-
config
: A dictionary, with key/value configuration. For the possible configuration keys, see Configuration. -
**defaults
: default values for configuration. These can be omitted in most cases, and take the least precedence.
elasticapm.get_client()
edit[small]#Added in v6.1.0.
Retrieves the Client
singleton. This is useful for many framework integrations,
where the client is instantiated automatically.
client = elasticapm.get_client() client.capture_message('foo')
Errors
editClient.capture_exception()
editAdded in v1.0.0. handled
added in v2.0.0.
Captures an exception object:
try: x = int("five") except ValueError: client.capture_exception()
-
exc_info
: A(type, value, traceback)
tuple as returned bysys.exc_info()
. If not provided, it will be captured automatically. -
date
: Adatetime.datetime
object representing the occurrence time of the error. If left empty, it defaults todatetime.datetime.utcnow()
. -
context
: A dictionary with contextual information. This dictionary must follow the Context schema definition. -
custom
: A dictionary of custom data you want to attach to the event. -
handled
: A boolean to indicate if this exception was handled or not.
Returns the id of the error as a string.
Client.capture_message()
editAdded in v1.0.0.
Captures a message with optional added contextual data. Example:
client.capture_message('Billing process succeeded.')
-
message
: The message as a string. -
param_message
: Alternatively, a parameterized message as a dictionary. The dictionary contains two values:message
, andparams
. This allows the APM Server to group messages together that share the same parameterized message. Example:client.capture_message(param_message={ 'message': 'Billing process for %s succeeded. Amount: %s', 'params': (customer.id, order.total_amount), })
-
stack
: If set toTrue
(the default), a stacktrace from the call site will be captured. -
exc_info
: A(type, value, traceback)
tuple as returned bysys.exc_info()
. If not provided, it will be captured automatically, ifcapture_message()
was called in anexcept
block. -
date
: Adatetime.datetime
object representing the occurrence time of the error. If left empty, it defaults todatetime.datetime.utcnow()
. -
context
: A dictionary with contextual information. This dictionary must follow the Context schema definition. -
custom
: A dictionary of custom data you want to attach to the event.
Returns the id of the message as a string.
Either the message
or the param_message
argument is required.
Transactions
editClient.begin_transaction()
editAdded in v1.0.0. trace_parent
support added in v5.6.0.
Begin tracking a transaction. Should be called e.g. at the beginning of a request or when starting a background task. Example:
client.begin_transaction('processors')
-
transaction_type
: (required) A string describing the type of the transaction, e.g.'request'
or'celery'
. -
trace_parent
: (optional) ATraceParent
object. See TraceParent generation. -
links
: (optional) A list ofTraceParent
objects to which this transaction is causally linked.
Client.end_transaction()
editAdded in v1.0.0.
End tracking the transaction. Should be called e.g. at the end of a request or when ending a background task. Example:
client.end_transaction('myapp.billing_process', processor.status)
-
name
: (optional) A string describing the name of the transaction, e.g.process_order
. This is typically the name of the view/controller that handles the request, or the route name. -
result
: (optional) A string describing the result of the transaction. This is typically the HTTP status code, or e.g.'success'
for a background task.
if name
and result
are not set in the end_transaction()
call,
they have to be set beforehand by calling elasticapm.set_transaction_name()
and elasticapm.set_transaction_result()
during the transaction.
TraceParent
editTransactions can be started with a TraceParent
object. This creates a
transaction that is a child of the TraceParent
, which is essential for
distributed tracing.
elasticapm.trace_parent_from_string()
editAdded in v5.6.0.
Create a TraceParent
object from the string representation generated by
TraceParent.to_string()
:
parent = elasticapm.trace_parent_from_string('00-03d67dcdd62b7c0f7a675424347eee3a-5f0e87be26015733-01') client.begin_transaction('processors', trace_parent=parent)
-
traceparent_string
: (required) A string representation of aTraceParent
object.
elasticapm.trace_parent_from_headers()
editAdded in v5.6.0.
Create a TraceParent
object from HTTP headers (usually generated by another
Elastic APM agent):
parent = elasticapm.trace_parent_from_headers(headers_dict) client.begin_transaction('processors', trace_parent=parent)
-
headers
: (required) HTTP headers formed as a dictionary.
elasticapm.get_trace_parent_header()
editAdded in v5.10.0.
Return the string representation of the current transaction TraceParent
object:
elasticapm.get_trace_parent_header()
Other APIs
editelasticapm.instrument()
editAdded in v1.0.0.
Instruments libraries automatically.
This includes a wide range of standard library and 3rd party modules.
A list of instrumented modules can be found in elasticapm.instrumentation.register
.
This function should be called as early as possibly in the startup of your application.
For supported frameworks, this is called automatically. Example:
import elasticapm elasticapm.instrument()
elasticapm.set_transaction_name()
editAdded in v1.0.0.
Set the name of the current transaction. For supported frameworks, the transaction name is determined automatically, and can be overridden using this function. Example:
import elasticapm elasticapm.set_transaction_name('myapp.billing_process')
-
name
: (required) A string describing name of the transaction -
override
: ifTrue
(the default), overrides any previously set transaction name. IfFalse
, only sets the name if the transaction name hasn’t already been set.
elasticapm.set_transaction_result()
editAdded in v2.2.0.
Set the result of the current transaction. For supported frameworks, the transaction result is determined automatically, and can be overridden using this function. Example:
import elasticapm elasticapm.set_transaction_result('SUCCESS')
-
result
: (required) A string describing the result of the transaction, e.g.HTTP 2xx
orSUCCESS
-
override
: ifTrue
(the default), overrides any previously set result. IfFalse
, only sets the result if the result hasn’t already been set.
elasticapm.set_transaction_outcome()
editAdded in v5.9.0.
Sets the outcome of the transaction. The value can either be "success"
, "failure"
or "unknown"
.
This should only be called at the end of a transaction after the outcome is determined.
The outcome
is used for error rate calculations.
success
denotes that a transaction has concluded successful, while failure
indicates that the transaction failed
to finish successfully.
If the outcome
is set to unknown
, the transaction will not be included in error rate calculations.
For supported web frameworks, the transaction outcome is set automatically if it has not been set yet, based on the
HTTP status code.
A status code below 500
is considered a success
, while any value of 500
or higher is counted as a failure
.
If your transaction results in an HTTP response, you can alternatively provide the HTTP status code.
While the outcome
and result
field look very similar, they serve different purposes.
Other than the result
field, which canhold an arbitrary string value,
outcome
is limited to three different values,
"success"
, "failure"
and "unknown"
.
This allows the APM app to perform error rate calculations on these values.
Example:
import elasticapm elasticapm.set_transaction_outcome("success") # Using an HTTP status code elasticapm.set_transaction_outcome(http_status_code=200) # Using predefined constants: from elasticapm.conf.constants import OUTCOME elasticapm.set_transaction_outcome(OUTCOME.SUCCESS) elasticapm.set_transaction_outcome(OUTCOME.FAILURE) elasticapm.set_transaction_outcome(OUTCOME.UNKNOWN)
-
outcome
: One of"success"
,"failure"
or"unknown"
. Can be omitted ifhttp_status_code
is provided. -
http_status_code
: if the transaction represents an HTTP response, its status code can be provided to determine theoutcome
automatically. -
override
: ifTrue
(the default), any previously setoutcome
will be overriden. IfFalse
, the outcome will only be set if it was not set before.
elasticapm.get_transaction_id()
editAdded in v5.2.0.
Get the id of the current transaction. Example:
import elasticapm transaction_id = elasticapm.get_transaction_id()
elasticapm.get_trace_id()
editAdded in v5.2.0.
Get the trace_id
of the current transaction’s trace. Example:
import elasticapm trace_id = elasticapm.get_trace_id()
elasticapm.get_span_id()
editAdded in v5.2.0.
Get the id of the current span. Example:
import elasticapm span_id = elasticapm.get_span_id()
elasticapm.set_custom_context()
editAdded in v2.0.0.
Attach custom contextual data to the current transaction and errors. Supported frameworks will automatically attach information about the HTTP request and the logged in user. You can attach further data using this function.
Before using custom context, ensure you understand the different types of metadata that are available.
Example:
import elasticapm elasticapm.set_custom_context({'billing_amount': product.price * item_count})
-
data
: (required) A dictionary with the data to be attached. This should be a flat key/valuedict
object.
.
, *
, and "
are invalid characters for key names and will be replaced with _
.
Errors that happen after this call will also have the custom context attached to them.
You can call this function multiple times, new context data will be merged with existing data,
following the update()
semantics of Python dictionaries.
elasticapm.set_user_context()
editAdded in v2.0.0.
Attach information about the currently logged in user to the current transaction and errors. Example:
import elasticapm elasticapm.set_user_context(username=user.username, email=user.email, user_id=user.id)
-
username
: The username of the logged in user -
email
: The email of the logged in user -
user_id
: The unique identifier of the logged in user, e.g. the primary key value
Errors that happen after this call will also have the user context attached to them.
You can call this function multiple times, new user data will be merged with existing data,
following the update()
semantics of Python dictionaries.
elasticapm.capture_span
editAdded in v4.1.0.
Capture a custom span.
This can be used either as a function decorator or as a context manager (in a with
statement).
When used as a decorator, the name of the span will be set to the name of the function.
When used as a context manager, a name has to be provided.
import elasticapm @elasticapm.capture_span() def coffee_maker(strength): fetch_water() with elasticapm.capture_span('near-to-machine', labels={"type": "arabica"}): insert_filter() for i in range(strength): pour_coffee() start_drip() fresh_pots()
-
name
: The name of the span. Defaults to the function name if used as a decorator. -
span_type
: (optional) The type of the span, usually in a dot-separated hierarchy oftype
,subtype
, andaction
, e.g.db.mysql.query
. Alternatively, type, subtype and action can be provided as three separate arguments, seespan_subtype
andspan_action
. -
skip_frames
: (optional) The number of stack frames to skip when collecting stack traces. Defaults to0
. -
leaf
: (optional) ifTrue
, all spans nested bellow this span will be ignored. Defaults toFalse
. -
labels
: (optional) a dictionary of labels. Keys must be strings, values can be strings, booleans, or numerical (int
,float
,decimal.Decimal
). Defaults toNone
. -
span_subtype
: (optional) subtype of the span, e.g. name of the database. Defaults toNone
. -
span_action
: (optional) action of the span, e.g.query
. Defaults toNone
. -
links
: (optional) A list ofTraceParent
objects to which this span is causally linked.
elasticapm.async_capture_span
editAdded in v5.4.0.
Capture a custom async-aware span.
This can be used either as a function decorator or as a context manager (in an async with
statement).
When used as a decorator, the name of the span will be set to the name of the function.
When used as a context manager, a name has to be provided.
import elasticapm @elasticapm.async_capture_span() async def coffee_maker(strength): await fetch_water() async with elasticapm.async_capture_span('near-to-machine', labels={"type": "arabica"}): await insert_filter() async for i in range(strength): await pour_coffee() start_drip() fresh_pots()
-
name
: The name of the span. Defaults to the function name if used as a decorator. -
span_type
: (optional) The type of the span, usually in a dot-separated hierarchy oftype
,subtype
, andaction
, e.g.db.mysql.query
. Alternatively, type, subtype and action can be provided as three separate arguments, seespan_subtype
andspan_action
. -
skip_frames
: (optional) The number of stack frames to skip when collecting stack traces. Defaults to0
. -
leaf
: (optional) ifTrue
, all spans nested bellow this span will be ignored. Defaults toFalse
. -
labels
: (optional) a dictionary of labels. Keys must be strings, values can be strings, booleans, or numerical (int
,float
,decimal.Decimal
). Defaults toNone
. -
span_subtype
: (optional) subtype of the span, e.g. name of the database. Defaults toNone
. -
span_action
: (optional) action of the span, e.g.query
. Defaults toNone
. -
links
: (optional) A list ofTraceParent
objects to which this span is causally linked.
asyncio
is only supported for Python 3.7+.
elasticapm.label()
editAdded in v5.0.0.
Attach labels to the the current transaction and errors.
Before using custom labels, ensure you understand the different types of metadata that are available.
Example:
import elasticapm elasticapm.label(ecommerce=True, dollar_value=47.12)
Errors that happen after this call will also have the labels attached to them.
You can call this function multiple times, new labels will be merged with existing labels,
following the update()
semantics of Python dictionaries.
Keys must be strings, values can be strings, booleans, or numerical (int
, float
, decimal.Decimal
)
.
, *
, and "
are invalid characters for label names and will be replaced with _
.
Avoid defining too many user-specified labels. Defining too many unique fields in an index is a condition that can lead to a mapping explosion.