rotel 🌶️ 🍅

Python package for Rotel.

Description

This package provides an embedded lightweight agent that can collect OpenTelemetry data, built on Rotel. When started, it spawns a background daemon that accepts OpenTelemetry metrics, traces, and logs. Designed for minimal overhead, Rotel reduces resource consumption while simplifying telemetry collection and processing in complex Python applications—without requiring additional sidecar containers.

Telemetry Type	Support
Metrics	Alpha
Traces	Alpha
Logs	Alpha

How it works

By default, the Rotel agent listens for OpenTelemetry data over gRPC (port 4317) and HTTP (port 4318) on localhost. It efficiently batches telemetry signals and forwards them to a configurable OpenTelemetry protocol (OTLP) compatible endpoint.

In your application, you use the OpenTelemetry Python SDK to add instrumentation for traces, metrics, and logs. The SDK by default will communicate over ports 4317 or 4318 on localhost to the Rotel agent. You can now ship your instrumented application and efficiently export OpenTelemetry data to your vendor or observability tool of choice with a single deployment artifact.

Future updates will introduce support for filtering data, transforming telemetry, and exporting to different vendors and tools.

Getting started

Rotel configuration

Add the rotel Python package to your project's dependencies. There are two approaches to configuring rotel:

1. typed config dicts
2. environment variables

Typed dicts

In the startup section of your main.py add the following code block. Replace the endpoint with the endpoint of your OpenTelemetry vendor and any required API KEY headers.

fromrotelimportConfig, Rotelrotel=Rotel(
enabled=True,
exporters= {
'otlp': Config.otlp_exporter(
endpoint="https://foo.example.com",
headers= {
"x-api-key" : settings.API_KEY,
"x-data-set": "testing"
}
),
},
# Define exporters per telemetry typeexporters_traces= ['otlp'],
exporters_metrics= ['otlp'],
exporters_logs= ['otlp']
)
rotel.start()

Environment variables

You can also configure rotel entirely with environment variables. In your application startup, insert:

importrotelrotel.start()

In your application deployment configuration, set the following environment variables. These match the typed configuration above:

• ROTEL_ENABLED=true
• ROTEL_EXPORTERS=otlp
• ROTEL_EXPORTER_OTLP_ENDPOINT=https://foo.example.com
• ROTEL_EXPORTER_OTLP_CUSTOM_HEADERS=x-api-key={API_KEY},x-data-set=testing
• ROTEL_EXPORTERS_TRACES=otlp
• ROTEL_EXPORTERS_METRICS=otlp
• ROTEL_EXPORTERS_LOGS=otlp

Any typed configuration options will override environment variables of the same name.

---

See the Configuration section for the full list of options.

OpenTelemetry SDK configuration

Once the Rotel agent is running, you may need to configure your application's instrumentation. If you are using the default rotel endpoints of localhost:4317 and localhost:4318, then you should not need to change anything.

To set the endpoint the OpenTelemetry SDK will use, set the following environment variable:

• OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317

Configuration

This is the full list of global options and their environment variable alternatives. Any defaults left blank in the table are either False or None.

Option Name	Type	Environ	Default	Options
enabled	bool	ROTEL_ENABLED
pid_file	str	ROTEL_PID_FILE	/tmp/rotel-agent.pid
log_file	str	ROTEL_LOG_FILE	/tmp/rotel-agent.log
log_format	str	ROTEL_LOG_FORMAT	text	json, text
debug_log	list[str]	ROTEL_DEBUG_LOG		traces, metrics, logs
debug_log_verbosity	str	ROTEL_DEBUG_LOG_VERBOSITY	basic	basic, detailed
otlp_grpc_endpoint	str	ROTEL_OTLP_GRPC_ENDPOINT	localhost:4317
otlp_http_endpoint	str	ROTEL_OTLP_HTTP_ENDPOINT	localhost:4318

For each exporter you would like to use, see the configuration options below. Exporters should be assigned to the exporters dict with a custom name.

OTLP Exporter

To construct an OTLP exporter, use the method Config.otlp_exporter() with the following options.

Option Name	Type	Default	Options
endpoint	str
protocol	str	grpc	grpc or http
headers	dict[str, str]
compression	str	gzip	gzip or none
request_timeout	str	5s
retry_initial_backoff	str	5s
retry_max_backoff	str	30s
retry_max_elapsed_time	str	300s
batch_max_size	int	8192
batch_timeout	str	200ms
tls_cert_file	str
tls_key_file	str
tls_ca_file	str
tls_skip_verify	bool

Datadog Exporter

Rotel provides an experimental Datadog exporter that supports traces at the moment. Construct a Datadog exporter with the method Config.datadog_exporter() using the following options.

Option Name	Type	Default	Options
region	str	us1	us1, us3, us5, eu, ap1
custom_endpoint	str
api_key	str

Clickhouse Exporter

Rotel provides a Clickhouse exporter with support metrics, logs, and traces. Construct a Clickhouse exporter with the method Config.clickhouse_exporter() using the following options.

Option Name	Type	Default	Options
endpoint	str
database	str	otel
table_prefix	str	otel
compression	str	lz4
async_insert	bool	true
user	str
password	str
enable_json	bool
json_underscore	bool

Kafka Exporter

Rotel provides a Kafka exporter with support for metrics, logs, and traces. Construct a Kafka exporter with the method Config.kafka_exporter() using the following options.

Option Name	Type	Default	Options
brokers	list	localhost:9092
traces_topic	str	otlp_traces
logs_topic	str	otlp_logs
metrics_topic	str	otlp_metrics
format	str	protobuf	json, protobuf
compression	str	none	gzip, snappy, lz4, zstd, none
acks	str	one	all, one, none
client_id	str	rotel
max_message_bytes	int	1000000
linger_ms	int	5
retries	int	2147483647
retry_backoff_ms	int	100
retry_backoff_max_ms	int	1000
message_timeout_ms	int	300000
request_timeout_ms	int	30000
batch_size	int	1000000
partitioner	str	consistent-random	consistent, consistent-randomm, murmur2-random, murmur2, fnv1a, fnv1a-random
partitioner_metrics_by_resource_attributes	str	1000
partitioner_logs_by_resource_attributes	str	1000
custom_config	str	1000
sasl_username	str
sasl_password	str
sasl_mechanism	str		plain, scram-sha256, scram-sha512
security_protocol	str	plaintext	plaintext, ssl, sasl-plaintext, sasl-ssl

Multiple exporters

Pyrotel supports multiple exporters, allowing you to send data to different destinations per telemetry type. Just set the exporters entry to a dict map of exporter definitions and then configure the exporters per telemetry type. For example, this will send metrics and logs to an OTLP endpoint while sending traces to Datadog:

fromrotelimportConfig, Rotelrotel=Rotel(
enabled=True,
exporters= {
'logs_and_metrics': Config.otlp_exporter(
endpoint="https://foo.example.com",
headers= {
"x-api-key" : settings.API_KEY,
"x-data-set": "testing"
}
),
'tracing': Config.datadog_exporter(
api_key="1234abcd",
),
},
# Define exporters per telemetry typeexporters_traces= ['tracing'],
exporters_metrics= ['logs_and_metrics'],
exporters_logs= ['logs_and_metrics']
)
rotel.start()

Processors

You can pass a list of Python files to Rotel that support the Python Processor SDK, by settting the following top-level config values. Each takes a list of absolute file paths to files implementing the processor SDK.

Option Name	Type
processors_metrics	list[str]
processors_traces	list[str]
processors_logs	list[str]

See the Python Processor SDK docs for more information.

Retries and timeouts

You can override the default request timeout of 5 seconds for the OTLP Exporter with the exporter setting:

• request_timeout: Takes a string time duration, so "250ms" for 250 milliseconds, "3s" for 3 seconds, etc.

Requests will be retried if they match retryable error codes like 429 (Too Many Requests) or timeout. You can control the behavior with the following exporter options:

• retry_initial_backoff: Initial backoff duration
• retry_max_backoff: Maximum backoff interval
• retry_max_elapsed_time: Maximum wall time a request will be retried for until it is marked as permanent failure

All options should be represented as string time durations.

Full OTEL example

To illustrate this further, here's a full example of how to use Rotel to send trace spans to Axiom from an application instrumented with OpenTelemetry.

The code sample depends on the following environment variables:

• ROTEL_ENABLED=true: Turn on or off based on the deployment environment
• AXIOM_DATASET: Name of an Axiom dataset
• AXIOM_API_TOKEN: Set to an API token that has access to the Axiom dataset

importosfromrotelimportConfig, Rotelfromopentelemetryimporttracefromopentelemetry.sdk.traceimportTracerProviderfromopentelemetry.sdk.resourcesimportResource, SERVICE_NAMEfromopentelemetry.exporter.otlp.proto.http.trace_exporterimportOTLPSpanExporterfromopentelemetry.sdk.trace.exportimportSimpleSpanProcessor# Enable at deploy time with ROTEL_ENABLED=trueifos.environ.get("ROTEL_ENABLED") =="true":
## Configure Rotel to export to Axiom#otlp_exporter=Config.otlp_exporter(
endpoint="https://api.axiom.co",
protocol="http", # Axiom only supports HTTPheaders={
"Authorization": f"Bearer {os.environ['AXIOM_API_TOKEN']}",
"X-Axiom-Dataset": os.environ["AXIOM_DATASET"],
},
)
rotel=Rotel(
enabled=True,
exporters= {
'axiom': otlp_exporter,
},
exporters_traces= ['axiom']
)
# Start the agentrotel.start()
## Configure OpenTelemetry SDK to export to the localhost Rotel## Define the service name resource for the tracer.resource=Resource(
attributes={
SERVICE_NAME: "pyrotel-test"
}
)
# Create a TracerProvider with the defined resource for creating tracers.provider=TracerProvider(resource=resource)
# Create the OTel exporter to send to the localhost Rotel agentexporter=OTLPSpanExporter(endpoint="http://localhost:4318/v1/traces")
# Create a processor with the OTLP exporter to send trace spans.## You could also use the BatchSpanProcessor, but since Rotel runs locally# and will batch, you can avoid double batching.processor=SimpleSpanProcessor(exporter)
provider.add_span_processor(processor)
# Set the TracerProvider as the global tracer provider.trace.set_tracer_provider(provider)

For the complete example, see the hello world application.

Debugging

If you set the option debug_log to ["traces"], or the environment variable ROTEL_DEBUG_LOG=traces, then rotel will log a summary to the log file /tmp/rotel-agent.log each time it processes trace spans. You can add also specify metrics to debug metrics and logs to debug logs.

FAQ

Do I need to call rotel.stop() when I exit?

In most deployment environments you do not need to call rotel.stop() and it is generally recommended that you don't. Calling rotel.stop() will terminate the running agent on a host, so any further export calls from OTEL instrumentation will fail. In a multiprocess environment, such as gunicorn, terminating the Rotel agent from one process will terminate it for all other processes. On ephemeral deployment platforms, it is usually fine to leave the agent running until the compute instance, VM/container/isolate, terminate.

Community

Want to chat about this project, share feedback, or suggest improvements? Join our Discord server! Whether you're a user of this project or not, we'd love to hear your thoughts and ideas. See you there! 🚀

Developing

See the DEVELOPING.md doc for building and development instructions.