Python

Collect HTTP server and CPython runtime metrics from your Python services and get curated dashboards and alerts for them with a single import.

Python apps instrumented with OpenTelemetry export their metrics over OTLP to your LeanSignal agent gateway (one per cluster or environment) — never to a remote or third-party endpoint. The gateway keeps full fidelity locally and forwards only the demanded subset to your central dataplane. This integration covers HTTP server request metrics (Flask/Django/FastAPI/WSGI/ASGI) plus CPython process/runtime metrics (memory, CPU, threads, GC).

Metric names follow the standard OTLP → Prometheus normalization (dots become underscores, unit/_total suffixes are added), so http.server.request.duration becomes http_server_request_duration_seconds and process.runtime.cpython.memory becomes process_runtime_cpython_memory_bytes. The demands below are built against those names.

Prerequisites

• A LeanSignal agent gateway deployed and connected — one per cluster or environment. If you haven’t deployed it yet, follow Install the agent (source: LeanSignal/leansignal-agent). It should show as Connected under Agents in the LeanSignal app.
• Network reachability to the gateway’s OTLP port — 4317 (gRPC) or 4318 (HTTP). See Agent configuration.
• Your Python app instrumented with the OpenTelemetry SDK, exporting metrics over OTLP to the gateway — configured in Setup.
• Editor or admin role in the LeanSignal app (importing a demand creates dashboards and alert rules).

Setup

Install the HTTP and system-metrics instrumentation plus the OTLP exporter, then point the SDK at your LeanSignal agent gateway (OTLP gRPC :4317 / HTTP :4318) — not a SaaS endpoint.

pip install opentelemetry-distro opentelemetry-exporter-otlp \
  opentelemetry-instrumentation-system-metrics
opentelemetry-bootstrap -a install   # adds Flask/Django/FastAPI/WSGI/ASGI instrumentors

export OTEL_SERVICE_NAME=my-python-app
export OTEL_EXPORTER_OTLP_ENDPOINT=http://<gateway-host>:4317   # :4318 for http/protobuf
export OTEL_EXPORTER_OTLP_PROTOCOL=grpc
export OTEL_METRICS_EXPORTER=otlp
export OTEL_SEMCONV_STABILITY_OPT_IN=http   # emit http.server.request.duration (seconds)

Run with auto-instrumentation:

opentelemetry-instrument python app.py

Enable the process/runtime metrics once in code:

from opentelemetry.instrumentation.system_metrics import SystemMetricsInstrumentor
SystemMetricsInstrumentor().instrument()

Opt-ins and caveats

The stable http_server_request_duration_seconds metric is emitted only when OTEL_SEMCONV_STABILITY_OPT_IN includes http; without it you get the legacy http_server_duration_milliseconds. Runtime metrics require SystemMetricsInstrumentor().instrument() and are CPython-only (skipped on PyPy). The process.runtime.* prefix is deprecated in semconv but still emitted today; the GC panel uses the cleaner cpython_gc_collections_total.

Where to send this

You have two options:

• Straight to the LeanSignal agent gateway. Point the app’s OTEL_EXPORTER_OTLP_ENDPOINT at it — nothing else in between.
• Via an OpenTelemetry Collector in agent mode. Point the app at a local collector and give that collector an otlp exporter to the gateway (leansignal-agent:4317).

Both are supported — pick whichever suits your topology. A local collector tends to fit when the app has no route to the gateway (a different network or cluster), or when you want local buffering, batching, or extra processing; exporting directly means one less process to operate.

Multiple instances

Give each running instance a stable identity so replicas stay distinct: set service.instance.id, exported as the instance label — for example OTEL_RESOURCE_ATTRIBUTES=service.instance.id=$(hostname), or your orchestrator’s pod name. The imported dashboard has an Instance filter — view all instances at once or focus on one.

Logs and traces

opentelemetry-instrument from Setup exports traces by default once OTEL_EXPORTER_OTLP_ENDPOINT is set. Logs are one opt-in away — the standard-library logging output is then captured and exported:

export OTEL_LOGS_EXPORTER="otlp"
export OTEL_PYTHON_LOGGING_AUTO_INSTRUMENTATION_ENABLED="true"

See the Python zero-code instrumentation docs.

Demand-driven, like metrics

Logs and traces land in the gateway’s edge stores (a roughly one-hour full-fidelity window) and are queryable immediately in the LeanSignal app under Logs and Traces with the source set to Available. Nothing is stored centrally until a dashboard demands it — the ready-made demands on this page stay metrics-only. See Logs and Traces.

Demands

Python ships as a ready-made demand in three variants. You don’t copy any JSON — in the LeanSignal app go to Demands, press the ˅ arrow on the Add Demand button (the arrow opens the import menu), choose Import from catalog…, and pick the variant. Each variant below lists the slug that identifies its published bundle. See Integrations for the full import flow.

The variants are nested — Standard is a superset of Essential, Extended a superset of Standard. Pick whichever is closest to what you need and treat it as a starting point: after import everything is a normal, editable copy, so retune thresholds, add or drop panels, and adjust queries for your environment — your edits reshape the demand automatically. Start small and re-import a larger variant later if you outgrow it.

Essential

The golden signals. 1 dashboard · 4 panels · 3 alerts — traffic, errors, latency, and memory at a glance.

Import this variant from the catalog. Its demand slug:

Demand slug

python-otel-demand-essential

Dashboard — Request rate, 5xx error rate, Request latency p95, and Process memory (RSS).

Alerts (3)

Alert	Severity	Fires when
High HTTP 5xx error rate	critical	5xx ratio > 5% for 5m
High request latency (p95)	warning	p95 latency > 1s for 5m
High process memory (RSS)	warning	RSS > 1 GiB for 10m

Review the bundle JSON →

Standard

Well-rounded coverage. 1 dashboard · 7 panels · 5 alerts — adds latency percentiles and per-route traffic. The recommended default.

Import this variant from the catalog. Its demand slug:

Demand slug

python-otel-demand-standard

Dashboard — everything in Essential, plus Request latency p50, Request latency p99, and Request rate by route.

Alerts (5)

Alert	Severity	Fires when
High HTTP 5xx error rate	critical	5xx ratio > 5% for 5m
High request latency (p95)	warning	p95 latency > 1s for 5m
High process memory (RSS)	warning	RSS > 1 GiB for 10m
Critical HTTP 5xx error rate	critical	5xx ratio > 20% for 5m
High request latency (p99)	warning	p99 latency > 2.5s for 5m

Review the bundle JSON →

Extended

Deep runtime visibility. 1 dashboard · 14 panels · 9 alerts — CPU, threads, context switches, and GC generations for services you need to diagnose, not just watch.

Import this variant from the catalog. Its demand slug:

Demand slug

python-otel-demand-extended

Dashboard — everything in Standard, plus In-flight requests, CPU time (cores), CPU utilization, Thread count, Context switches, Virtual memory (VMS), and GC collections by generation.

Alerts (9)

Alert	Severity	Fires when
High HTTP 5xx error rate	critical	5xx ratio > 5% for 5m
High request latency (p95)	warning	p95 latency > 1s for 5m
High process memory (RSS)	warning	RSS > 1 GiB for 10m
Critical HTTP 5xx error rate	critical	5xx ratio > 20% for 5m
High request latency (p99)	warning	p99 latency > 2.5s for 5m
Thread count elevated	warning	> 300 threads for 10m
Frequent gen-2 garbage collection	warning	> 1 gen-2 GC/s for 10m
Involuntary context-switch storm	warning	> 1000 involuntary switches/s for 10m
No inbound traffic	warning	no requests for 15m

Review the bundle JSON →

Dashboards

Each variant imports one Python dashboard, its panels grouped into collapsible sections. Every dashboard carries an Instance filter (on the instance label), so one import covers the whole fleet — view every instance at once, or focus on one. Legends are sortable tables showing last/mean/max per series, so the outlier instance stands out. Dashboards are demand-driven: importing one tells your agents to forward exactly the timeseries its panels query, and nothing else. Panels are ordinary Perses panels — edit queries, add panels, or retune them after import, and your changes reshape the demand automatically.

Alerts

Thresholds are conservative starting points — tune them to your workload.

Caution

Imported alert rules arrive without a notification channel and start out paused — with nowhere to deliver, they do not evaluate. After importing, create or pick a channel under Alerts → Manage Channels, attach it to each rule, and resume. Rules without a channel are marked with a warning icon in the alerts list.

Next steps

• Demand-driven observability — why importing a demand also shapes what your agents collect.
• Agent configuration — the full agent config surface.