Cloud Trace / Cloud Profiler Complete Guide: Distributed Tracing, APM, and OpenTelemetry on GCP

Cloud Trace and Cloud Profiler are GCP's APM (Application Performance Monitoring) components. Trace provides distributed request tracing, while Profiler provides function-level profiling. OpenTelemetry integration lets you instrument your code with the industry standard.

Cloud Trace

Key Features

• Distributed tracing (request tracking across microservices)
• Timeline view of spans (units of work)
• Automatic latency distribution (P50 / P95 / P99)
• Latency anomaly detection (Insights)
• OpenTelemetry / OpenCensus support

OpenTelemetry Instrumentation Example (Python)

from opentelemetry import trace
from opentelemetry.sdk.trace import TracerProvider
from opentelemetry.sdk.trace.export import BatchSpanProcessor
from opentelemetry.exporter.cloud_trace import CloudTraceSpanExporter

trace.set_tracer_provider(TracerProvider())
trace.get_tracer_provider().add_span_processor(
    BatchSpanProcessor(CloudTraceSpanExporter())
)

tracer = trace.get_tracer(__name__)

@app.route('/order')
def create_order():
    with tracer.start_as_current_span("validate_order"):
        validate()
    with tracer.start_as_current_span("save_to_db"):
        save()
    with tracer.start_as_current_span("send_notification"):
        notify()
    return {"status": "ok"}

Cloud Profiler

Key Features

• Function-level CPU usage profiling
• Heap (memory) profiling
• Flame Graph visualization
• Low overhead (safe to run continuously in production)
• Languages: Java / Python / Go / Node.js
• Pricing: completely free

Setup Example (Python)

# pip install google-cloud-profiler

import googlecloudprofiler

googlecloudprofiler.start(
    service='my-api',
    service_version='1.0.0',
    verbose=3,  # 0=quiet, 4=verbose
)

Pricing

Sampling Strategy

OpenTelemetry Integration Best Practices

• Instrument with the SDK → send via the OTel Collector or directly to Cloud Trace
• Use auto-instrumentation to cover HTTP / gRPC / DB with zero code changes
• Split views by environment using service name / version
• Propagate trace context (W3C TraceContext) for cross-service tracking
• Attach resource attributes such as k8s.cluster.name / cloud.region

Typical Use Cases

• Pinpoint latency bottlenecks across microservices
• Investigate production performance regressions
• Drill down when SLOs are breached
• Detect regressions after new releases
• Cost optimization (CPU hot spots → refactoring)

Comparison with Other APM Tools

Google Cloud is a trademark of Google LLC. For the latest information, see the official Cloud Trace docs.