GCP for DevOps — Error Reporting & Profiler

Error Reporting & Profiler

Error Reporting automatically groups and deduplicates errors from your applications. Cloud Profiler continuously profiles production applications to identify CPU and memory hotspots with minimal overhead.

Error Reporting

How It Works

Error Reporting Flow:
┌──────────────┐    ┌──────────────┐    ┌──────────────────┐
│ Application   │───▶│ Cloud         │───▶│ Error Reporting   │
│ (exception)   │    │ Logging       │    │                   │
│               │    │               │    │ • Group errors    │
│ Stack trace   │    │ Log entry     │    │ • Count / trend   │
│ logged to     │    │ with severity │    │ • Notify          │
│ stderr/stdout │    │ ERROR         │    │ • Link to source  │
└──────────────┘    └──────────────┘    └──────────────────┘

Error Reporting automatically detects errors from:

• Cloud Functions, Cloud Run, App Engine, GKE
• Compute Engine VMs (with Ops Agent)
• Any service writing structured error logs

Report Errors from Code

from google.cloud import error_reporting

client = error_reporting.Client()

try:
    result = 1 / 0
except ZeroDivisionError:
    client.report_exception()

# Report a custom error
client.report("Custom error: database connection timeout")

Structured Error Logging

Cloud Logging entries with severity ERROR and a stack trace are automatically picked up:

import traceback
import json

def handle_request():
    try:
        # Application logic
        process_data()
    except Exception as e:
        error_entry = {
            "severity": "ERROR",
            "message": str(e),
            "stack_trace": traceback.format_exc(),
            "serviceContext": {
                "service": "my-api",
                "version": "v1.2.3",
            },
        }
        print(json.dumps(error_entry))

Error Groups

Error Reporting groups similar errors together:

Feature	Description
Grouping	Same stack trace = same group
Count	Total occurrences over time
Trend	Increasing, decreasing, stable
First/Last seen	When error first/last occurred
Affected services	Which services are impacted
Resolution status	Open, acknowledged, resolved, muted

Cloud Profiler

How Profiler Works

Profiler Architecture:
┌──────────────────┐    ┌──────────────┐
│ Application       │    │ Cloud          │
│                   │    │ Profiler       │
│ ┌──────────────┐ │    │ Service        │
│ │ Profiler      │─┼───▶│               │
│ │ Agent         │ │    │ • Collect      │
│ │               │ │    │ • Analyze      │
│ │ ~0.5% CPU    │ │    │ • Flame graph  │
│ │ overhead      │ │    │ • Comparison   │
│ └──────────────┘ │    └──────────────┘
└──────────────────┘

Enable Profiler (Python)

# pip install google-cloud-profiler
import googlecloudprofiler

def main():
    googlecloudprofiler.start(
        service="my-api",
        service_version="1.0.0",
        verbose=0,
    )

    # Your application code
    start_server()

Enable Profiler (Go)

import "cloud.google.com/go/profiler"

func main() {
    cfg := profiler.Config{
        Service:        "my-api",
        ServiceVersion: "1.0.0",
    }
    if err := profiler.Start(cfg); err != nil {
        log.Fatalf("Profiler failed: %v", err)
    }

    // Your application code
}

Profile Types

Profile Type	What It Measures	Use Case
CPU time	Time spent executing code	Find CPU-intensive functions
Wall time	Real elapsed time	Find I/O waiting, blocking
Heap	Memory allocation	Find memory leaks
Allocated heap	Total bytes allocated	Find allocation hotspots
Threads	Active goroutines/threads	Find concurrency issues
Contention	Lock contention time	Find synchronization bottlenecks

Analyzing Flame Graphs

In the Cloud Console:

1. Navigate to Profiler
2. Select service, profile type, and time range
3. Read the flame graph:
   • Width = time spent in that function
   • Depth = call stack depth
   • Color = different packages/modules
4. Compare across versions or time periods

Web Security Scanner

# Create a scan config
$ gcloud beta web-security-scanner scan-configs create \
    --display-name="My App Scan" \
    --starting-urls="https://myapp.example.com"

# Run a scan
$ gcloud beta web-security-scanner scan-runs start SCAN_CONFIG_ID

# List scan results
$ gcloud beta web-security-scanner scan-runs list SCAN_CONFIG_ID

Common Mistakes

Mistake	Why It’s Wrong	Fix
No stack traces in error logs	Errors not grouped properly	Always log full stack traces
Missing serviceContext	Can’t filter by service	Add service name and version
Not resolving error groups	Stale errors pile up	Regularly triage and resolve
Profiler not enabled in prod	Can’t diagnose prod issues	Enable with minimal overhead
No version labeling	Can’t compare across deploys	Always set service version

Quick Reference

Feature	Access
Error Reporting	Console > Error Reporting
Cloud Profiler	Console > Profiler
Web Security Scanner	Console > Security > Web Security Scanner
Error Reporting API	google.cloud.error_reporting client
Profiler Agent	google-cloud-profiler package

Next → Practice Exercises