Troubleshooting Guide

Common Cloud Run issues and their solutions.

Deployment Issues

Container Failed to Start

Error	Cause	Solution
Container failed to start	App not listening on PORT	Ensure app listens on $PORT env var
Memory limit exceeded	OOM during startup	Increase --memory flag
Container failed to start within 4 minutes	Slow initialization	Increase --timeout or optimize startup

# Check container logs
gcloud logging read "resource.type=cloud_run_revision AND textPayload:failed" --limit 20

# Increase memory
gcloud run services update SERVICE --region REGION --memory 2Gi

# Increase timeout
gcloud run services update SERVICE --region REGION --timeout 600

Permission Denied

# Grant Cloud Run admin role
gcloud projects add-iam-policy-binding PROJECT_ID \
  --member user:YOUR_EMAIL \
  --role roles/run.admin

# Grant Artifact Registry reader
gcloud projects add-iam-policy-binding PROJECT_ID \
  --member serviceAccount:PROJECT_NUMBER-compute@developer.gserviceaccount.com \
  --role roles/artifactregistry.reader

Image Not Found

# Verify image exists
gcloud artifacts docker images list us-central1-docker.pkg.dev/PROJECT/REPO

# Check image path format
# Correct: us-central1-docker.pkg.dev/project/repo/image:tag
# Wrong:   gcr.io/project/image (old format, may still work)

Runtime Issues

Cold Start Latency

Solution	Impact
--min-instances 1	Eliminates cold starts, increases cost
--startup-cpu-boost	Faster startup, slight cost increase
Reduce image size	Faster container pull
Use multi-stage builds	Smaller final image

# Keep instance warm
gcloud run services update SERVICE --region REGION --min-instances 1

# Enable CPU boost
gcloud run services update SERVICE --region REGION --startup-cpu-boost

Request Timeouts

# Increase timeout (max 3600s for HTTP, 86400s for streaming)
gcloud run services update SERVICE --region REGION --timeout 900

# Check current timeout
gcloud run services describe SERVICE --region REGION \
  --format 'value(spec.template.spec.timeoutSeconds)'

Memory Errors (OOM)

# Check memory usage in logs
gcloud logging read "resource.type=cloud_run_revision AND textPayload:memory" --limit 50

# Increase memory
gcloud run services update SERVICE --region REGION --memory 4Gi

# Memory limits
# Min: 128Mi
# Max: 32Gi (or 64Gi with specific configs)

Concurrency Issues

SYMPTOMS

• Requests queuing
• 503 errors under load
• High latency spikes

# Reduce concurrency (for CPU-bound workloads)
gcloud run services update SERVICE --region REGION --concurrency 10

# Increase max instances
gcloud run services update SERVICE --region REGION --max-instances 100

# Check current settings
gcloud run services describe SERVICE --region REGION \
  --format 'yaml(spec.template.spec.containerConcurrency)'

Networking Issues

VPC Connector Errors

# Check connector status
gcloud compute networks vpc-access connectors describe CONNECTOR \
  --region REGION

# Recreate connector
gcloud compute networks vpc-access connectors delete CONNECTOR --region REGION
gcloud compute networks vpc-access connectors create CONNECTOR \
  --network default \
  --region REGION \
  --range 10.8.0.0/28

Cannot Connect to Database

Cause	Solution
No VPC connector	Create and attach VPC connector
Private IP only	Use Cloud SQL Proxy or VPC
Firewall rules	Allow Cloud Run IP ranges

# For Cloud SQL (recommended)
gcloud run services update SERVICE \
  --region REGION \
  --add-cloudsql-instances PROJECT:REGION:INSTANCE

CORS Errors

Add CORS headers in your application:

// Express.js example
const cors = require('cors');
app.use(cors({
  origin: ['https://your-frontend.com'],
  methods: ['GET', 'POST', 'PUT', 'DELETE'],
  allowedHeaders: ['Content-Type', 'Authorization']
}));

GPU Issues

GPU Not Available

# Check quota
gcloud compute regions describe us-central1 \
  --format 'value(quotas[?metric==NVIDIA_L4_GPUS])'

# Request quota increase
# Go to: https://console.cloud.google.com/iam-admin/quotas

CUDA Errors

Error	Solution
CUDA out of memory	Reduce batch size or model size
CUDA not available	Verify --gpu 1 flag in deploy
Driver version mismatch	Use compatible CUDA base image

# Verify GPU attached
gcloud run services describe SERVICE --region REGION \
  --format 'yaml(spec.template.spec.containers[0].resources)'

Logging & Debugging

Enable Detailed Logs

# Set log level in app
gcloud run services update SERVICE \
  --region REGION \
  --set-env-vars "LOG_LEVEL=debug"

# View structured logs
gcloud logging read 'resource.type="cloud_run_revision"' \
  --format json \
  --limit 100

Debug Locally

# Run container locally with same env
docker run -p 8080:8080 \
  -e PORT=8080 \
  -e K_SERVICE=local \
  YOUR_IMAGE

# Test
curl http://localhost:8080

Check Service Health

#!/bin/bash
SERVICE_URL=$(gcloud run services describe SERVICE \
  --region REGION --format 'value(status.url)')

echo "Testing $SERVICE_URL"

# Health check
curl -w "\nStatus: %{http_code}\nTime: %{time_total}s\n" \
  -o /dev/null -s "$SERVICE_URL/health"

# Load test (requires hey)
hey -n 100 -c 10 "$SERVICE_URL/"

Common Error Codes

HTTP Code	Meaning	Common Cause
403	Forbidden	IAM permissions
404	Not Found	Wrong URL or service not deployed
429	Too Many Requests	Rate limiting / quota exceeded
500	Internal Error	Application crash
502	Bad Gateway	Container not responding
503	Unavailable	Scaling / cold start issue
504	Timeout	Request exceeded timeout