Production Logging

Pixflare can automatically push Worker logs to R2 for long-term storage. This gives you historical logs for compliance, troubleshooting production issues, and analyzing patterns over time. Cloudflare keeps real-time logs for only 24 hours, so Logpush is essential for preserving your log history.

Logpush captures errors, warnings, and slow requests - the stuff you actually care about - while skipping routine debug logs that would just waste storage and money.

Setup

Enable Logpush by setting three environment variables and creating R2 API tokens:

bash

# Enable Logpush
ENABLE_LOGPUSH='true'

# Create R2 API tokens (instructions below)
R2_ACCESS_KEY_ID='your-access-key-id'
R2_SECRET_ACCESS_KEY='your-secret-access-key'

# Optional: Customize storage location
LOGPUSH_R2_BUCKET='my-logs-bucket'        # Defaults to: {main-bucket}-logs
LOGPUSH_PATH_PREFIX='logs/workers'        # Defaults to: logs/workers

Creating R2 API Tokens

Go to your Cloudflare dashboard: R2 > Manage R2 API Tokens
Click Create API Token
Set permissions to Admin Read & Write
Copy the Access Key ID and Secret Access Key
Add them to your .env file

Deploy

After adding the environment variables:

bash

# Regenerate wrangler config with Logpush enabled
make generate-wrangler

# Deploy the updated worker
make deploy-workers

# Run the setup script to create the Logpush job
make non-terraform-setup

The setup script will:

Create a separate R2 bucket for logs (keeps them isolated from your images)
Configure a Logpush job to send Worker logs to R2
Verify everything is working correctly

Within 5-10 minutes of generating traffic, logs will start appearing in your R2 bucket.

How It Works

Here's what happens when your Worker receives requests:

mermaid

graph LR
    A[User Request] --> B[Worker]
    B --> C{Log Level}
    C -->|debug| D[Observability Tab Only]
    C -->|info/warn/error| E[Cloudflare Logs]
    E -->|10% sample| F[Logpush Service]
    F -->|Every 5 min| G[R2 Bucket]

    style D fill:#e3f2fd
    style E fill:#fff3e0
    style G fill:#e8f5e9

What Gets Logged

Logpush captures important events and skips the noise:

Captured (sent to R2):

Errors and exceptions with stack traces
Slow requests (>5 seconds)
HTTP errors (4xx, 5xx status codes)
Request metadata (timestamp, method, path, status)

Filtered out (not sent to R2):

Debug logs ("Incoming request", "User ensured")
Successful routine operations
Internal implementation details

The head_sampling_rate = 0.1 means only 10% of logs are saved, which keeps costs minimal while giving you plenty of data to work with. Even at scale, you'll stay well under $5/month.

Storage Layout

Logs are organized by date in your R2 bucket:

r2://your-bucket-logs/
  logs/
    workers/
      20260108/
        20260108T000852Z_20260108T000940Z_6393db3e.log.gz
        20260108T001352Z_20260108T001440Z_7b24cd4f.log.gz
        ...

Each file contains 5 minutes of compressed logs in newline-delimited JSON format.

Using the Logs

Quick Health Check

See recent log files to verify Logpush is working:

bash

# List today's logs
npx wrangler r2 object list your-bucket-logs \
  --prefix logs/workers/$(date +%Y%m%d) | head -20

Download and View Logs

Grab a specific log file and inspect it:

bash

# Download a log file
npx wrangler r2 object get \
  your-bucket-logs/logs/workers/20260108/FILE.log.gz \
  --file /tmp/logs.gz

# Decompress and view
gunzip /tmp/logs.gz
cat /tmp/logs | jq '.' | less

Find Errors

Search for errors from a specific time period:

bash

# Get errors with timestamps
cat /tmp/logs | jq -r '
  select(.Outcome == "exception") |
  "\(.EventTimestampMs | tonumber / 1000 | strftime("%Y-%m-%d %H:%M:%S")) - \(.Exceptions[0].message)"
'

Example output:

2026-01-08 12:34:56 - Failed to upload image: Invalid file format
2026-01-08 13:22:10 - Database query timeout

Analyze Slow Requests

Find performance bottlenecks:

bash

# Extract slow request warnings
cat /tmp/logs | jq -r '
  .Logs[]? |
  select(.message == "Slow request detected") |
  "\(.timestamp) - \(.path) - \(.duration)ms"
'

Monitor Error Rates

Count errors by type over time:

bash

# Download all logs from yesterday
DATE=$(date -d "yesterday" +%Y%m%d)
for file in $(npx wrangler r2 object list your-bucket-logs --prefix logs/workers/$DATE | awk '{print $1}'); do
  npx wrangler r2 object get your-bucket-logs/$file --file - | gunzip
done > /tmp/yesterday.log

# Count errors by exception type
cat /tmp/yesterday.log | jq -r '.Exceptions[]?.name' | sort | uniq -c | sort -rn

Check Request Patterns

See what endpoints are being hit:

bash

# Extract request paths and count them
cat /tmp/logs | jq -r '.Logs[]? | select(.path) | .path' | sort | uniq -c | sort -rn

Costs

With 10% sampling (the default), Logpush costs are minimal:

Logpush: $0.05 per million log entries x 10% = negligible
R2 Storage: ~$0.015/GB/month for stored logs
R2 Operations: Free for Class B reads

Most deployments will spend $1-3/month even with heavy traffic. You're already paying $5/month for Workers Paid (required for Logpush), so this is a tiny addition.

If you need to reduce costs further, you can:

Lower sampling rate in wrangler.toml: head_sampling_rate = 0.05 (5%)
Set up log rotation to delete old logs after 90 days
Use a smaller R2 bucket in a different region

Troubleshooting

Logs not appearing in R2?

Check the Logpush job status:

bash

curl -s -X GET \
  "https://api.cloudflare.com/client/v4/accounts/${CLOUDFLARE_ACCOUNT_ID}/logpush/jobs" \
  -H "Authorization: Bearer ${CLOUDFLARE_API_TOKEN}" | \
  jq '.result[] | {name, enabled, last_complete, last_error}'

Look for:

enabled: true (job is active)
last_complete: <timestamp> (job has run successfully)
last_error: null (no errors)

Still not working?

Verify you're on Workers Paid plan ($5/month minimum)
Check R2 API token has Admin Read & Write permissions
Ensure logpush = true is in your wrangler.production.toml
Generate some traffic and wait 10 minutes for the first batch

Need more help?

Check the Cloudflare Logpush documentation or open an issue on GitHub.

Multiple Destinations

You can send logs to multiple destinations simultaneously - R2 for long-term storage plus a monitoring service for real-time alerts.

Cloudflare Logpush supports:

DataDog - Real-time monitoring and alerting
Splunk - Enterprise log aggregation
New Relic - Application performance monitoring
SentinelOne - Security monitoring
IBM Cloud Logs - IBM cloud integration
S3-compatible providers - Any S3 API (Backblaze B2, MinIO, etc.)
HTTP endpoint - Custom log processing pipeline

To add another destination, create a second Logpush job:

bash

# Example: Send logs to DataDog
curl -X POST \
  "https://api.cloudflare.com/client/v4/accounts/${CLOUDFLARE_ACCOUNT_ID}/logpush/jobs" \
  -H "Authorization: Bearer ${CLOUDFLARE_API_TOKEN}" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "pixflare-datadog-logs",
    "dataset": "workers_trace_events",
    "destination_conf": "datadog://<DATADOG_ENDPOINT>?header_DD-API-KEY=<API_KEY>",
    "logpull_options": "fields=EventTimestampMs,EventType,Outcome,ScriptName,Logs,Exceptions&timestamps=rfc3339",
    "enabled": true,
    "filter": "{\"where\":{\"and\":[{\"key\":\"ScriptName\",\"operator\":\"eq\",\"value\":\"pixflare-production-api\"}]}}"
  }'

Each destination has its own configuration format. Check the Cloudflare Logpush destinations docs for specific setup instructions.

Production Logging ​

Setup ​

Creating R2 API Tokens ​

Deploy ​

How It Works ​

What Gets Logged ​

Storage Layout ​

Using the Logs ​

Quick Health Check ​

Download and View Logs ​

Find Errors ​

Analyze Slow Requests ​

Monitor Error Rates ​

Check Request Patterns ​

Costs ​

Troubleshooting ​

Multiple Destinations ​

Production Logging

Setup

Creating R2 API Tokens

Deploy

How It Works

What Gets Logged

Storage Layout

Using the Logs

Quick Health Check

Download and View Logs

Find Errors

Analyze Slow Requests

Monitor Error Rates

Check Request Patterns

Costs

Troubleshooting

Multiple Destinations