Postfix Rate Limiting and Throttling

Rate limiting in Postfix prevents abuse, protects your IP reputation, and ensures fair resource usage. Whether you need to limit outbound sending rates to avoid being flagged as a spammer, or restrict inbound connections to prevent denial-of-service, Postfix provides multiple throttling mechanisms. This guide covers all rate limiting options with practical configurations.

Why Rate Limit?

• IP reputation — sending too fast triggers spam filters at receiving servers
• Abuse prevention — prevent compromised accounts from sending bulk spam
• Resource protection — avoid overwhelming your server or remote servers
• Warm-up — gradually increase sending volume for new IP addresses

Outbound Rate Limiting

Per-Destination Rate Delay

# /etc/postfix/main.cf
# Delay between messages to the same destination
default_destination_rate_delay = 1s

# Per-transport overrides
smtp_destination_rate_delay = 1s

# This means Postfix waits 1 second between delivering messages
# to the same remote domain — effective for not overwhelming recipients

Per-Destination Concurrency

# Maximum simultaneous connections to any single destination
default_destination_concurrency_limit = 20
smtp_destination_concurrency_limit = 20

# For sensitive destinations (e.g., known rate-limiters)
# Use transport maps with per-destination settings
# /etc/postfix/main.cf
transport_maps = hash:/etc/postfix/transport

# /etc/postfix/transport
gmail.com    slow-smtp:
outlook.com  slow-smtp:
yahoo.com    slow-smtp:

# /etc/postfix/master.cf
slow-smtp    unix  -       -       n       -       5       smtp
    -o smtp_destination_concurrency_limit=5
    -o smtp_destination_rate_delay=2s
    -o smtp_extra_recipient_limit=5

Global Sending Rate

# Maximum messages per second (across all destinations)
# Controlled via active queue size
qmgr_message_active_limit = 500

# Limit recipients per delivery
default_extra_recipient_limit = 50
smtp_extra_recipient_limit = 50

# Maximum recipients per message
smtpd_recipient_limit = 100

Inbound Rate Limiting

Connection Rate Limiting

# /etc/postfix/main.cf
# Rate limit client connections
smtpd_client_connection_rate_limit = 30     # Max connections per minute per client
smtpd_client_connection_count_limit = 10    # Max simultaneous connections per client
smtpd_client_message_rate_limit = 60        # Max messages per minute per client
smtpd_client_recipient_rate_limit = 100     # Max recipients per minute per client
smtpd_client_new_tls_session_rate_limit = 30

# Rate limit cache
anvil_rate_time_unit = 60s                  # Time window for rate counting
anvil_status_update_time = 600s             # How often to log anvil stats

# Exempt trusted clients
smtpd_client_event_limit_exceptions = $mynetworks

Per-User Sending Limits (SASL)

# Limit authenticated users
# Requires Postfix 3.1+ with smtpd_relay_restrictions

# Using policy service for per-user limits
smtpd_sender_restrictions =
    check_policy_service unix:private/policy-spf,
    permit_sasl_authenticated,
    reject

# For more granular per-user limits, use a policy daemon
# Example: postfix-policyd-spf-python or custom policy script

Policy Daemon for Advanced Rate Limiting

#!/usr/bin/env python3
# /usr/local/bin/rate-limit-policy.py
# Simple per-sender rate limiting policy daemon

import sys
import time
from collections import defaultdict

RATE_LIMIT = 100  # messages per hour
WINDOW = 3600     # 1 hour in seconds

counters = defaultdict(list)

def check_rate(sender):
    now = time.time()
    # Clean old entries
    counters[sender] = [t for t in counters[sender] if now - t < WINDOW]
    if len(counters[sender]) >= RATE_LIMIT:
        return "DEFER_IF_PERMIT Rate limit exceeded. Try again later."
    counters[sender].append(now)
    return "DUNNO"

while True:
    attrs = {}
    for line in sys.stdin:
        line = line.strip()
        if not line:
            break
        key, _, value = line.partition("=")
        attrs[key] = value

    sender = attrs.get("sasl_username", attrs.get("sender", ""))
    action = check_rate(sender)
    sys.stdout.write(f"action={action}\n\n")
    sys.stdout.flush()

# /etc/postfix/master.cf
policy-ratelimit unix - n n - 0 spawn
    user=nobody argv=/usr/bin/python3 /usr/local/bin/rate-limit-policy.py

# /etc/postfix/main.cf
smtpd_data_restrictions = check_policy_service unix:private/policy-ratelimit

IP Warm-Up Schedule

# For new IP addresses, gradually increase sending volume
# Week 1: 50 messages/hour
# Week 2: 200 messages/hour
# Week 3: 500 messages/hour
# Week 4: 1000 messages/hour
# Week 5+: Full speed

# Adjust these settings weekly during warm-up:
# /etc/postfix/main.cf
default_destination_rate_delay = 60s   # Week 1: very slow
# Gradually reduce to 0 over weeks

Monitoring Rate Limits

# Check anvil statistics
postconf -d anvil_status_update_time
# Anvil logs to mail.log every 600s by default

# View current rate data
postfix check

# Check for rate-limited connections in logs
grep "rate limit" /var/log/mail.log
grep "anvil" /var/log/mail.log

# Monitor with postfix_exporter (Prometheus)
# Tracks connections, message rates, and rejections per client

Handling Legitimate Bulk Senders

# Whitelist known bulk senders from rate limits
smtpd_client_event_limit_exceptions =
    $mynetworks,
    hash:/etc/postfix/rate-limit-exceptions

# /etc/postfix/rate-limit-exceptions
# IP addresses of trusted bulk senders
10.0.0.50   OK
10.0.0.51   OK

Best Practices

• Start with conservative limits and increase based on observed delivery success
• Use per-destination rate delays for outbound mail to major providers (Gmail, Outlook, Yahoo)
• Exempt $mynetworks from inbound rate limits to avoid blocking trusted sources
• Monitor anvil statistics and mail logs for signs of rate limiting hitting legitimate traffic
• Implement per-user rate limits for shared hosting environments to prevent one user from consuming all resources
• Warm up new IP addresses gradually over 4-6 weeks
• Log rate limit events for debugging and abuse investigation