When working with HAProxy's round-robin load balancing algorithm, you can fine-tune traffic distribution by assigning different weights to backend servers. This is particularly useful when you need to:

• Prioritize newer hardware with more capacity
• Gradually introduce new servers into production
• Handle maintenance windows gracefully

For your specific requirement of 70%/15%/15% distribution across three MySQL servers, here's the modified configuration:

listen MySQL 10.10.10.14:3306
   mode tcp
   balance roundrobin
   option persist
   server sql1 10.10.10.4:3306 weight 70
   server sql2 10.10.10.5:3306 weight 15
   server sql3 10.10.10.6:3306 weight 15

After implementing the weighted configuration, you should:

1. Check the runtime weights using the HAProxy stats page or CLI
2. Monitor the actual traffic distribution with tools like show stat
3. Verify that the weight adjustments don't exceed server capacities

When implementing weighted round-robin in production:

• Make gradual weight changes (10% increments) when possible
• Consider using on-marked-down and on-marked-up options for failover scenarios
• Remember that weights only affect new connections, not existing ones

For more dynamic control, you can adjust weights on the fly using runtime API:

echo "set server MySQL/sql1 weight 80" | socat stdio /var/run/haproxy.sock
echo "set server MySQL/sql2 weight 10" | socat stdio /var/run/haproxy.sock
echo "set server MySQL/sql3 weight 10" | socat stdio /var/run/haproxy.sock

The runtime API allows for zero-downtime adjustments, which is crucial for production environments.

When working with HAProxy's round-robin algorithm, the weight parameter is exactly what you need to achieve uneven distribution of traffic. This becomes particularly useful when you have servers with different capacities or when implementing canary deployments.


Here's how to implement your desired 70/15/15 distribution:

listen MySQL 10.10.10.14:3306
   mode tcp
   balance roundrobin
   option persist
   server sql1 10.10.10.4:3306 weight 70
   server sql2 10.10.10.5:3306 weight 15
   server sql3 10.10.10.6:3306 weight 15
After implementing weighted round robin, verify the distribution using HAProxy's stats page or by analyzing logs. Important considerations:

Weights are relative - values could be 140/30/30 for same ratio
Dynamic weight adjustment is possible via runtime API
Consider on-marked-down and on-marked-up for failover scenarios

For production changes, I recommend this rollout sequence:

First add weights to non-production environment
Test with gradual weight changes (e.g., 40/30/30 → 50/25/25 → 70/15/15)
Use HAProxy's runtime API for zero-downtime adjustments

For dynamic environments, consider these additional parameters:
server sql1 10.10.10.4:3306 weight 70 check inter 5s rise 2 fall 3
server sql2 10.10.10.5:3306 weight 15 backup
server sql3 10.10.10.6:3306 weight 15 disabled