Absolutely! HAProxy has exactly the tools you need to prioritize high-priority URLs within the same backend—no need to waste resources on dedicated servers for your low-volume, infrequent high-priority traffic. You can use queue weighting, connection allocation, and timeout adjustments to ensure these requests don’t get stuck waiting behind low-priority traffic. Here’s how to implement it:

1. Queue Weighting (Direct Priority for Queued Requests) #

The most direct way to prioritize requests in the backend queue is to define multiple queues with different weights. HAProxy processes requests from higher-weight queues first when servers are busy.

Example Configuration: #

backend web_service
    # Your existing backend servers
    server web1 10.0.0.1:80 check
    server web2 10.0.0.2:80 check

    # Define queues with different weights
    queue default weight 10   # For low-priority traffic
    queue high_prio weight 100 # For high-priority traffic

    # Match your high-priority URLs
    acl is_high_prio path_beg /api/critical/ /admin/urgent/

    # Route matching requests to the high-priority queue
    use-server web1 queue high_prio if is_high_prio
    use-server web2 queue high_prio if is_high_prio

How It Works: #

When your backend servers are at capacity, incoming requests are routed to their respective queues. HAProxy will pull requests from the high_prio queue (weight 100) far more frequently than the default queue (weight 10), ensuring critical requests get processed first.

2. Allocate More Concurrent Connections to High-Priority Traffic #

You can also reserve a portion of your server’s concurrent connections for high-priority requests, preventing low-priority traffic from hogging all available slots.

Example Configuration: #

backend web_service
    server web1 10.0.0.1:80 check maxconn 100 # Total max connections per server
    server web2 10.0.0.2:80 check maxconn 100

    acl is_high_prio path_beg /api/critical/

    # Reserve 40% of connections for high-priority requests
    use-server web1 maxconn 40 if is_high_prio
    use-server web2 maxconn 40 if is_high_prio

How It Works: #

Each server will allow up to 40 concurrent connections for high-priority requests, even if the remaining 60 slots are filled with low-priority traffic. This guarantees critical requests always have access to server resources.

3. Adjust Timeouts for Priority Traffic #

Tweak timeout settings to prioritize high-priority requests further:

• Let high-priority requests wait longer in the queue before being rejected.
• Shorten timeouts for low-priority traffic to free up resources faster.

Example Configuration: #

backend web_service
    server web1 10.0.0.1:80 check
    server web2 10.0.0.2:80 check

    acl is_high_prio path_beg /api/critical/

    # High-priority requests can wait longer in the queue
    timeout queue 30s if is_high_prio
    # Low-priority requests get a shorter queue timeout
    timeout queue 5s

    # Optional: Shorten idle timeout for low-priority connections
    timeout client 60s if !is_high_prio
    timeout client 5m if is_high_prio

How It Works: #

Low-priority requests that can’t be processed quickly will be dropped after 5 seconds, freeing up queue space for critical requests. High-priority requests get more time to wait and stay connected longer if needed.

Final Notes #

You can combine these strategies for even better results. For example, use queue weighting alongside connection allocation to ensure both queued and incoming high-priority requests get preferential treatment. This approach keeps your resource utilization high while guaranteeing critical traffic isn’t delayed.

内容的提问来源于stack exchange，提问作者user2492286