Scaling Past the Breaking Point: High Availability with HAProxy 1.4
There is nothing quite as soul-crushing as watching `top` report a load average of 50.00 while your Apache error logs scream "MaxClients reached." It’s 2011, and if your infrastructure still relies on a single massive server to handle all your web traffic, you are gambling with your uptime. I've seen it happen too often: a marketing email goes out, traffic spikes, and the server melts just when it needs to perform.
It doesn't matter how much RAM you throw at a single box. Eventually, you hit a wall—whether it's I/O contention or CPU stealing. The solution isn't a bigger server; it's smarter architecture. Today, we are going to look at HAProxy 1.4, the gold standard for software load balancing, and how to use it to horizontally scale your PHP applications.
The Scenario: The "Thundering Herd"
Last month, I was called in to troubleshoot a high-profile eZ Publish deployment here in Norway. They were running a standard LAMP stack (Linux, Apache, MySQL, PHP) on a single physical server. Every time they ran a campaign, the database locked up, and Apache processes consumed every bit of swap.
The fix wasn't more hardware—it was splitting the load. We deployed two frontend web servers and placed a lightweight Load Balancer (LB) in front. While hardware load balancers like F5 are great, they are prohibitively expensive for most startups. HAProxy offers comparable performance for free, provided you have the Linux chops to configure it.
Configuring HAProxy for Layer 7 Balancing
HAProxy runs beautifully on a minimal Linux install. For this setup, we are using CentOS 5.5, but the commands are similar for Debian 6 (Squeeze).
First, install the package:
yum install haproxy
The magic happens in /etc/haproxy/haproxy.cfg. Unlike simple round-robin DNS, HAProxy allows us to inspect HTTP headers and maintain session persistence—critical for keeping users logged in.
Here is a battle-tested configuration snippet for a standard web cluster:
global
log 127.0.0.1 local0
maxconn 4096
user haproxy
group haproxy
daemon
defaults
log global
mode http
option httplog
option dontlognull
retries 3
option redispatch
timeout connect 5000
timeout client 50000
timeout server 50000
frontend http_front
bind *:80
# ACL to separate static content if needed
acl url_static path_beg /static /images /css
default_backend web_servers
backend web_servers
mode http
balance roundrobin
# Sticky sessions via cookie
cookie SERVERID insert indirect nocache
option httpchk HEAD /health_check.php HTTP/1.0
server web01 192.168.10.11:80 cookie web01 check inter 2000 rise 2 fall 3
server web02 192.168.10.12:80 cookie web02 check inter 2000 rise 2 fall 3
Breaking Down the Config
- option httpchk: HAProxy constantly pings
health_check.php. Ifweb01goes down (kernel panic, power outage, or bad cable), HAProxy instantly removes it from the pool. No more dead pages for users. - cookie SERVERID: This injects a cookie so that once a user is connected to
web01, they stay there. This is mandatory for most shopping carts unless you are storing sessions in Memcached.
The Hardware Reality: Latency and I/O
Software is only as fast as the iron it runs on. Even with HAProxy, if your backend servers are struggling with disk I/O, your "Time to First Byte" will suffer. In Norway, we also have to consider geography. Routing traffic from Oslo to a server in Frankfurt adds latency that your users can feel.
Pro Tip: Network latency within Norway (via NIX - the Norwegian Internet Exchange) is typically under 10ms. Routing outside the country can triple that. For latency-sensitive applications, hosting locally is not just patriotic; it's a technical requirement.
This is where the choice of VPS matters. Many budget hosts oversell their nodes, leading to "noisy neighbor" issues where another customer's cron job kills your performance. At CoolVDS, we utilize KVM virtualization rather than OpenVZ. This ensures that the RAM and CPU resources assigned to your load balancer are actually yours.
Storage Speed
While standard 7.2k RPM SATA drives are fine for backups, your database and web roots need speed. We are seeing a massive shift towards SSD storage in the enterprise space. While still expensive compared to spinning rust, the IOPS (Input/Output Operations Per Second) gain is astronomical. CoolVDS offers high-performance storage tiers that ensure your MySQL queries aren't waiting on the disk head to seek.
Compliance and Data Location
Beyond performance, we have the legal landscape. The Personopplysningsloven (Personal Data Act) places strict requirements on how we handle Norwegian citizens' data. Datatilsynet is very clear about the responsibilities of data controllers.
Hosting your infrastructure on CoolVDS servers located physically in Norway simplifies this compliance. You know exactly where the data lives—not "somewhere in the cloud," but in a secure datacenter in Oslo. This matters for everything from medical records to simple e-commerce transactions.
Conclusion
High availability isn't just for Google or Facebook. With open-source tools like HAProxy and robust VPS hosting, you can build an architecture that survives traffic spikes and hardware failures. Don't wait for your single server to crash during your next big sale.
Ready to stabilize your stack? Spin up a VPS Norway instance on CoolVDS today. With our low latency network and dedicated resources, it’s the perfect foundation for your load-balanced cluster.