Healthcare & Data
Real-Time Data Failover for Linux Servers
The Challenge
The client demanded a robust, self-healing system where any file change on one server would be instantly mirrored to another. This setup had to handle heavy Input/Output (I/O) operations from FTP, Samba, and web traffic while adhering to strict security protocols for remote execution.
Our Strategic Approach
Skynats approached this challenge with a focus on open-source reliability and performance under load:
- Technology Evaluation: We rigorously tested various open-source replication systems. GlusterFS emerged as the clear winner for its scalability and handle on high I/O loads.
- Architecture Design: We designed a trusted storage pool where servers act as peers, eliminating single points of failure.
- Stress Testing: Before production, we simulated heavy FTP and Web traffic in a staging environment to fine-tune latency and throughput.
- Implementation: The finalized, optimized architecture was deployed to the OVH bare-metal infrastructure.
Technical Implementation Details
The solution was built on a trusted peer architecture using GlusterFS:
- Trusted Storage Pool: Configured two Linux servers as a trusted peer probe, ensuring secure communication channels.
- Brick & Volume Creation: Defining storage 'bricks' on each server and creating a replicated volumes across them.
- Mounting: The replicated volume was mounted as a local directory on both servers, making the replication transparent to applications.
- Automatic Failover: If one server goes down, the other continues to serve data without interruption. Upon recovery, the system automatically self-heals by syncing missed data.
- Performance Tuning: Optimized GlusterFS parameters to handle the specific small-file heavy workload of the client's web applications.
The Outcome
The project was delivered in just 2 weeks. The system has now been running in production for over a year with zero data discrepancies. It automatically heals from network blips and server restarts, providing a maintenance-free, high-availability storage layer for critical healthcare data.