Stop letting your NIC buffer overflows kill your packet throughput
Performance Tuning & Kernel Parameters (Sysctl)
Stop letting your NIC buffer overflows kill your packet throughput
🧩 The Challenge
You look at netstat and see zero errors, yet your high-throughput app is dropping packets like it’s going out of style. I spent an entire weekend staring at logs before realizing the NIC hardware buffers were just getting hammered by incoming spikes.
💡 The Fix
Bumping up the ring buffer size via ethtool gives the kernel more breathing room to process frames during traffic bursts. You’ll want to check what your specific NIC supports first, since hardware limits vary wildly.
ethtool -g eth0
ethtool -G eth0 rx 4096 tx 4096
⚙️ Why It Works
Expanding these rings provides a deeper staging area for packets, preventing the interface from tossing them when the CPU is busy handling other interrupts. Most defaults are absurdly conservative for modern 10G or 25G links.
🚀 Pro-Tip: Always verify the change sticks after a reboot by adding it to your interface configuration or a post-up script.
Linux Tips & Tricks | © ngelinux.com | 10/30/2026
