Networking & Firewall (Ss/Netstat/Iptables/Nftables/Curl)

Stop mystery packet drops when your buffer sizes are too small

🧩 The Challenge

Dealing with a production app that suddenly stops talking to the world when traffic spikes is a special kind of nightmare. You end up staring at tcpdump for hours only to realize the OS is just dumping packets on the floor before they even hit your application.

💡 The Fix

You need to peek at the interface statistics to see if the rx_dropped counter is ticking up, which usually points to a receive queue overflow. Bumping up the ring buffer settings will give your interface a bit more breathing room during those sudden bursts.

ethtool -g eth0
ethtool -G eth0 rx 4096

⚙️ Why It Works

Modern high-speed NICs often ship with conservative default ring buffer sizes that choke once you hit a decent amount of concurrent traffic. By increasing these values, you’re telling the kernel to allocate more memory to buffer incoming packets, preventing them from being dropped at the hardware level.

🚀 Pro-Tip: Always check your kernel ring buffer with dmesg | grep eth0 too, because sometimes the hardware will tell you exactly why it’s puking up packets.

Linux Tips & Tricks | © ngelinux.com | 8/27/2026