Learn to watch Linux interrupt counters, spot noisy devices and rebalance IRQs before they throttle your system. 29.11.2025 | reading time: 2 min Want to know which device is hammering your CPUs? Learn to read `cat /proc/interrupts` and use `watch` to observe realtime changes so you can identify IRQ storms and act quickly. A noisy NIC: reproduce and observe Run a quick live check to see interrupts move: ```bash $ sudo watch -n1 cat /proc/interrupts CPU0 CPU1 16: 123456 102 PCI-MSI-edge eth0 17: 234 0 IO-APIC-edge timer 18: 3456 2000 PCI-MSI-edge mlx5 ```Watch will refresh the counts every second and you immediately see which IRQ line increases fastest. Interpret the counts Columns map to CPUs and each line is an IRQ vector or chip line; high, rapidly rising counts on one vector point at the offending device; check `/proc/softirqs` to distinguish softirq activity from hardware IRQs. Tune IRQ handling If a NIC saturates one CPU, adjust affinity or enable balancing: set `/proc/irq/<N>/smp_affinity` to a hex mask or start `irqbalance`; for example change a vector to CPU1 with `echo 2 > /proc/irq/45/smp_affinity` but test carefully on production systems. Other diagnostics to try Use `dmesg` for driver errors, `mpstat -I ALL 1` to watch per-interrupt CPU load, and `perf` to profile interrupt context if counts alone don't reveal the root cause. Next steps Monitoring interrupts is a small skill with big payoff: once you can spot and rebalance IRQs you reduce latency and improve throughput; keep practicing and consider formal study to master Linux operations at scale with courses like bitsandbytes.academy that prepare for CompTIA Linux+ or LPIC-1. Join Bits & Bytes Academy First class LINUX exam preparation. utilities processes troubleshooting boot-process infrastructure
