proctap scrapes high-signal, low-overhead metrics straight from /proc and /sys, and exposes them at GET /metrics for Prometheus scraping. Built in Rust with axum and the Prometheus client crate.

• diskstat — Block device I/O stats from /sys/class/block//stat.
• ethtool — Driver/firmware NIC counters via ethtool -S <iface>.
• interrupts — IRQ distribution from /proc/interrupts (per IRQ × CPU).
• memstat — Memory info from /proc/meminfo.
• netdev_stat — Per-interface NIC stats from /sys/class/net//statistics/*.
• proc — Process scheduler stats from /proc//sched (per matching process).
• queues — Per-queue NIC sysfs values from /sys/class/net//queues/{rx,tx}-*/.
• snmp — Kernel TCP/UDP/IP counters from /proc/net/snmp.
• softirqs — SoftIRQ totals from /proc/softirqs (per CPU).
• softnet_stat — Per-CPU networking backlog stats from /proc/net/softnet_stat.

# Requires Rust (stable)
cargo build --release

# Run the exporter
./target/release/proctap \
  --interval 5 \
  --proc-prefixes ping

Scrape at: http://<host>:9000/metrics.

Finding the right --proc-prefixes: it matches Linux comm (truncated to 15 chars). Check with: cat /proc/<PID>/comm Example: if you run ping 127.0.0.1, the comm is likely ping.

Note: Linux truncates comm to 15 chars.