Lightweight system metrics collector written in Zig 0.15.2. Reads kernel metrics from /proc and /sys, stores them in SQLite with JSON payloads.

Collected Metrics#

Build#

# Requires Zig 0.15.2 and libsqlite3-dev
zig build

# Release build (static, musl)
zig build -Doptimize=ReleaseSafe -Dtarget=x86_64-linux-musl

Usage#

# Default: writes to /var/lib/sysmon/metrics.db every 5 seconds
sudo mkdir -p /var/lib/sysmon
sudo sysmon

# Custom path and interval
sysmon --db ./metrics.db --interval 10

Query#

# Use the included helper
chmod +x sysmon-query.sh
SYSMON_DB=./metrics.db ./sysmon-query.sh --summary

# Or query directly
sqlite3 metrics.db "SELECT datetime(ts,'unixepoch','localtime'), json
                     FROM metrics WHERE kind='cpu'
                     ORDER BY ts DESC LIMIT 5;"

Schema#

CREATE TABLE metrics (
    ts      INTEGER NOT NULL,   -- unix epoch
    kind    TEXT NOT NULL,       -- cpu, load, mem, net, diskio, diskspace
    device  TEXT NOT NULL DEFAULT '',  -- eth0, sda, /, ...
    json    TEXT NOT NULL        -- metric payload
);

JSON payloads keep the schema extensible – no migrations needed when adding container metrics later.

Architecture Notes#

• Delta-based metrics (CPU, net, disk I/O): First sample establishes baseline, rates appear from the second tick onward.
• SQLite WAL mode: Allows concurrent reads while the collector writes.
• Batch inserts: All metrics per tick wrapped in a single transaction.
• Writergate (Zig 0.15): All I/O uses explicit buffers + flush.
• Zero allocations in the hot path: Stack buffers throughout.

Future: Container Metrics#

Container CPU/memory/IO from cgroups v2:

/sys/fs/cgroup/<container>/cpu.stat
/sys/fs/cgroup/<container>/memory.current
/sys/fs/cgroup/<container>/io.stat

The JSON payload approach makes adding these trivial – just add a new kind value (e.g. container_cpu) with the container name as device.