What causes "failed to get clock ticks per second: %v" in prometheus/node_exporter?

Verify the AIX system's CLK_TCK configuration is sane (getconf CLK_TCK should return a positive number)

failed to get clock ticks per second

Error message

failed to get clock ticks per second: %v

What it means

On AIX, tickPerSecond calls the libc sysconf(_SC_CLK_TCK) via cgo to determine clock ticks per second, needed to convert CPU time samples into seconds. A return value of -1 or a non-nil error means the system call failed, so the CPU collector cannot be created correctly and this error is propagated.

Solutions

Verify the AIX system's CLK_TCK configuration is sane (getconf CLK_TCK should return a positive number)
Rebuild with a working cgo toolchain for AIX
If unresolved, disable the cpu collector on that host

Example fix

// before
ticks, err := C.sysconf(C._SC_CLK_TCK) // -1
// after
check `getconf CLK_TCK` on the host; fix OS config or upgrade AIX before running the exporter

Defensive patterns

Strategy: try-catch

Try / catch

ticks, err := tickPerSecond()
if err != nil {
    // verify OS-level configuration before retrying
    out, _ := exec.LookPath("getconf")
    _ = out
    log.Error("CLK_TCK unavailable; cpu metrics disabled", "err", err)
    return nil
}

Prevention

Verify `getconf CLK_TCK` returns a sane value on target AIX hosts before deploying
Keep the AIX cgo toolchain and OS patches current
Treat repeated sysconf failures as an OS-level problem, not an exporter bug

When it happens

Trigger: C.sysconf(C._SC_CLK_TCK) returns -1 or sets errno during NewCpuCollector on AIX.

Common situations: Corrupted/unusual system configuration on AIX; cgo environment issues at runtime; running an AIX-targeted build on an environment where sysconf misbehaves (rare, mostly a platform-level fault).

Related errors

AI-assisted analysis of prometheus/node_exporter@17ddd77c59 (2026-09-07). Data as JSON: /api/errors/a8c66de6b401b29d. Report an issue: GitHub.

Appendix: source

Thrown at collector/cpu_aix.go:73

	cpu              typedDesc
	cpuPhysical      typedDesc
	cpuRunQueue      typedDesc
	cpuFlags         typedDesc
	cpuContextSwitch typedDesc

	logger             *slog.Logger
	tickPerSecond      float64
	purrTicksPerSecond float64
}

func init() {
	registerCollector("cpu", defaultEnabled, NewCpuCollector)
}

func tickPerSecond() (float64, error) {
	ticks, err := C.sysconf(C._SC_CLK_TCK)
	if ticks == -1 || err != nil {
		return 0, fmt.Errorf("failed to get clock ticks per second: %v", err)
	}
	return float64(ticks), nil
}

func NewCpuCollector(logger *slog.Logger) (Collector, error) {
	ticks, err := tickPerSecond()
	if err != nil {
		return nil, err
	}

	pconfig, err := perfstat.PartitionStat()

	if err != nil {
		return nil, err
	}

	return &cpuCollector{
		cpu:                typedDesc{nodeCPUSecondsDesc, prometheus.CounterValue},

View on GitHub (pinned to 17ddd77c59)