node_exporter/docs/node-mixin/rules/rules.libsonnet
Ben Kochie d90b2d83d7
Replace load average with PSI metric
The load average metric is misleading as a representation of CPU
saturation. Normal CPU utilization is a better real representation of
saturation.

On newer Linux, there is a new Pressure Stall Information[0] metric that
better represents CPU over saturation. This is also useful as it can
make single-core saturation more visible.

[0]: https://www.kernel.org/doc/html/latest/accounting/psi.html

Signed-off-by: Ben Kochie <superq@gmail.com>
2021-08-19 12:19:26 +02:00

107 lines
4.3 KiB
Plaintext

{
prometheusRules+:: {
groups+: [
{
name: 'node-exporter.rules',
rules: [
{
// This rule gives the number of CPUs per node.
record: 'instance:node_num_cpu:sum',
expr: |||
count without (cpu) (
count without (mode) (
node_cpu_seconds_total{%(nodeExporterSelector)s}
)
)
||| % $._config,
},
{
// CPU utilisation is % CPU is not idle. This represents CPU saturation.
record: 'instance:node_cpu_utilisation:rate%(rateInterval)s' % $._config,
expr: |||
1 - avg without (cpu, mode) (
rate(node_cpu_seconds_total{%(nodeExporterSelector)s, mode="idle"}[%(rateInterval)s])
)
||| % $._config,
},
{
// CPU pressure represents over-saturation. This is the amount of CPU seconds
// requested, but the kernel was not able to schedule.
// NOTE: This is only availalbe on Linux >= 4.19 and `CONFIG_PSI` is enabled.
// See also:
// - https://www.kernel.org/doc/html/latest/accounting/psi.html
// - https://facebookmicrosites.github.io/psi/docs/overview
expr: |||
rate(node_pressure_cpu_waiting_seconds_total{%(nodeExporterSelector)s}[%(rateInterval)s])
||| % $._config,
},
{
// Memory utilisation (ratio of used memory per instance).
record: 'instance:node_memory_utilisation:ratio',
expr: |||
1 - (
node_memory_MemAvailable_bytes{%(nodeExporterSelector)s}
/
node_memory_MemTotal_bytes{%(nodeExporterSelector)s}
)
||| % $._config,
},
{
record: 'instance:node_vmstat_pgmajfault:rate%(rateInterval)s' % $._config,
expr: |||
rate(node_vmstat_pgmajfault{%(nodeExporterSelector)s}[%(rateInterval)s])
||| % $._config,
},
{
// Disk utilisation (seconds spent, 1 second rate).
record: 'instance_device:node_disk_io_time_seconds:rate%(rateInterval)s' % $._config,
expr: |||
rate(node_disk_io_time_seconds_total{%(nodeExporterSelector)s, %(diskDeviceSelector)s}[%(rateInterval)s])
||| % $._config,
},
{
// Disk saturation (weighted seconds spent, 1 second rate).
record: 'instance_device:node_disk_io_time_weighted_seconds:rate%(rateInterval)s' % $._config,
expr: |||
rate(node_disk_io_time_weighted_seconds_total{%(nodeExporterSelector)s, %(diskDeviceSelector)s}[%(rateInterval)s])
||| % $._config,
},
{
record: 'instance:node_network_receive_bytes_excluding_lo:rate%(rateInterval)s' % $._config,
expr: |||
sum without (device) (
rate(node_network_receive_bytes_total{%(nodeExporterSelector)s, device!="lo"}[%(rateInterval)s])
)
||| % $._config,
},
{
record: 'instance:node_network_transmit_bytes_excluding_lo:rate%(rateInterval)s' % $._config,
expr: |||
sum without (device) (
rate(node_network_transmit_bytes_total{%(nodeExporterSelector)s, device!="lo"}[%(rateInterval)s])
)
||| % $._config,
},
// TODO: Find out if those drops ever happen on modern switched networks.
{
record: 'instance:node_network_receive_drop_excluding_lo:rate%(rateInterval)s' % $._config,
expr: |||
sum without (device) (
rate(node_network_receive_drop_total{%(nodeExporterSelector)s, device!="lo"}[%(rateInterval)s])
)
||| % $._config,
},
{
record: 'instance:node_network_transmit_drop_excluding_lo:rate%(rateInterval)s' % $._config,
expr: |||
sum without (device) (
rate(node_network_transmit_drop_total{%(nodeExporterSelector)s, device!="lo"}[%(rateInterval)s])
)
||| % $._config,
},
],
},
],
},
}