33 KiB

Raw Permalink Blame History

description

Intel PowerStat Input Plugin

This input plugin monitors power statistics on Intel-based platforms and assumes presence of Linux based OS.

Not all CPUs are supported, please see the software and hardware dependencies sections below to ensure platform support.

Main use cases are power saving and workload migration. Telemetry frameworks allow users to monitor critical platform level metrics. Key source of platform telemetry is power domain that is beneficial for MANO Monitoring&Analytics systems to take preventive/corrective actions based on platform busyness, CPU temperature, actual CPU utilization and power statistics.

Global configuration options

In addition to the plugin-specific configuration settings, plugins support additional global and plugin configuration settings. These settings are used to modify metrics, tags, and field or create aliases and configure ordering, etc. See the CONFIGURATION.md for more details.

Configuration

# Intel PowerStat plugin enables monitoring of platform metrics (power, TDP)
# and per-CPU metrics like temperature, power and utilization. Please see the
# plugin readme for details on software and hardware compatibility.
# This plugin ONLY supports Linux.
[[inputs.intel_powerstat]]
  ## The user can choose which package metrics are monitored by the plugin with
  ## the package_metrics setting:
  ## - The default, will collect "current_power_consumption",
  ##   "current_dram_power_consumption" and "thermal_design_power".
  ## - Leaving this setting empty means no package metrics will be collected.
  ## - Finally, a user can specify individual metrics to capture from the
  ##   supported options list.
  ## Supported options:
  ##   "current_power_consumption", "current_dram_power_consumption",
  ##   "thermal_design_power", "max_turbo_frequency", "uncore_frequency",
  ##   "cpu_base_frequency"
  # package_metrics = ["current_power_consumption", "current_dram_power_consumption", "thermal_design_power"]

  ## The user can choose which per-CPU metrics are monitored by the plugin in
  ## cpu_metrics array.
  ## Empty or missing array means no per-CPU specific metrics will be collected
  ## by the plugin.
  ## Supported options:
  ##   "cpu_frequency", "cpu_c0_state_residency", "cpu_c1_state_residency",
  ##   "cpu_c3_state_residency", "cpu_c6_state_residency", "cpu_c7_state_residency",
  ##   "cpu_temperature", "cpu_busy_frequency", "cpu_c0_substate_c01",
  ##   "cpu_c0_substate_c02", "cpu_c0_substate_c0_wait"
  # cpu_metrics = []

  ## Optionally the user can choose for which CPUs metrics configured in cpu_metrics array should be gathered.
  ## Can't be combined with excluded_cpus.
  ## Empty or missing array means CPU metrics are gathered for all CPUs.
  ## e.g. ["0-3", "4,5,6"] or ["1-3,4"]
  # included_cpus = []

  ## Optionally the user can choose which CPUs should be excluded from gathering metrics configured in cpu_metrics array.
  ## Can't be combined with included_cpus.
  ## Empty or missing array means CPU metrics are gathered for all CPUs.
  ## e.g. ["0-3", "4,5,6"] or ["1-3,4"]
  # excluded_cpus = []

  ## Filesystem location of JSON file that contains PMU event definitions.
  ## Mandatory only for perf-related metrics (cpu_c0_substate_c01, cpu_c0_substate_c02, cpu_c0_substate_c0_wait).
  # event_definitions = ""

  ## The user can set the timeout duration for MSR reading.
  ## Enabling this timeout can be useful in situations where, on heavily loaded systems,
  ## the code waits too long for a kernel response to MSR read requests.
  ## 0 disables the timeout (default).
  # msr_read_timeout = "0ms"

Configuration notes

The configuration of included_cpus or excluded_cpus may affect the ability to collect package_metrics. Some of them (max_turbo_frequency, cpu_base_frequency, and uncore_frequency) need to read data from exactly one processor for each package. If included_cpus or excluded_cpus exclude all processors from the package, reading the mentioned metrics for that package will not be possible.
event_definitions JSON file for specific architecture can be found at perfmon. A script to download the event definition that is appropriate for current environment (event_download.py) is available at pmu-tools. For perf-related metrics supported by this plugin, an event definition JSON file with events for the core is required.

For example: sapphirerapids_core.json or GenuineIntel-6-8F-core.json.

Example: Configuration with no per-CPU telemetry

This configuration allows getting default processor package specific metrics, no per-CPU metrics are collected:

[[inputs.intel_powerstat]]
  cpu_metrics = []

Example: Configuration with no per-CPU telemetry - equivalent case

This configuration allows getting default processor package specific metrics, no per-CPU metrics are collected:

[[inputs.intel_powerstat]]

Example: Configuration for CPU Temperature and CPU Frequency

This configuration allows getting default processor package specific metrics, plus subset of per-CPU metrics (CPU Temperature and CPU Frequency) which will be gathered only for cpu_id = 0:

[[inputs.intel_powerstat]]
  cpu_metrics = ["cpu_frequency", "cpu_temperature"]
  included_cpus = ["0"]

Example: Configuration for CPU Temperature and CPU Frequency without default package metrics

This configuration allows getting only a subset of per-CPU metrics (CPU Temperature and CPU Frequency) which will be gathered for all cpus except cpu_id = ["1-3"]:

[[inputs.intel_powerstat]]
  package_metrics = []
  cpu_metrics = ["cpu_frequency", "cpu_temperature"]
  excluded_cpus = ["1-3"]

Example: Configuration with all available metrics

This configuration allows getting all processor package specific metrics and all per-CPU metrics:

[[inputs.intel_powerstat]]
  package_metrics = ["current_power_consumption", "current_dram_power_consumption", "thermal_design_power", "max_turbo_frequency", "uncore_frequency", "cpu_base_frequency"]
  cpu_metrics = ["cpu_frequency", "cpu_c0_state_residency", "cpu_c1_state_residency", "cpu_c3_state_residency", "cpu_c6_state_residency", "cpu_c7_state_residency", "cpu_temperature", "cpu_busy_frequency", "cpu_c0_substate_c01", "cpu_c0_substate_c02", "cpu_c0_substate_c0_wait"]
  event_definitions = "/home/telegraf/.cache/pmu-events/GenuineIntel-6-8F-core.json"

SW Dependencies

Kernel modules

Plugin is mostly based on Linux Kernel modules that expose specific metrics over sysfs or devfs interfaces. The following dependencies are expected by plugin:

intel-rapl kernel module which exposes Intel Runtime Power Limiting metrics over sysfs (/sys/devices/virtual/powercap/intel-rapl),
msr kernel module that provides access to processor model specific registers over devfs (/dev/cpu/cpu%d/msr),
cpufreq kernel module - which exposes per-CPU Frequency over sysfs (/sys/devices/system/cpu/cpu%d/cpufreq/scaling_cur_freq),
intel-uncore-frequency kernel module exposes Intel uncore frequency metrics over sysfs (/sys/devices/system/cpu/intel_uncore_frequency).

Make sure that required kernel modules are loaded and running. Modules might have to be manually enabled by using modprobe. Depending on the kernel version, run commands:

# rapl modules:
## kernel < 4.0
sudo modprobe intel_rapl
## kernel >= 4.0
sudo modprobe rapl
sudo modprobe intel_rapl_common
sudo modprobe intel_rapl_msr

# msr module:
sudo modprobe msr

# cpufreq module:
### integrated in kernel 

# intel-uncore-frequency module:
## only for kernel >= 5.6.0
sudo modprobe intel-uncore-frequency

Kernel's perf interface

For perf-related metrics, when Telegraf is not running as root, the following capability should be added to the Telegraf executable:

sudo setcap cap_sys_admin+ep <path_to_telegraf_binary>

Alternatively, /proc/sys/kernel/perf_event_paranoid has to be set to value less than 1.

Depending on environment and configuration (number of monitored CPUs and number of enabled metrics), it might be required to increase the limit on the number of open file descriptors allowed. This can be done for example by using ulimit -n command.

Dependencies of metrics on system configuration

Details of these dependencies are discussed above:

Configuration option	Type	Dependency
`current_power_consumption`	`package_metrics`	`rapl` kernel module(s)
`current_dram_power_consumption`	`package_metrics`	`rapl` kernel module(s)
`thermal_design_power`	`package_metrics`	`rapl` kernel module(s)
`max_turbo_frequency`	`package_metrics`	`msr` kernel module
`uncore_frequency`	`package_metrics`	`intel-uncore-frequency`/`msr` kernel modules*
`cpu_base_frequency`	`package_metrics`	`msr` kernel module
`cpu_frequency`	`cpu_metrics`	`cpufreq` kernel module
`cpu_c0_state_residency`	`cpu_metrics`	`msr` kernel module
`cpu_c1_state_residency`	`cpu_metrics`	`msr` kernel module
`cpu_c3_state_residency`	`cpu_metrics`	`msr` kernel module
`cpu_c6_state_residency`	`cpu_metrics`	`msr` kernel module
`cpu_c7_state_residency`	`cpu_metrics`	`msr` kernel module
`cpu_busy_cycles` (DEPRECATED - superseded by `cpu_c0_state_residency_percent`)	`cpu_metrics`	`msr` kernel module
`cpu_temperature`	`cpu_metrics`	`msr` kernel module
`cpu_busy_frequency`	`cpu_metrics`	`msr` kernel module
`cpu_c0_substate_c01`	`cpu_metrics`	kernel's `perf` interface
`cpu_c0_substate_c02`	`cpu_metrics`	kernel's `perf` interface
`cpu_c0_substate_c0_wait`	`cpu_metrics`	kernel's `perf` interface

*for all metrics enabled by the configuration option uncore_frequency, starting from kernel version 5.18, only the intel-uncore-frequency module is required. For older kernel versions, the metric uncore_frequency_mhz_cur requires the msr module to be enabled.

Root privileges

Telegraf with Intel PowerStat plugin enabled may require root privileges to read all the metrics (depending on OS type or configuration).

Alternatively, the following capabilities can be added to the Telegraf executable:

#without perf-related metrics:
sudo setcap cap_sys_rawio,cap_dac_read_search+ep <path_to_telegraf_binary>

#with perf-related metrics:
sudo setcap cap_sys_rawio,cap_dac_read_search,cap_sys_admin+ep <path_to_telegraf_binary>

HW Dependencies

Specific metrics require certain processor features to be present, otherwise Intel PowerStat plugin won't be able to read them. The user can detect supported processor features by reading /proc/cpuinfo file. Plugin assumes crucial properties are the same for all CPU cores in the system.

The following processor properties are examined in more detail in this section:

vendor_id
cpu family
model
flags

The following processor properties are required by the plugin:

Processor vendor_id must be GenuineIntel and cpu family must be 6 - since data used by the plugin are Intel-specific.
The following processor flags shall be present:
- msr shall be present for plugin to read platform data from processor model specific registers and collect the following metrics:
  - cpu_c0_state_residency
  - cpu_c1_state_residency
  - cpu_c3_state_residency
  - cpu_c6_state_residency
  - cpu_c7_state_residency
  - cpu_busy_cycles (DEPRECATED - superseded by cpu_c0_state_residency_percent)
  - cpu_busy_frequency
  - cpu_temperature
  - cpu_base_frequency
  - max_turbo_frequency
  - uncore_frequency (for kernel < 5.18)
- aperfmperf shall be present to collect the following metrics:
  - cpu_c0_state_residency
  - cpu_c1_state_residency
  - cpu_busy_cycles (DEPRECATED - superseded by cpu_c0_state_residency_percent)
  - cpu_busy_frequency
- dts shall be present to collect:
  - cpu_temperature
Please consult the table of supported CPU models for the processor package. | MHz |

Known issues

Starting from Linux kernel version v5.4.77, due to this kernel change, resources such as /sys/devices/virtual/powercap/intel-rapl//*/energy_uj can only be accessed by the root user for security reasons. Therefore, this plugin requires root privileges to gather rapl metrics correctly.

If such strict security restrictions are not relevant, reading permissions for files in the /sys/devices/virtual/powercap/intel-rapl/ directory can be manually altered, for example, using the chmod command with custom parameters. For instance, read and execute permissions for all files in the intel-rapl directory can be granted to all users using:

sudo chmod -R a+rx /sys/devices/virtual/powercap/intel-rapl/

Example Output

powerstat_package,host=ubuntu,package_id=0 thermal_design_power_watts=160 1606494744000000000
powerstat_package,host=ubuntu,package_id=0 current_power_consumption_watts=35 1606494744000000000
powerstat_package,host=ubuntu,package_id=0 cpu_base_frequency_mhz=2400i 1669118424000000000
powerstat_package,host=ubuntu,package_id=0 current_dram_power_consumption_watts=13.94 1606494744000000000
powerstat_package,host=ubuntu,package_id=0,active_cores=0 max_turbo_frequency_mhz=3000i 1606494744000000000
powerstat_package,host=ubuntu,package_id=0,active_cores=1 max_turbo_frequency_mhz=2800i 1606494744000000000
powerstat_package,die=0,host=ubuntu,package_id=0,type=initial uncore_frequency_limit_mhz_min=800,uncore_frequency_limit_mhz_max=2400 1606494744000000000
powerstat_package,die=0,host=ubuntu,package_id=0,type=current uncore_frequency_mhz_cur=800i,uncore_frequency_limit_mhz_min=800,uncore_frequency_limit_mhz_max=2400 1606494744000000000
powerstat_core,core_id=0,cpu_id=0,host=ubuntu,package_id=0 cpu_frequency_mhz=1200.29 1606494744000000000
powerstat_core,core_id=0,cpu_id=0,host=ubuntu,package_id=0 cpu_temperature_celsius=34i 1606494744000000000
powerstat_core,core_id=0,cpu_id=0,host=ubuntu,package_id=0 cpu_c0_state_residency_percent=0.8 1606494744000000000
powerstat_core,core_id=0,cpu_id=0,host=ubuntu,package_id=0 cpu_c1_state_residency_percent=6.68 1606494744000000000
powerstat_core,core_id=0,cpu_id=0,host=ubuntu,package_id=0 cpu_c3_state_residency_percent=0 1606494744000000000
powerstat_core,core_id=0,cpu_id=0,host=ubuntu,package_id=0 cpu_c6_state_residency_percent=92.52 1606494744000000000
powerstat_core,core_id=0,cpu_id=0,host=ubuntu,package_id=0 cpu_c7_state_residency_percent=0 1606494744000000000
powerstat_core,core_id=0,cpu_id=0,host=ubuntu,package_id=0 cpu_busy_frequency_mhz=1213.24 1606494744000000000
powerstat_core,core_id=0,cpu_id=0,host=ubuntu,package_id=0 cpu_c0_substate_c01_percent=0 1606494744000000000
powerstat_core,core_id=0,cpu_id=0,host=ubuntu,package_id=0 cpu_c0_substate_c02_percent=5.68 1606494744000000000
powerstat_core,core_id=0,cpu_id=0,host=ubuntu,package_id=0 cpu_c0_substate_c0_wait_percent=43.74 1606494744000000000

Supported CPU models

Model number	Processor name	`cpu_c1_state_residency` `cpu_c6_state_residency` `cpu_temperature` `cpu_base_frequency`	`cpu_c3_state_residency`	`cpu_c7_state_residency`	`uncore_frequency`
0x1E	Intel Nehalem	✓	✓
0x1F	Intel Nehalem-G	✓	✓
0x1A	Intel Nehalem-EP	✓	✓
0x2E	Intel Nehalem-EX	✓	✓
0x25	Intel Westmere	✓	✓
0x2C	Intel Westmere-EP	✓	✓
0x2F	Intel Westmere-EX	✓	✓
0x2A	Intel Sandybridge	✓	✓	✓
0x2D	Intel Sandybridge-X	✓	✓	✓
0x3A	Intel Ivybridge	✓	✓	✓
0x3E	Intel Ivybridge-X	✓	✓	✓
0x3C	Intel Haswell	✓	✓	✓
0x3F	Intel Haswell-X	✓	✓	✓
0x45	Intel Haswell-L	✓	✓	✓
0x46	Intel Haswell-G	✓	✓	✓
0x3D	Intel Broadwell	✓	✓	✓
0x47	Intel Broadwell-G	✓	✓	✓	✓
0x4F	Intel Broadwell-X	✓	✓		✓
0x56	Intel Broadwell-D	✓	✓		✓
0x4E	Intel Skylake-L	✓	✓	✓
0x5E	Intel Skylake	✓	✓	✓
0x55	Intel Skylake-X	✓			✓
0x8E	Intel KabyLake-L	✓	✓	✓
0x9E	Intel KabyLake	✓	✓	✓
0xA5	Intel CometLake	✓	✓	✓
0xA6	Intel CometLake-L	✓	✓	✓
0x66	Intel CannonLake-L	✓		✓
0x6A	Intel IceLake-X	✓			✓
0x6C	Intel IceLake-D	✓			✓
0x7D	Intel IceLake	✓
0x7E	Intel IceLake-L	✓		✓
0x9D	Intel IceLake-NNPI	✓		✓
0xA7	Intel RocketLake	✓		✓
0x8C	Intel TigerLake-L	✓		✓
0x8D	Intel TigerLake	✓		✓
0x8F	Intel Sapphire Rapids X	✓			✓
0xCF	Intel Emerald Rapids X	✓			✓
0xAD	Intel Granite Rapids X	✓
0xAE	Intel Granite Rapids D	✓
0x8A	Intel Lakefield	✓		✓
0x97	Intel AlderLake	✓		✓	✓
0x9A	Intel AlderLake-L	✓		✓	✓
0xB7	Intel RaptorLake	✓		✓	✓
0xBA	Intel RaptorLake-P	✓		✓	✓
0xBF	Intel RaptorLake-S	✓		✓	✓
0xAC	Intel MeteorLake	✓		✓	✓
0xAA	Intel MeteorLake-L	✓		✓	✓
0xC6	Intel ArrowLake	✓		✓
0xBD	Intel LunarLake	✓		✓
0x37	Intel Atom® Bay Trail	✓
0x4D	Intel Atom® Avaton	✓
0x4A	Intel Atom® Merrifield	✓
0x5A	Intel Atom® Moorefield	✓
0x4C	Intel Atom® Airmont	✓	✓
0x5C	Intel Atom® Apollo Lake	✓	✓	✓
0x5F	Intel Atom® Denverton	✓
0x7A	Intel Atom® Goldmont	✓	✓	✓
0x86	Intel Atom® Jacobsville	✓
0x96	Intel Atom® Elkhart Lake	✓		✓
0x9C	Intel Atom® Jasper Lake	✓		✓
0xBE	Intel AlderLake-N	✓		✓
0xAF	Intel Sierra Forest	✓
0xB6	Intel Grand Ridge	✓
0x57	Intel Xeon® PHI Knights Landing	✓
0x85	Intel Xeon® PHI Knights Mill	✓

33 KiB Raw Permalink Blame History