mirror of
https://github.com/prometheus-community/windows_exporter.git
synced 2026-10-18 18:49:05 +02:00
Jan-Otto KröpkeandClaude
da7bebc6b3
Co-authored-by: Claude <noreply@anthropic.com> Signed-off-by: Claude <noreply@anthropic.com>
83 lines
4.6 KiB
Markdown
83 lines
4.6 KiB
Markdown
# `cpu` collector
|
|
|
|
The collector reads the multi-instance PDH object `Processor Information`
|
|
with `pdh.CounterTypeRaw`. The object exists on every Windows host, so no role
|
|
is needed. These notes were recorded on Windows Server 2025 (build 26100), a
|
|
Hyper-V guest with 12 logical processors in one processor group.
|
|
|
|
- Instances are `<group>,<processor>`, for example `0,0` to `0,11`. Besides
|
|
`_Total` the object has one aggregate per processor group, `<group>,_Total`
|
|
(here `0,_Total`). `pdh.IsTotalInstance` treats both as totals, so with
|
|
`pdh.InstancesAll` they are dropped and only per-processor rows remain. The
|
|
`core` label is the raw instance name, e.g. `0,3`.
|
|
- `% C1/C2/C3 Time`, `% DPC Time`, `% Idle Time`, `% Interrupt Time`,
|
|
`% Privileged Time` and `% User Time` are `PERF_100NSEC_TIMER`. Their raw
|
|
value is cumulative 100 ns ticks; `internal/pdh` scales them to seconds.
|
|
These raw values can drop by one clock tick (1/64 s) between two samples
|
|
([#1383](https://github.com/prometheus-community/windows_exporter/issues/1383)),
|
|
so `Collect` holds the previous value of `time_seconds_total` and
|
|
`cstate_seconds_total` for drops below `maxTimeDrop` (1 s).
|
|
- `% Processor Time` and `% Priority Time` are `PERF_100NSEC_TIMER_INV`. The
|
|
raw value is cumulative idle (non-busy) time in 100 ns ticks, not busy time.
|
|
`internal/pdh` scales it to seconds like the other 100 ns timers. In the
|
|
sample, the raw value of `% Processor Time` equals `% Idle Time`.
|
|
- `% Processor Utility`, `% Processor Performance` and `% Privileged Utility`
|
|
are `PERF_AVERAGE_BULK`. Raw and base are both cumulative. The displayed
|
|
value is delta raw / delta base and is already a percentage (it can exceed
|
|
100). A 2 s sample on core `0,0` gave 63.4 for utility and 121.9 for
|
|
performance. The raw values of utility and performance are almost equal;
|
|
their bases differ. `% Processor Utility` and `% Privileged Utility` have
|
|
nearly equal bases that grew about 630 000 per second; the performance
|
|
base grew about 330 000 per second.
|
|
- `Interrupts/sec`, `DPCs Queued/sec` and `Clock Interrupts/sec`
|
|
(`PERF_COUNTER_COUNTER`) and `C1/C2/C3 Transitions/sec`,
|
|
`Idle Break Events/sec` (`PERF_COUNTER_BULK_COUNT`) are cumulative counts.
|
|
- `Processor Frequency` is a gauge in MHz (3700 here); the collector multiplies
|
|
it by 1e6 for `windows_cpu_core_frequency_hertz`. `Parking Status` is a
|
|
gauge, 0 when the processor is not parked. `% Performance Limit` is a gauge
|
|
percentage (100 here). `% C3 Time` and `C3 Transitions/sec` stayed 0 on this
|
|
virtual machine.
|
|
|
|
## Counter reference
|
|
|
|
Recorded on Windows Server 2025 (build 26100). `Raw` and `Base` are the first and second
|
|
values that `pdh.CounterTypeRaw` returns for the sample instance.
|
|
|
|
### `Processor Information`
|
|
|
|
Multi-instance. Instances: `_Total`, `0,_Total`, `0,0`, `0,1`, `0,10`, `0,11`, `0,2`, `0,3`, ... (14 in total). Sample instance: `0,0`.
|
|
|
|
Counters the collector reads:
|
|
|
|
| Counter | Type | Raw | Base |
|
|
|---|---|---|---|
|
|
| `% C1 Time` | `PERF_100NSEC_TIMER` | 3366958649 | |
|
|
| `% C2 Time` | `PERF_100NSEC_TIMER` | 21091589064 | |
|
|
| `% C3 Time` | `PERF_100NSEC_TIMER` | 0 | |
|
|
| `% DPC Time` | `PERF_100NSEC_TIMER` | 185312500 | |
|
|
| `% Idle Time` | `PERF_100NSEC_TIMER` | 25109218750 | |
|
|
| `% Interrupt Time` | `PERF_100NSEC_TIMER` | 257812500 | |
|
|
| `% Performance Limit` | `PERF_COUNTER_RAWCOUNT` | 100 | |
|
|
| `% Priority Time` | `PERF_100NSEC_TIMER_INV` | 26964687500 | |
|
|
| `% Privileged Time` | `PERF_100NSEC_TIMER` | 4745937500 | |
|
|
| `% Privileged Utility` | `PERF_AVERAGE_BULK` | 35875675583 | 2142017948 |
|
|
| `% Processor Performance` | `PERF_AVERAGE_BULK` | 80044724120 | 669031135 |
|
|
| `% Processor Time` | `PERF_100NSEC_TIMER_INV` | 25109218750 | |
|
|
| `% Processor Utility` | `PERF_AVERAGE_BULK` | 80044725697 | 2142017679 |
|
|
| `% User Time` | `PERF_100NSEC_TIMER` | 4394687500 | |
|
|
| `C1 Transitions/sec` | `PERF_COUNTER_BULK_COUNT` | 2303708 | |
|
|
| `C2 Transitions/sec` | `PERF_COUNTER_BULK_COUNT` | 2337838 | |
|
|
| `C3 Transitions/sec` | `PERF_COUNTER_BULK_COUNT` | 0 | |
|
|
| `Clock Interrupts/sec` | `PERF_COUNTER_COUNTER` | 1571877 | |
|
|
| `DPCs Queued/sec` | `PERF_COUNTER_COUNTER` | 561134 | |
|
|
| `Idle Break Events/sec` | `PERF_COUNTER_BULK_COUNT` | 4641550 | |
|
|
| `Interrupts/sec` | `PERF_COUNTER_COUNTER` | 7795204 | |
|
|
| `Parking Status` | `PERF_COUNTER_RAWCOUNT` | 0 | |
|
|
| `Processor Frequency` | `PERF_COUNTER_RAWCOUNT` | 3700 | |
|
|
|
|
Not read by the collector:
|
|
|
|
- `PERF_AVERAGE_BULK`: `Actual Frequency`
|
|
- `PERF_COUNTER_RAWCOUNT`: `% of Maximum Frequency`, `DPC Rate`, `Performance Limit Flags`, `Processor State Flags`
|
|
- `PERF_PRECISION_100NS_TIMER`: `Average Idle Time`
|