sendnrw/windows_exporter
2
0
Fork 0
mirror of https://github.com/prometheus-community/windows_exporter.git synced 2026-10-17 10:09:05 +02:00
Code Issues Packages Projects Releases Wiki Activity

netframework: add the counter reference to AGENTS.md (#2751)

Browse Source 
Signed-off-by: Jan-Otto Kröpke <mail@jkroepke.de>
This commit is contained in:
Jan-Otto Kröpke
2026-10-16 10:31:12 +02:00
committed by GitHub
parent 43d77f6869
commit 82e8e425df
1 changed files with 204 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
internal/collector/netframework/AGENTS.md
+204
View File
@@ -0,0 +1,204 @@
# `netframework` collector
The collector reads the eight .NET Framework PDH objects `.NET CLR Exceptions`,
`.NET CLR Interop`, `.NET CLR Jit`, `.NET CLR Loading`,
`.NET CLR LocksAndThreads`, `.NET CLR Memory`, `.NET CLR Remoting` and
`.NET CLR Security` with `pdh.CounterTypeRaw`, all in one shared query. The
objects come from the .NET Framework performance counter provider
(`mscoree.dll`), not from .NET (Core). The objects are missing on the German
development machine although the provider is registered, so the reference
below comes from a CI probe.
- Every object has one instance per process that loaded the .NET Framework
CLR, named after the process image without `.exe`, plus `_Global_`. Two
processes with the same image report the same instance name, e.g.
`powershell` twice. The collector skips `_Global_` and publishes later
duplicates with an occurrence suffix (`powershell#1`).
- `% Time in Jit`, `% Time in RT checks` and `% Time in GC` are
`PERF_RAW_FRACTION`. Their value is raw / base (`,secondvalue`), not a
percentage and not a cumulative value. Per the .NET documentation, they
cover the time since the end of the previous JIT, check or GC phase and are
updated when the next one ends. Skip the ratio when the base is 0. On
`_Global_` the base is `0xFFFFFFFF` and the ratio is meaningless. 0.31 read
`% Time in Jit` and `% Time in RT checks` without the base and divided the
raw value by the performance counter frequency, which is wrong; only
`% Time in GC` used the base.
- `% Time Loading` and `% Time Sig. Authenticating` are
`PERF_COUNTER_TIMER`, not fractions. The collector does not read them.
- Every rate counter (`/ sec`, `/sec`, `Rate of ...`) is
`PERF_COUNTER_COUNTER` or `PERF_COUNTER_BULK_COUNT`, whose raw value is a
cumulative count since the process started, except the Interop
`# of TLB exports / sec` and `# of TLB imports / sec`, which are
`PERF_COUNTER_RAWCOUNT`.
- The `.NET CLR Memory` sizes and counts are `PERF_COUNTER_LARGE_RAWCOUNT`
gauges in bytes or counts; `# Gen N Collections` and `# Induced GC` are
cumulative. The object also has a `Process ID` counter.
## Counter reference
Recorded on the `windows-2022` GitHub Actions runner. `Raw` and `Base` are the
first and second values that `pdh.CounterTypeRaw` returns for a `powershell`
instance; `Base` is shown for fraction counters only.
### `.NET CLR Exceptions`
One instance per process that loaded the CLR, plus `_Global_`.
Counters the collector reads:
| Counter | Type | Raw | Base |
|---|---|---|---|
| `# of Exceps Thrown` | `PERF_COUNTER_RAWCOUNT` | 6 | |
| `# of Filters / sec` | `PERF_COUNTER_COUNTER` | 0 | |
| `# of Finallys / sec` | `PERF_COUNTER_COUNTER` | 14 | |
| `Throw To Catch Depth / sec` | `PERF_COUNTER_COUNTER` | 30 | |
Not read by the collector:
- `PERF_COUNTER_COUNTER`: `# of Exceps Thrown / sec`
### `.NET CLR Interop`
One instance per process that loaded the CLR, plus `_Global_`.
Counters the collector reads:
| Counter | Type | Raw | Base |
|---|---|---|---|
| `# of CCWs` | `PERF_COUNTER_RAWCOUNT` | 5 | |
| `# of marshalling` | `PERF_COUNTER_RAWCOUNT` | 0 | |
| `# of Stubs` | `PERF_COUNTER_RAWCOUNT` | 226 | |
Not read by the collector:
- `PERF_COUNTER_RAWCOUNT`: `# of TLB exports / sec`, `# of TLB imports / sec`
### `.NET CLR Jit`
One instance per process that loaded the CLR, plus `_Global_`.
Counters the collector reads:
| Counter | Type | Raw | Base |
|---|---|---|---|
| `# of Methods Jitted` | `PERF_COUNTER_RAWCOUNT` | 49 | |
| `% Time in Jit` | `PERF_RAW_FRACTION` | 0 | 10000 |
| `Standard Jit Failures` | `PERF_COUNTER_RAWCOUNT` | 0 | |
| `Total # of IL Bytes Jitted` | `PERF_COUNTER_RAWCOUNT` | 6261 | |
Not read by the collector:
- `PERF_COUNTER_RAWCOUNT`: `# of IL Bytes Jitted`
- `PERF_COUNTER_COUNTER`: `IL Bytes Jitted / sec`
### `.NET CLR Loading`
One instance per process that loaded the CLR, plus `_Global_`.
Counters the collector reads:
| Counter | Type | Raw | Base |
|---|---|---|---|
| `Bytes in Loader Heap` | `PERF_COUNTER_LARGE_RAWCOUNT` | 1957888 | |
| `Current appdomains` | `PERF_COUNTER_RAWCOUNT` | 1 | |
| `Current Assemblies` | `PERF_COUNTER_RAWCOUNT` | 17 | |
| `Current Classes Loaded` | `PERF_COUNTER_RAWCOUNT` | 352 | |
| `Total # of Load Failures` | `PERF_COUNTER_RAWCOUNT` | 0 | |
| `Total Appdomains` | `PERF_COUNTER_RAWCOUNT` | 1 | |
| `Total appdomains unloaded` | `PERF_COUNTER_RAWCOUNT` | 0 | |
| `Total Assemblies` | `PERF_COUNTER_RAWCOUNT` | 17 | |
| `Total Classes Loaded` | `PERF_COUNTER_RAWCOUNT` | 352 | |
Not read by the collector:
- `PERF_COUNTER_COUNTER`: `Rate of appdomains`, `Rate of appdomains unloaded`, `Rate of Assemblies`, `Rate of Classes Loaded`, `Rate of Load Failures`
- `PERF_COUNTER_TIMER`: `% Time Loading`
- `PERF_COUNTER_RAWCOUNT`: `Assembly Search Length`
### `.NET CLR LocksAndThreads`
One instance per process that loaded the CLR, plus `_Global_`.
Counters the collector reads:
| Counter | Type | Raw | Base |
|---|---|---|---|
| `# of current logical Threads` | `PERF_COUNTER_RAWCOUNT` | 10 | |
| `# of current physical Threads` | `PERF_COUNTER_RAWCOUNT` | 8 | |
| `# of current recognized threads` | `PERF_COUNTER_RAWCOUNT` | 2 | |
| `# of total recognized threads` | `PERF_COUNTER_RAWCOUNT` | 2 | |
| `Current Queue Length` | `PERF_COUNTER_RAWCOUNT` | 0 | |
| `Queue Length Peak` | `PERF_COUNTER_RAWCOUNT` | 0 | |
| `Total # of Contentions` | `PERF_COUNTER_RAWCOUNT` | 0 | |
Not read by the collector:
- `PERF_COUNTER_COUNTER`: `Contention Rate / sec`, `Queue Length / sec`, `rate of recognized threads / sec`
### `.NET CLR Memory`
One instance per process that loaded the CLR, plus `_Global_`.
Counters the collector reads:
| Counter | Type | Raw | Base |
|---|---|---|---|
| `# GC Handles` | `PERF_COUNTER_LARGE_RAWCOUNT` | 328 | |
| `# Gen 0 Collections` | `PERF_COUNTER_LARGE_RAWCOUNT` | 1 | |
| `# Gen 1 Collections` | `PERF_COUNTER_LARGE_RAWCOUNT` | 0 | |
| `# Gen 2 Collections` | `PERF_COUNTER_LARGE_RAWCOUNT` | 0 | |
| `# Induced GC` | `PERF_COUNTER_LARGE_RAWCOUNT` | 0 | |
| `# of Pinned Objects` | `PERF_COUNTER_LARGE_RAWCOUNT` | 2 | |
| `# of Sink Blocks in use` | `PERF_COUNTER_LARGE_RAWCOUNT` | 20 | |
| `# Total committed Bytes` | `PERF_COUNTER_LARGE_RAWCOUNT` | 17883136 | |
| `# Total reserved Bytes` | `PERF_COUNTER_LARGE_RAWCOUNT` | 402644992 | |
| `% Time in GC` | `PERF_RAW_FRACTION` | 530853 | 3964773039 |
| `Allocated Bytes/sec` | `PERF_COUNTER_BULK_COUNT` | 17877616 | |
| `Finalization Survivors` | `PERF_COUNTER_LARGE_RAWCOUNT` | 126 | |
| `Gen 0 heap size` | `PERF_COUNTER_LARGE_RAWCOUNT` | 134217728 | |
| `Gen 0 Promoted Bytes/Sec` | `PERF_COUNTER_BULK_COUNT` | 10683176 | |
| `Gen 1 heap size` | `PERF_COUNTER_LARGE_RAWCOUNT` | 16777480 | |
| `Gen 1 Promoted Bytes/Sec` | `PERF_COUNTER_BULK_COUNT` | 0 | |
| `Gen 2 heap size` | `PERF_COUNTER_LARGE_RAWCOUNT` | 24 | |
| `Large Object Heap size` | `PERF_COUNTER_LARGE_RAWCOUNT` | 1100176 | |
| `Process ID` | `PERF_COUNTER_LARGE_RAWCOUNT` | 13232 | |
Not read by the collector:
- `PERF_COUNTER_LARGE_RAWCOUNT`: `# Bytes in all Heaps`, `Promoted Finalization-Memory from Gen 0`, `Promoted Memory from Gen 0`, `Promoted Memory from Gen 1`
### `.NET CLR Remoting`
One instance per process that loaded the CLR, plus `_Global_`.
Counters the collector reads:
| Counter | Type | Raw | Base |
|---|---|---|---|
| `Channels` | `PERF_COUNTER_RAWCOUNT` | 0 | |
| `Context Proxies` | `PERF_COUNTER_RAWCOUNT` | 0 | |
| `Context-Bound Classes Loaded` | `PERF_COUNTER_RAWCOUNT` | 0 | |
| `Context-Bound Objects Alloc / sec` | `PERF_COUNTER_COUNTER` | 0 | |
| `Contexts` | `PERF_COUNTER_RAWCOUNT` | 1 | |
| `Total Remote Calls` | `PERF_COUNTER_RAWCOUNT` | 0 | |
Not read by the collector:
- `PERF_COUNTER_COUNTER`: `Remote Calls/sec`
### `.NET CLR Security`
One instance per process that loaded the CLR, plus `_Global_`.
Counters the collector reads:
| Counter | Type | Raw | Base |
|---|---|---|---|
| `# Link Time Checks` | `PERF_COUNTER_RAWCOUNT` | 0 | |
| `% Time in RT checks` | `PERF_RAW_FRACTION` | 128 | 325 |
| `Stack Walk Depth` | `PERF_COUNTER_RAWCOUNT` | 1 | |
| `Total Runtime Checks` | `PERF_COUNTER_RAWCOUNT` | 4288 | |
Not read by the collector:
- `PERF_COUNTER_TIMER`: `% Time Sig. Authenticating`