document DOTNET_EventPipeThreadSamplingRate

docs/core/diagnostics/dotnet-trace.md

@@ -237,6 +237,9 @@ dotnet-trace collect
   | `gc-collect` | Tracks GC collections only at very low overhead.         |
   | `database`   | Captures ADO.NET and Entity Framework database commands. |
 
+  > [!TIP]
+  > The sampling rate used by `dotnet-sampled-thread-time` can be changed with the `DOTNET_EventPipeThreadSamplingRate` environment variable (value in milliseconds). This setting is process-global and affects all EventPipe sessions. See [Trace using environment variables](./eventpipe.md#trace-using-environment-variables) for details.
+
   > [!NOTE]
   > In past versions of the dotnet-trace tool, the collect verb supported a profile called `cpu-sampling`. This profile was removed because the name was misleading. It sampled all threads regardless of their CPU usage. You can achieve a similar result now using `--profile dotnet-sampled-thread-time,dotnet-common`. If you need to match the former `cpu-sampling` behavior exactly, use `--profile dotnet-sampled-thread-time --providers "Microsoft-Windows-DotNETRuntime:0x14C14FCCBD:4"`.
 

docs/core/diagnostics/eventpipe.md

@@ -84,6 +84,8 @@ However, you can use the following environment variables to set up an EventPipe
 
 * `DOTNET_EventPipeProcNumbers`: Set this to `1` to enable capturing processor numbers in EventPipe event headers. The default value is `0`.
 
+* `DOTNET_EventPipeThreadSamplingRate`: Available in .NET 11 and later. Sets the interval, in milliseconds, for the EventPipe thread time sampling profiler. When set to `0` or omitted, the runtime uses its built-in default of 10 ms (~100 Hz). This setting is process-global and affects all EventPipe sessions, including on-demand traces started by tools such as [dotnet-trace](./dotnet-trace.md). Setting a large value reduces sampling overhead but also reduces the resolution of any trace collected during the process lifetime.
+
 * `DOTNET_EventPipeConfig`: Sets up the EventPipe session configuration when starting an EventPipe session with `DOTNET_EnableEventPipe`.
   The syntax is as follows:
 

docs/core/runtime-config/debugging-profiling.md

@@ -81,6 +81,8 @@ The following table compares perf maps and jit maps.
 | **runtimeconfig.json** | N/A          | N/A    |
 | **Environment variable** | `DOTNET_PerfMapEnabled` | `0` - disabled<br/>`1` - perf maps and jit dumps both enabled<br/>`2` - jit dumps enabled<br/>`3` - perf maps enabled |
 
+For EventPipe-related environment variables, see [Trace using environment variables](../diagnostics/eventpipe.md#trace-using-environment-variables).
+
 ## Perf log markers
 
 - Enables or disables the specified signal to be accepted and ignored as a marker in the perf logs.