Closed duskembayev closed 4 years ago
@dotnet/dotnet-diag
/cc @sywhang @josalem @noahfalk
Are there any workarounds? I tried to set "EnableDiagnostics" environment variable, but it doesn't help.
Is there any chance you can supply a sampling profiler trace of the process, say using ETW through PerfView? The shutdown path you're seeing there suggests there was a stack trace taken with no symbols, not that the shutdown is actually happening. Code review suggests that one thread is waiting on the call to Accept waiting for a connection, but I'd be surprised if that used this much CPU.
@hoyosjs I've collected profile data, but can not publish them here. Could you give me some email address to send it?
Thanks @duskembayev
An option to disable EventPipe's thread from spawning has been added in master and 3.0 through dotnet/coreclr#27137 and dotnet/coreclr#27140 respectively setting the environment variable COMPlus_EnableDiagnostics
to 0 prior to launching the process. @duskembayev did you get a chance to test if this solves your problem? If not, please feel free to reopen this issue/send me an email.
Hi all. I have an application with some security limits that executes managed .NET Core library (using custom CLR hosting). Everything works fine, but I found that process utilization permanently about 20%. After investigation, using ProcessExplorer, I found thread with this stack trace:
Looks like coreclr cannot create named pipe (named pipes are limited by restricted token of the process). Could you help me with this problem?
dotnet version: 3.0.100-preview8-013656 os: win-x64 platform: x64