GPU metrics error

Developer Tools Nsight Systems Profiling Linux Targets
1

When I check GPU metrics following error occurs.
What does “[QuadDCommon::tag_error_code*] = 17” mean…?

AnalysisService: failed to start event source Perf: /build/agent/work/20a3cfcd1c25021d/QuadD/Target/quadd_d/quadd_d/jni/EventSource/GpuMetricsInternal.cpp(800): Throw in function void QuadDDaemon::EventSource::GpuMetrics::Internal::Start(const std::shared_ptr&, uint32_t, const std::function&)
Dynamic exception type: boost::wrapexcept
std::exception::what: InternalErrorException
[QuadDCommon::tag_error_code*] = 17
[QuadDCommon::tag_error_text*] = Insufficient privilege
[QuadDCommon::tag_api_function*] = NVPW_Device_PeriodicSampler_BeginSession(&beginSessionParams)
[QuadDCommon::tag_debug*] = 1

Host
Ubuntu 18.04.5 LTS
Nsight Systems 2021.2.1

Device
Ubuntu 18.04.5 LTS
Tesla T4
CUDA 10.2

2

If I set SSH ID as root, profiling is stopped by itself with following message.

Number of incomplete CUPTI events dropped: 465. Incomplete events appear when GPU timestamp information have not been retrieved at the time the profiling session was stopped.

3

I’m suffering from the same error message. Have you heard anything back from anyone?

4

not yet… :(

5

same error. unsolved yet…

6

I’m having the same issue.

7

I am sorry that there was no response to this earlier, your forum post was dropped in an orphaned category that the Nsys team was unaware of until this afternoon.

@andrey_trachenko can you answer this?

8

Hello, GPU metrics sampling needs administrative/root privileges of special setup for non-privileged users. For details please see https://developer.nvidia.com/ERR_NVGPUCTRPERM.

1 Like
9

same error using 2022.1.1 ver in orin-x… any updates?