Failure to launch daemon during device validation

Developer Tools Nsight Systems Profiling Linux Targets
1

I am new to Nsight Systems and GPU profiling in general.

I am having trouble to connect to a machine on Google Cloud within Nsight Systems, I can’t get past the device validation phase. I can see through `ps aux’ that things are happening on the target during this phase (so the ssh connection must have been established), but Nsight Systems gets stuck on “launching the tools app on the device”. I have not been able to find what this means, but I am getting the following detailed message:

NVIDIA Nsight Systems
2018.3.1.29-2abcab9 OSX

- <username>@<ip_address>:

[Error] Failed to launch daemon
Failed to launch daemon executable on target device.
Please kill all running instances of quadd_d manually or reboot the target device.

Here are the details of the image I selected from Google Cloud:
Google Deep Learning VM
Version: m17
Based on: Debian GNU/Linux 9.6 (stretch) (GNU/Linux 4.9.0-8-amd64 x86_64\n)

Does anybody know what the problem could be? Thanks!

2

I figured it out. This problem arises when the host and daemon cannot communicate. You need to allow the following traffic:
port 22 ingress
port 45555 ingress

If you are using Google Cloud you need to set these firewall rules on the console, not on your target machine.

3

Makes sense, those are the ports we use for data.

I’ll add a blurb about this to our documentation for use in Cloud systems. Thanks for letting us know!