amd_smi_exporter results disagree with rocm-smi

[yx-lamini]

amd_smi_exporter cannot detect any any GPUs [1], rocm-smi shows 4 gpus [2], and amd-smi got messed up, showing internal errors on the screen (just a symptom of potential internal corruption of the system)

[1] amd_smi_exporter shows no GPUs

[2] rocm-smi shows 4 GPUs

[3] amd-smi was messed up during the installation process

[muralimk-amd]

Hi lamini, We haven't observed any issues.
Could you please checkout the latest https://github.com/ROCm/amdsmi and install and then try with amd_smi_exporter.
will provide steps if needed.

[muthusAMD]

Hi @yx-lamini,
As @muralimk-amd recommends, please move to the latest https://github.com/ROCm/amdsmi and the steps are given below.

We also optimized our amd_smi_exporter repos and dependencies, and we moved amd_smi_exporter to fetch data(cpu, gpu data) from amdsmi library recently, so that we can seamlessly query both CPU/GPU data across processors (please refer README.md).

For the ease of testing, please create a testing folder in your home directory
(For Example: We created "mytest" folder in ours home directory "/home/user"), then go to "mytest" folder,
#cd /home/user/mytest
then follow the below steps,

1. Ensure we have Driver in place:
a) ensure amd_hsmp drivers are available. installed : https://github.com/amd/amd_hsmp
#modprobe amd_hsmp
#lsmod | grep amd_hsmp
b) ensure amdgpu drivers are available.
#modprobe amdgpu
#lsmod | grep amdgpu

2. Cleanup Old repos if required[optional]:
#sudo rm -rf /opt/goamdsmi/ /opt/e-sms/ /opt/rocm

3. Export LD LIBRARY PATH:
#export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/opt/rocm/lib:/opt/rocm/lib64:
#echo $LD_LIBRARY_PATH

4. Installed library:
a) amdsmi library: from https://github.com/ROCm/amdsmi
#git clone https://github.com/ROCm/amdsmi.git
#cd amdsmi/;mkdir build;cd build;cmake ../.;make;sudo make install;cd ../..;
Expected Result:
* Latest amd-smi will get installed in "/opt/rocm/bin",
* amd-smi installation path will be displayed during the previous step, "sudo make install" of amdsmi repo.

5. Installed go lang:
Install from https://golang.org/dl/go1.20.12.linux-amd64.tar.gz
#git clone https://github.com/udhos/update-golang
#cd update-golang
#sudo ./update-golang.sh
#source /etc/profile.d/golang_path.sh
#go version;cd ..;

6. Install Exporter:
From https://github.com/amd/amd_smi_exporter
#git clone https://github.com/amd/amd_smi_exporter.git
#cd amd_smi_exporter/src/; make -j64;sudo make install;cd ../../;
Expected Result: amd_smi_exporter should be created under amd_smi_exporter/src

• Please let us know if you face any issues while building amd_smi_exporter and amd-smi by following the above steps.
• amd_smi_exporter and amd-smi results should match each other.
  thanks.