Building Real-Time Video AI Applications。NGC CLI fails on DLI Jupyter Platform: GLIBC_2.28 not found

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Hi all,

I’m currently working through the “Building Real-Time Video AI Applications” certification on NVIDIA’s DLI Jupyter platform. During Step 2.2 of the assessment notebook, I encountered a critical issue when executing the ngc registry model list command.

Here’s the exact error message:
Error loading Python lib ‘/dli/task/ngc_assets/ngccli/ngc-cli/libpython3.11.so.1.0’:
dlopen: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.28’ not found
(required by /dli/task/ngc_assets/ngccli/ngc-cli/libpython3.11.so.1.0)

It seems that the default NGC CLI used in the notebook expects a newer version of glibc than what is currently available in the container environment.

❓What I’ve Tried:

  • Re-running Step 2.1 to re-download and extract the CLI.
  • Verifying system GLIBC version (it’s below 2.28).
  • Attempted to run ngc CLI from the shell — same result.

❓My Questions:

  1. Is this a known issue with the current DLI environment?
  2. Is there a way to manually switch to an older version of the NGC CLI (e.g., ngccli_cat_linux) that is compatible with this GLIBC version?
  3. If so, could anyone share the steps or workaround?

Any help would be greatly appreciated. I’m currently blocked and cannot proceed with downloading the DashCamNet model for the next steps in the course.

Thanks in advance!

1 Like
3

Could you attach the link of this Building Real-Time Video AI Applications?

5

I am also facing the same issue.

Building Real-Time Video AI Applications

6

same here

8

Same here. I didn’t have any issue when doing part 1 or part 2 of the interactive content, but it started to happen during the coding assignment. So I went back to my previous assignment and they got the same error when running ngc.

9

@yuweiw are you able to take another look at this?

10

Hi,

Thank you for taking the course. It looks like NGC made an update recent that’s causing this issue.

Manually downloading an older version of the NGC CLI would work. For example, I tested it with v3.60.0 found here. Please manually install a working version until the course is patched.

Best,

Kevin

1 Like
11

y’all should have automated tests on these courses, it’s pretty embarrassing that these courses break randomly and remain broken for months.

13

okay i got it to work.

  1. change the download url to:
    !wget “https://api.ngc.nvidia.com/v2/resources/nvidia/ngc-apps/ngc_cli/versions/3.60.0/files/ngccli_linux.zip” -P $NGC_DIR/ngccli

  2. delete the existing “ngc_assets/ngccli” folder

  3. rerun the notebook

2 Likes
14

i keep getting this error at 4.4

Kernel Restarting
The kernel for assessment.ipynb appears to have died. It will restart automatically.

this is the output before the error happens:

image
image1470×458 47 KB

any idea how to solve it?

15

Yes this worked, Thanks.

16

Yes something like this worked perfectly, if you replace the code cell with the below:

# DO NOT CHANGE THIS CELL
import os
os.environ['NGC_DIR']='/dli/task/ngc_assets'
ngc_cli_path = os.path.join(os.environ['NGC_DIR'], 'ngccli')
ngc_zip_file = os.path.join(ngc_cli_path, 'ngccli_linux.zip') # Define the specific zip filename

# Clean up previous attempts FIRST
!rm -rf $ngc_cli_path

# Download and install specific NGC CLI version
!mkdir -p $ngc_cli_path
!wget "https://api.ngc.nvidia.com/v2/resources/nvidia/ngc-apps/ngc_cli/versions/3.60.0/files/ngccli_linux.zip" -O $ngc_zip_file # Use -O to force filename
!unzip -o -u "$ngc_zip_file" -d $ngc_cli_path/
!rm $ngc_zip_file
os.environ["PATH"]="{}/ngccli/ngc-cli:{}".format(os.getenv("NGC_DIR", ""), os.getenv("PATH", ""))