INSTALLATION FAILED(SDK Manager 1.3 & JetPack 4.3)

dandelion1124 · October 2, 2020, 3:57am

The other day, I received the following mail from NVIDIA.

Version 1.3.0 of the NVIDIA SDK Manager client installer is now available and replaces all previous releases. 
Some older releases do not support our latest login and authentication system 
which means they cannot download files that normally require membership.

So, I tried to install JetPack 4.3 using SDK Manager 1.3.0(sdkmanager_1.3.0-7105_amd64.deb).
But, the following message was displayed.

And, NVIDIA does not support previous version of SDK Manager.
So, there is no workaround for this problem.

Please investigate immediately to take an appropriate action.

dandelion1124 · October 2, 2020, 4:02am

I checked log file.

12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): Selecting previously unselected package nvidia-container-csv-cudnn.
12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): (Reading database ... 135681 files and directories currently installed.)
12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): Preparing to unpack .../nvidia-container-csv-cudnn_7.6.3.28-1+cuda10.0_arm64.deb ...
12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): Unpacking nvidia-container-csv-cudnn (7.6.3.28-1+cuda10.0) ...
12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): dpkg: dependency problems prevent configuration of nvidia-container-csv-cudnn:
12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): nvidia-container-csv-cudnn depends on libcudnn7 (= 7.6.3.28-1+cuda10.0); however:
12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): Package libcudnn7 is not installed.
12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): dpkg: error processing package nvidia-container-csv-cudnn (--install):
12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): dependency problems - leaving unconfigured
12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): Errors were encountered while processing:
12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): nvidia-container-csv-cudnn
12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): Unhandled error when running sudo dpkg -i /opt/nvidia/deb_repos/nvidia-container-csv-cudnn_7.6.3.28-1+cuda10.0_arm64.deb
12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): [ Package Install Finished with Error ]
12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): [ NV_DOCKER_TARGET_POST_INSTALL_COMP Install took 19s ]
12:00:31 ERROR: NVIDIA Container Runtime with Docker integration (Beta): [error]: dpkg install on target failed; [exec_command]: safe_apt_install "sudo dpkg -i /opt/nvidia/deb_repos/nvidia-container-csv-cudnn_7.6.3.28-1+cuda10.0_arm64.deb" 5; sleep 5; ; [deb_path]: /home/nvidia/Downloads/nvidia/sdkm_downloads/nvidia-container-csv-cudnn_7.6.3.28-1+cuda10.0_arm64.deb
12:00:31 ERROR: NVIDIA Container Runtime with Docker integration (Beta): command terminated with error
12:00:31 SUMMARY: NVIDIA Container Runtime with Docker integration (Beta): Failed to install debian package [target]

dandelion1124 · October 2, 2020, 4:13am

12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): Unpacking nvidia-container-csv-cudnn (7.6.3.28-1+cuda10.0) ...
12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): dpkg: dependency problems prevent configuration of nvidia-container-csv-cudnn:
12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): nvidia-container-csv-cudnn depends on libcudnn7 (= 7.6.3.28-1+cuda10.0); however:
12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): Package libcudnn7 is not installed.
12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): dpkg: error processing package nvidia-container-csv-cudnn (--install):
12:00:31 INFO: NVIDIA Container Runtime with Docker integration (Beta): dependency problems - leaving unconfigured

After this installation, I checked installed package on Jetson.
I checked that libcudnn7(7.6.3.28-1+cuda10.0) was installed.

carolyuu · October 5, 2020, 2:03am

Hi dandelion1124,

Are you still got the same error when press “RETRY” from sdkmanager?

dandelion1124 · October 5, 2020, 5:47am

Are you still got the same error when press “RETRY” from sdkmanager?

I’ll try again. Please wait a little.
By the way, do you know this problem as known issue?

igalNV · October 5, 2020, 2:30pm

Issue is fixed at server side. Please try again.
The issue origin was not related to the new 1.3 client but older settings issue that require to click “retry” to succeed. It is now fixed to work at first run.

carolyuu · October 6, 2020, 3:24am

Hi dandelion1124,

This issue fixed and work.
Please check again and let us know the result.
Thanks!

dandelion1124 · October 7, 2020, 5:55am

Thank you for your reply. Which version should I use for this check?
Because, SDK Manager is updated to 1.3.1.
And, I heard that this issue is happening on the server side.

carolyuu · October 7, 2020, 5:59am

Hi dandelion1124,

Please use the latest SDK Manager 1.3 version.

dandelion1124 · October 8, 2020, 3:41pm

I could install using SDK Manager 1.3.1.