Failure of Installing SDK Components on Jetson TX2

Robotics & Edge Computing Jetson & Embedded Systems Jetson TX2
1

Hi All,
I was able to flash my Jetson TX2 with the Ubuntu OS and when I proceed with Installing SDK Components, there are below errors in the logs and it fails to install NViDiA container runtime and Multimedia API on Target device.

Below are the Errors for these two:

(NVidia container runtime)

18:28:46 INFO : NVIDIA Container Runtime with Docker integration (Beta) : dpkg: dependency problems prevent configuration of nvidia-docker2:
18:28:46 INFO : NVIDIA Container Runtime with Docker integration (Beta) : nvidia-docker2 depends on docker.io (= 18.09.2-0ubuntu1~18.04.1); however:
18:28:46 INFO : NVIDIA Container Runtime with Docker integration (Beta) : Version of docker.io on system is 18.09.7-0ubuntu1~18.04.3.
18:28:46 INFO : NVIDIA Container Runtime with Docker integration (Beta) :
18:28:46 INFO : NVIDIA Container Runtime with Docker integration (Beta) : dpkg: error processing package nvidia-docker2 (–install):
18:28:46 INFO : NVIDIA Container Runtime with Docker integration (Beta) : dependency problems - leaving unconfigured
18:28:46 INFO : NVIDIA Container Runtime with Docker integration (Beta) : Errors were encountered while processing:
18:28:46 INFO : NVIDIA Container Runtime with Docker integration (Beta) : nvidia-docker2
18:28:46 INFO : NVIDIA Container Runtime with Docker integration (Beta) : Unhandled error when running sudo dpkg -i /opt/nvidia/deb_repos/nvidia-docker2_0.9.0_beta.1+docker18.09.2-1_arm64.deb
18:28:46 INFO : NVIDIA Container Runtime with Docker integration (Beta) : exit status 1
18:28:46 INFO : NVIDIA Container Runtime with Docker integration (Beta) : [ Package Install Finished with Error ]
18:28:46 ERROR : NVIDIA Container Runtime with Docker integration (Beta) : NV_DOCKER_TARGET_POST_INSTALL_COMP command scp -q -o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no /tmp/tmp_NV_DOCKER_TARGET_POST_INSTALL_COMP.sh nvidia@192.168.55.1:~; ssh -t -q -o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no nvidia@192.168.55.1 “~/tmp_NV_DOCKER_TARGET_POST_INSTALL_COMP.sh && rm -f ~/tmp_NV_DOCKER_TARGET_POST_INSTALL_COMP.sh” finished with error
18:28:46 INFO : NVIDIA Container Runtime with Docker integration (Beta) :
18:28:46 INFO : NVIDIA Container Runtime with Docker integration (Beta) : [ 92.00 KB used. Disk Avail: 35.04 GB ]
18:28:46 INFO : NVIDIA Container Runtime with Docker integration (Beta) : [ NV_DOCKER_TARGET_POST_INSTALL_COMP Install took 45s ]
18:28:46 INFO : NVIDIA Container Runtime with Docker integration (Beta) : command scp -q -o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no /tmp/tmp_NV_DOCKER_TARGET_POST_INSTALL_COMP.sh nvidia@192.168.55.1:~; ssh -t -q -o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no nvidia@192.168.55.1 “~/tmp_NV_DOCKER_TARGET_POST_INSTALL_COMP.sh && rm -f ~/tmp_NV_DOCKER_TARGET_POST_INSTALL_COMP.sh” finished with error
18:28:46 ERROR : NVIDIA Container Runtime with Docker integration (Beta) : command terminated with error

Multimedia API:

:30:17 INFO : Multimedia API : Linking: camera_jpeg_capture
18:30:17 INFO : Multimedia API : main.o: In function ArgusSamples::ConsumerThread::threadInitialize()': 18:30:17 INFO : Multimedia API : main.cpp:(.text+0xb4): undefined reference to EGLStream::FrameConsumer::create(Argus::OutputStream*, Argus::Status*)’
18:30:17 INFO : Multimedia API : collect2: error: ld returned 1 exit status
18:30:17 INFO : Multimedia API : Makefile:60: recipe for target ‘camera_jpeg_capture’ failed
18:30:17 INFO : Multimedia API : make[1]: *** [camera_jpeg_capture] Error 1
18:30:17 INFO : Multimedia API : make[1]: Leaving directory ‘/usr/src/tegra_multimedia_api/samples/09_camera_jpeg_capture’
18:30:17 INFO : Multimedia API : Makefile:53: recipe for target ‘all’ failed
18:30:17 INFO : Multimedia API : make: *** [all] Error 1
18:30:17 INFO : Multimedia API : exit status 2
18:30:17 INFO : Multimedia API : [ Package Install Finished with Error ]
18:30:17 ERROR : Multimedia API : NV_MULTIMEDIA_API_TARGET_POST_INSTALL_TX2_COMP command scp -q -o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no /tmp/tmp_NV_MULTIMEDIA_API_TARGET_POST_INSTALL_TX2_COMP.sh nvidia@192.168.55.1:~; ssh -t -q -o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no nvidia@192.168.55.1 “~/tmp_NV_MULTIMEDIA_API_TARGET_POST_INSTALL_TX2_COMP.sh && rm -f ~/tmp_NV_MULTIMEDIA_API_TARGET_POST_INSTALL_TX2_COMP.sh” finished with error
18:30:17 INFO : Multimedia API :
18:30:17 INFO : Multimedia API : [ 2.06 GB released. Disk Avail: 37.10 GB ]
18:30:17 INFO : Multimedia API : [ NV_MULTIMEDIA_API_TARGET_POST_INSTALL_TX2_COMP Install took 1m31s ]
18:30:17 INFO : Multimedia API : command scp -q -o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no /tmp/tmp_NV_MULTIMEDIA_API_TARGET_POST_INSTALL_TX2_COMP.sh nvidia@192.168.55.1:~; ssh -t -q -o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no nvidia@192.168.55.1 “~/tmp_NV_MULTIMEDIA_API_TARGET_POST_INSTALL_TX2_COMP.sh && rm -f ~/tmp_NV_MULTIMEDIA_API_TARGET_POST_INSTALL_TX2_COMP.sh” finished with error
18:30:17 ERROR : Multimedia API : command terminated with error

Please recommend what should be done in this case.

2

Hi,
Please try to flash system image and install packages through SDKManager.
https://docs.nvidia.com/sdk-manager/download-run-sdkm/index.html

3

Hi,
that’s exactly what I did. All the installation worked on Target and Host with the two exceptions stated above on the Target machine. I have shared the logs too. Please let me know a good solution to get this working.

4

Hi jwasan,

Please try using ‘eth0’ IP address to install SDK components.
(The default is using usb mode IP address)
Connect the ethernet cable to your TX2 and using ‘ifconfig’ to check eth0 ip address.

5

okay, I have the eth0 IP Addr, now how do I install the failed SDK Components?

6

I’m hitting these two errors as well when trying to install the SDK components (to get the updated 32.2 L4T).

7

thats sad, let me know if you find a solution please.

8

You could edit “~/.nvsdkm/dist/sdkml3_jetpack_l4t_42.json”. Everywhere you see the address “192.168.55.1” (this is the USB virtual wired ethernet) you could put your known IP address as a replacement. Be sure to save an unedited copy before starting.

9

Hi jwasan,

Please run the sdkmanager, the IP address default is show 192.168.55.1, replace to your eth0 ip address.
Then enter the username and password to start install SDK components.

10

On my side, I was already installing over ethernet. The first error seems related to this thread:

https://devtalk.nvidia.com/default/topic/1057847/jetson-tx2/nvidia-docker2-broken-dependency-on-older-version-of-docker-io/

I’d rather not have to reflash though, especially since I only want the updated MMAPI and don’t care about docker. Is it possible to modify the install script to not try to build the MMAPI sample apps during installation (to avoid the second error)? Or to update the MMAPI manually without the sdkmanager?

11

For me, I do not want to reflash my device and go through all the other installations i did on my system but just the two that failed. is there a way to do that?

12

Hi
Building tegra_multimedia_api samples requires packages such as cuDNN, CUDA, TensorRT. Suggest you install all through sdkmanager.

13

As far as I know, multimedia API upgrades with BSP. In other words, it may cause incompatibility issue if you install it on a Jetson device with old BSP. To solve the problem, please flash your device first, then install multimedia API component.

14

Right, I see the issue. The libraries that the MMAPI code depends on (libnvargus, libnvbuf_utils, etc…) are actually in the OS package Jetson_Linux_R32.2.0_aarch64. So it isn’t possible to install the newer MMAPI from 4.2.1 without also flashing the OS from 4.2.1. Will stick to 32.1 for now.

15

my problem was due to broken package.
this fix my problem.

go to target machine
sudo apt-get install -f