Upgrade/Update strategy


In my company we plan to deploy multiple Jetsons Nano into the edge (possibly more than 100).
What would be a good strategy for upgrading the full image, and also only our software. We would like something like full firmware upgrade, and incremental updates too.
These boards won’t have access to internet.
We don’t have a lot of experience with these kind of installations.
I’ve read about busybox, yocto, etc, but I need opinions.
We’d like some kind of readonly system, and keep the configuration somewhere (maybe a NVME disk into our future carrier board, or a RW partition on emmc for logs/config, including network config, so that it will be possible to change IP etc via an API).

thanks in advance for sharing experience


Please refer to the document first.

Honestly speaking, we don’t know if this could meet your goal because it seems you are not yet familiar with jetson platforms.
Maybe you can get hands-on this first to know how jetson works, and then come back to forum to have another discussion.

1 Like

Thanks @WayneWWW for the reply.
It’s true we don’t have much experience in L4T. We are developping applications for some time now with TensorRT, and we are starting to think about updrade strategy etc, I’ll read about the bootloader and what we could do.


Thanks for the good content.
I am also thinking about update and upgrade.

It’s a terminology question, what does edge mean here?

Thank you.

Hi @forumuser, I meant “deploy at the edge”, sorry for my English. It means we plan to have many jetsons remotely installed at different physical locations.
For now we have not started research about how to do it. Probably we’ll use deb packages. I think we’ll discard full image upgrade, because of limited space on emmc memory. I’m not an expert but I read about having 2 bootable systems at the same time (if something goes wrong you can fallback to the previous version), for full upgrade, I think it’s used a lot for IoT devices.
I have to investigate more about it, packages updates has its drawbacks too, to keep the system coherent, but it allows low bandwidth updates.

If you have ideas/interesting information to share I’ll be happy to learn about it, if you don’t mind.

1 Like


We are also struggling with this problem you mentioned below.

It is difficult to maintain if our firmware is affected by the update.

Thank you :)

This thread is old, but at the end I discovered GitHub - OE4T/meta-tegra: BSP layer for NVIDIA Jetson platforms, based on L4T and we now generate our own distribution with yocto.