jba

Finally found the better answer to my initial question. The process is described in the wiki: https://docs.armbian.com/User-Guide_Recovery/#flashing-boot-loader However, there is one mistake. Instead of the last line ~ $ bash pack/usr/lib/u-boot/platform_install.sh pack/usr/lib/linux-u-boot-nanopineo2-current /dev/XXX # replace XXX with the actual device you need to run ~ $ source pack/usr/lib/u-boot/platform_install.sh ~ $ write_uboot_platform pack/usr/lib/linux-u-boot-nanopineo2-current /dev/XXX # replace XXX with the actual device /dev/sdb In addition it is important to start the partition at block 8192 (at least for my odroid HC2, may be a differnet number on other devices). This way you can easily write your backups to an sd. Regards, Jürgen

I am using an odroid HC2 as home NAS for some years now. Software is armbian/XU4 in debian flavour. At the moment I am running bookworm (which was updated from buster to bullseye and then to bookworm - I know its not supported but it worked fine. And the problem described here started only some month after the last update). Some days ago I wanted to reboot the unit but it was stuck in the boot process. Not available by ssh or any other services. I then found out that the boot stopped because my hard disk could not be mounted from ftsab (while the sd card as root file system could be mounted). After commenting out the line it boots normally. And when logged in by ssh I can mount the hard disk without any problem. As a workaround I put the "noauto" option in fstab and mount the drive after a sleep of some seconds from rc.local. That works fine but I am wondering what is causing the problem? Has something in armbian changed so that mounting on boot time doesnt work anymore? Maybe some problem with initializing the sata controller? Or is this a sign of degrading hardware? Any hints welcome. Juergen

Hi bedna, I have to admit that I didnt look too deep in your script. This is mainly because my backaup strategy works now and is established for a lot of system, not only the odroid hc2. Therefore I dont want to change it. But maybe our misunderstanding is based on the terms image and files. For me you can backup on image level or on file level. There is no in-between. Of course you can mount an image and access the files, but its still a different thing. And taking an image-based backup of a running system is not a good idea in my opinion. > But restoring it is only possible on exactly the same type of sd, which will typically not be the case if you want to make a clone. You cited this sentence but removed the sentence before which is important. I was refering to a backup of the first 4 MB of a sd (and not your script). When you restore the first 4 MB to a different sd, you will delete the partition table. So at least you have to rewrite the partition table. But even then it didnt work for me. Did not find out the reason for that, maybe I did something else wrong. Regards, Jürgen

Hi Bedna, this is not exactly what I was looking for. I regularly run backups via rsync - so it is on a file basis, not image. This allows me easily access old states for every file - just like timemachine. Jürgen

Took me a while to find this out. Finally I found this threat that describes the procedure: Seems you have to write a vanilla image to your sd in order to get it bootable. After that you can delete all files and put your own backup on it. Important point is to then change the current uuid in boot/boot.ini (although it says you shouldnt). That worked for me. In many posts it is proposed to take a backup of the first 4 MB from your sd by dd. But restoring it is only possible on exactly the same type of sd, which will typically not be the case if you want to make a clone. I think this is because the first 4 MB do not only cover the u-boot, but also the partition table. So the partition/file system may be corrupted after writing the first 4 MB on a new sd. You have to recreate the partition table afterwards. But the system will not boot. Even when putting the right UUID in boot/boot.ini. Maybe fdisk is somehow corrupting the u-boot when writing the partition table. So I see no point in backing up the first block of the sd. It will help you only if you want to use exactly the same sd again. Regards, Jürgen

I finally did the update. I used the same procedure as I wrote in the thread I linked in the first post. It worked without problems. However, in the last step nothing was updated. I think this means that the armbian repos for bookworm point to the same versions as for bullseye (kernel, dts, ...). I guess this will change soon. Only then I will get a new kernel. I hope this will not break the systems. Lets see. Jürgen

I take regular backups from my linux boxes (rsyncing the file system). To resume, I boot from usb, copy the files to the disk of the box and then make it bootable by chrooting into the box and executing grub-install or update-grub. That works fine. But how can I do this on my odroid hc2 running armbian/bullseye? I cannot boot from usb. So I need to make the sd card bootable on my debian-amd64 pc after copying the files on it. Is there some command like "update-uboot" I can execute on amd64? Or do I need a virtual host running arm architecture to do so? Jüregen

As bookworm is stable now I want to update my odroid HC2 from bullseye to bookworm? Any experiences welcome. I know this is not supported. But in principle it should work. I did the update from buster to bullseye and that worked perfectly (see link to threat below). I would just do the same for the new update. But maybe there is one special fact to remember that wasn't necessary for the last update. Jürgen

I had the same problem on first boot (buster on HC2). Second boot worked. Didnt find out the reason for failure on first boot (suspectzed an error in ssh config, but that wasnt the case). Did you try a second boot with bullseye? jba