I'm using both. The installer first creates the card in the USB adapter, then instructs me to power down, move this prepared card into the microSD card slot and then reboot.What is the reason of using an extra component? And not just use the RPi SD-card inserted in the microSD card-slot.
Then, a second card is placed into the USB adapter and the install procedure creates a second bootable system there. These two systems are my "maintenance" systems should boot from the production system fail.
Finally, the installer creates the "production" system that contains rootfs and bootfs (and several additional partitions for /var, /tmp, /home, swap etc) as MD Raid devices. Fortunately the boot loader accepts a MD RAID partition as long as the bootfs resides on a raid 1 array with type 0.9 superblock. The initrd needs to be rebuilt with the necessary modules, of course.
Using the autoboot.txt (in a separate tiny partition) and TRYBOOT feature, I switch between maintenance and production modes.
I'm using this setup with SD cards, external USB SSD drives or NVMe drives in several headless systems. The USB drives have the advantage of being hot-pluggable. Usually I can replace and re-mirror a failed drive/SD card without or with only minimal service interruption.
Regards
Marcel
Statistics: Posted by abuelomg — Sat Nov 23, 2024 12:21 pm