I have a base disk image with a 30GB rootfs. I run mender-convert with MENDER_STORAGE_TOTAL_SIZE_MB=“131072”, expecting the A/B partitions to be about 64GB each. I also pass mender-convert an overlay with ~18GB of data.
I realized that mender-convert-modify writes the 18GB of data into the rootfs of the original image, which only has 30GB of total space, which is why it’s running out of space. Do you have any advice for making mender-convert support the case where the input image’s rootfs is much smaller than the output rootfs? Some ideas off the top of my head:
expand the original rootfs size to the calculated final rootfs size
Inject the overlays into rootfsA in mender-convert-package, and then dd rootfsA into rootfsB
@thienandangthanh I don’t know if this has been solved for mender-convert, but I ended up writing my own build tools that don’t have this problem. So I’m no longer using mender-convert.
I finally figured a way to make it build successfully.
I create a custom config file inherit from configs/beaglebone_black_base_config with 2 variable: