I’m in the process of upgrading my custom host os to a newer version using hostapp-update, but it fails with error:
Error processing tar file(exit status 1): write /lib/modules/4.19.71/kernel/drivers/gpu/drm/vc4/vc4.ko: no space left on device
It seems to fail when trying to balena load
the image and I can reproduce the issue on the command line:
root@hltest-mc-2:~# DOCKER_HOST="unix:///var/run/balena-host.sock" balena load -i /resin-data/data/helioslite-raspberrypi-2.46.1+rev5~hl1-dockerimg.tar
f1d9e8a0f1fd: Loading layer [===================> ] 85.23MB/223MB
Error processing tar file(exit status 1): write /lib/modules/4.19.71/kernel/drivers/gpu/drm/vc4/vc4.ko: no space left on device
This is weird because I have plenty of space on the /mnt/sysroot/inactive
partition:
root@hltest-mc-2:~# df -h /mnt/sysroot/inactive
Filesystem Size Used Avail Use% Mounted on
/dev/mmcblk0p3 300M 2.2M 278M 1% /mnt/sysroot/inactive
The image I’m trying to update is not that big so I think this should fit in the /mnt/sysroot/inactive partition:
root@hltest-mc-2:~# du -h /resin-data/data/helioslite-raspberrypi-2.46.1+rev5~hl1-dockerimg.tar
210M /resin-data/data/helioslite-raspberrypi-2.46.1+rev5~hl1-dockerimg.tar
I’ve noticed this similar issue on the forum but in my case the docker storage has already been cleaned up:
root@hltest-mc-2:~# DOCKER_HOST="unix:///var/run/balena-host.sock" balena system df
TYPE TOTAL ACTIVE SIZE RECLAIMABLE
Images 0 0 0B 0B
Containers 0 0 0B 0B
Local Volumes 0 0 0B 0B
My custom host os image is very similar to the official one (just a few services have been added, nothing taking up lot of space). One point to notice though, is that I’m using overlay2 docker storage instead of the default aufs one. Could it explain this upgrade issue?
I’ve tried to watch how the balena load
is progressing on /mnt/sysroot/inactive
partition with command watch -n 5 df -h /mnt/sysroot/inactive
and I noticed that the used space of that partition seemed to increase quicker than the balena load was progressing: even before balena load
progress bar was displayed, the “used space” had already increased from 2.2M to 216M:
Every 5s: df -h /mnt/sysroot/inactive 2020-01-29 15:02:54
Filesystem Size Used Avail Use% Mounted on
/dev/mmcblk0p3 300M 216M 64M 78% /mnt/sysroot/inactive
Then the balena load progress bar is displayed and /mnt/sysroot/inactive used space continues to increase until 100%…
So it’s like if the docker host os image was loaded twice…
Any idea?