I just installed the latest (mercury) version of endeavouros, replacing a windows install (just linux, NOT dual-boot).
I allocated 1022MiB to the /efi partition. However, after the first install I tried creating a new /home partition and discovered I could no longer login to the new install. Being a newbie I reinstalled the OS. I then figured out how to move the existing /home to the new /home partition without breaking the system.
With the two installs, my /efi partition is down to 172MiB free (84% used). When I tried doing a general package upgrade today, the update appears to be trying to rebuild the boot images causing the package upgrade to fail.
I don’t need the two boot images – is there an easy way to remove the original boot image to free up space in /efi?
Should 1022MiB be large enough for a single boot image, or should the /efi partition need to be expanded?
( 8/13) Reloading system bus configuration...
( 9/13) eos-nvidia-fix.hook
Installing kernel 6.17.6-arch1-1
dracut[I]: Executing: /usr/bin/dracut --no-hostonly --force /efi/c05154185b1e4909aa6bf3796834d8f0/6.17.6-arch1-1/i
nitrd-fallback 6.17.6-arch1-1
<snip>
dracut[I]: *** Creating image file '/efi/c05154185b1e4909aa6bf3796834d8f0/6.17.6-arch1-1/initrd-fallback.tmp' ***
dracut[I]: *** Hardlinking files ***
dracut[I]: *** Hardlinking files done ***
cp: error writing '/efi/c05154185b1e4909aa6bf3796834d8f0/6.17.6-arch1-1/initrd-fallback.tmp': No space left on dev
ice
dracut[F]: Creation of /efi/c05154185b1e4909aa6bf3796834d8f0/6.17.6-arch1-1/initrd-fallback.tmp failed
/usr/lib/kernel/install.d/50-dracut-fallback.install failed with exit status 1.
error: command failed to execute correctly
If you aren’t reformatting the ESP, you probably have multiple installs worth of initrds in there.
Take in a look in /efi. You will see some long strings in there that look like c05154185b1e4909aa6bf3796834d8f0. Type cat /etc/machine-id. That will return an ID string. That is the one you want to keep. Any other long ID strings are from past installs.
So I can just delete the directory in /efi that does not correspond to /etc/machine-id (the unused instance)? It looks like there are also a couple of files in /efi/loader/entries that correspond to this unused instance. I just want to make sure there isn’t a “correct” way to remove it besides just deleting the directory and its contents.
Also, I noticed that the unused instance is identified as 6.17.5-arch-1.1, while the used one is 6.17.6-arch-1.1. Does that mean that the 6.17.6-arch-1.1 (the running instance) has been successfully updated, while the 6.17.5-arch-1.1 (the unused instance) was not updated? In other words, if I deleted the 6.17.5-arch-1.1 instance then I shouldn’t need to worry about the post-package-update failure in dracut (since the running instance has been updated to version 6.17.5)?
I deleted the unused install, leaving me with 42% used in my 1022M /efi partition. Hopefully that’s OK for now. If there’s no immediate danger, I’d like to see how my root and home partitions grow before resizing partitions. I have 64 GiB allocated to root and that’s only 24% used so far but I’d like to get a feel for how fast that will grow after regular system updates.
I was able to do a pacman system update and reboot and all looks well so far. Thanks for your help!
no I got sysd bootloader…and have only the current kernel and the fallback kernel afaik. I agree I almost feel I should have a bigger /efi. But all works.