PR#556: bump the root filesystem size - fedora-kickstarts

fedora-kickstarts

#556 bump the root filesystem size

Merged 4 years ago by mohanboddu. Opened 4 years ago by chrismurphy.

Unknown source devel into master

bump the root filesystem size

Chris Murphy • 4 years ago

d9ee921

fedora-live-workstation.ks

file modified

+1 -1

		`@@ -10,7 +10,7 @@`
		`#`
		`#include snippets/packagekit-cached-metadata.ks`

		`- part / --size 6656`
		`+ part / --size 7680`

		`%post`

chrismurphy commented 4 years ago

a. pre-empt future compose failure due to running out of space
b. existing l-i-t-d persistent root feature depends on the free space on this file system, right now free space is just under 650M and is a bit tight
c. fairly certain free space dropping below 90% is why I'm running into https://bugzilla.redhat.com/show_bug.cgi?id=1715699

pbrobinson commented 4 years ago

It would be nice to see why it's so large and do some analysis to see if it could be reduced back

chrismurphy commented 4 years ago

It seems to be growing at a fairly linear rate, release to release. The slightly larger bump F28 to F29 was font related and itself about 200MB, iirc.

Free space:
Fedora 31, 656M
Fedora 30, 859M
Fedora 29, 1.1G
Fedora 28, 1.5G

There are a couple changes coming that will shrink it, but they're not due until Fedora 32 at the earliest.

chrismurphy commented 4 years ago

Testing suggests bumping this 1GiB will not increase the embedded rootfs.img or ISO size once run through mksquashfs. There's a bit more metadata from ext4 formatting the larger volume, but it's wiped away entirely by squashfs.
# ls -ls
69052 -rw-r--r--. 1 root root 6979321856 Jul 31 18:42 e1.img
32 -rw-r--r--. 1 chris chris 32768 Jul 31 18:49 e1.squashed
69564 -rw-r--r--. 1 root root 8053063680 Jul 31 18:42 e2.img
28 -rw-r--r--. 1 chris chris 28672 Jul 31 18:50 e2.squashed

I think the bump is worth doing, even though the bug on line c) mentioned at the top is now fixed. The current device-mapper overlay, whether RAM or disk based, causes the LiveOS rootfs size to be set at mkfs.ext4 time. That will eventually be decoupled when overlayfs is used instead as part of https://bugzilla.redhat.com/show_bug.cgi?id=1717728 but that's not happening in this cycle.

kevin commented 4 years ago

Can you rebase this?

chrismurphy commented 4 years ago

I don't understand the request. I'm not sure how to rebase in pagure, only make a new commit and pull request, and when I make a new commit on master, the diff stats are identical to this PR.

mohanboddu commented 4 years ago

I don't understand the request. I'm not sure how to rebase in pagure, only make a new commit and pull request, and when I make a new commit on master, the diff stats are identical to this PR.

@chrismurphy On your local devel branch, run a rebase git pull <main_repo> master --rebase and then push it to your fork's devel branch with force as the history gets change git push <your_fork> devel -f