http://bugzilla.opensuse.org/show_bug.cgi?id=999776
http://bugzilla.opensuse.org/show_bug.cgi?id=999776#c2
--- Comment #2 from Carlos Robinson ---
Welcome. :-)
That question was (partially) answered in the mail thread where this started,
in the factory mail list. One reason is that the package list is bigger in
Leap. Another reason is that Leap's kernel has more modules.
You posted there your ideas in the issue - I'll insert it here for ease of
access:
+++·············
BTW, we worked on reducing the install memory footprint recently:
https://lizards.opensuse.org/2016/07/27/highlights-of-yast-development-sprin...
For this I collected some numbers (see attached file). The things mentioned
there have all been implemented meanwhile.
You can see that tumbleweed (I would place leap close to tw) needs more
space than sles12 but not _that_ much more. And as sles12 _does_ install on
0.5 GB machines, it's not clear why 0.75 GB should be a problem for leap.
Even if you account for larger repositories.
It is true that there are hooks still in linuxrc for quite some
counter-measures including activating swap and even creating a swap file on
a windows partition (haven't tried this for a decade but the code is still
there...).
But the memory limits for those haven't been updated for ages. Mainly
because it is a bit unpractical: you have to tune them again for every
release and every architecture. And yes, it changes from beta to beta.
So, basically we've given up on the fine-tuning and just say 'go for 1GB'.
That said, I think that leap failing to install on 0.75 GB would be a bug.
·············++-
The mentioned file is this:
+++·············
Reaching the 512 MiB installation limit
=======================================
A way to save some space would be to remove all files in inst-sys that are
already part of the initrd.
ATM both initrd and inst-sys are self-contained; that is, all lib
dependencies are resolved. This leads to some duplicates: e.g. glibc is both
in initrd and inst-sys.
We can remove the duplicate files from inst-sys and save quite some space.
But this will mean that only the inst-sys that was created together with the
initrd can be used. linuxrc can enforce this (the check exists but is
disabled by default).
I've done some test builds and here are the numbers:
- total image size (initrd + instsys) in MiB
before after saved
factory i586 224.5 194.1 30.4
factory x64_64 215.6 185.6 30.0
sle12-sp2 aarch64 162.6 147.0 15.6
sle12-sp2 ppc64le 162.6 145.2 17.4
sle12-sp2 s390x 147.0 129.2 17.8
sle12-sp2 x86_64 176.1 158.4 17.7
sle12-ga aarch64 119.8
sle12-ga ppc64le 138.8
sle12-ga s390x 122.1
sle12-ga x86_64 151.0
As you can see, quite some part of the 512 MiB are already taken by file
system images - and it's growing.
If things get worse, a major blob to get rid of (except on s390x) would be
the kernel modules (after udev has loaded everything that's needed). With
the added pitfall that things like raid or lvm that are loaded later must be
exempt from this.
- kernel modules (compressed) in MiB
factory i586 49.1
factory x86_64 50.0
sle12-sp2 aarch64 26.9
sle12-sp2 ppc64le 16.8
sle12-sp2 s390x 5.9
sle12-sp2 x86_64 28.8
·············++-
--
You are receiving this mail because:
You are on the CC list for the bug.