Commit Graph

339 Commits

Author SHA1 Message Date
Harald Sitter
61fb76f19d fully balance before shrink
when looking at a heatmap of the partition it's still not quite ideal,
when I manually shrink it I get a better block layout, so let's try to
balance the blocks before shrinking. if nothing else it should mean less
relocation during shrinking
2024-11-07 07:21:23 +01:00
Harald Sitter
75bc73d176 create torrent from raw basename
the file on the servers we want is the raw not the output (that's the
name without the .raw suffix)
2024-11-06 16:27:02 +01:00
Harald Sitter
ad734333a6 use smarter logic for shrinking the btrfs
size is only approximate and depending on how big a number it is the
512M buffer may not be enough
2024-11-06 15:48:17 +01:00
Harald Sitter
40b3f76a17 do not balance
it seems entirely superfluous because resizing needs to squeeze the data
anyway (there seems to be no size diff in the outcome if we balance
first or not). meanwhile doing the balance (at least the convert) does
tend to break things resulting in errors inside the kernel
2024-11-06 15:48:17 +01:00
Harald Sitter
e41f5043e9 debug++ 2024-11-06 15:48:17 +01:00
Harald Sitter
b5a1d6e6e4 make sure the first resize works 2024-11-06 15:48:17 +01:00
Harald Sitter
6bec498f17 sync more aggressively
to make sure size data is applied immediately
2024-11-06 15:48:17 +01:00
Harald Sitter
1696b4ff74 write the btrfs.json to the script dir
no sense handing it from one script to the other
2024-11-06 15:48:17 +01:00
Harald Sitter
f08e33623d reduce the esp size
shouldn't really be 512m, in fact even 256 is rather a lot
2024-11-06 15:48:17 +01:00
Harald Sitter
b90e767757 auto determine size when rebuilding image
since we know the size of the input artifacts the automatic size should
be very close to ideal
2024-11-06 15:48:17 +01:00
Harald Sitter
57101a2f14 make sure to sync subvol changes before shrinking
otherwise btrfs might fall over because of the pending data. presents
primarily in it turning read-only
2024-11-06 15:48:17 +01:00
Harald Sitter
bc9fb1fc99 Revert "disable btrfs shrinking for now"
This reverts commit f925969063.
2024-11-06 15:48:17 +01:00
Harald Sitter
c8a7eafdc2 make outputs +r
torrent files are also 600 for some reason. be more liberal in chmodding
in general
2024-11-06 13:40:30 +01:00
Harald Sitter
b0000a5333 torrent: construct web urls completely
the base urls aren't particularly useful on their own, we need to append
our path
2024-11-06 03:27:32 +01:00
Justin Zobel
bc8d4ed64d Sort flatpak list for easier viewing 2024-11-06 12:48:35 +10:30
Harald Sitter
a63ec33d5d throw away useless packages from build env
s3cmd is no longer used because we don't use s3 storage.

vim was never used
2024-11-06 02:01:04 +01:00
Harald Sitter
ce692b289f create torrent file for raw image 2024-11-06 02:01:01 +01:00
Harald Sitter
f925969063 disable btrfs shrinking for now
doesn't quite work as expected. needs more development
2024-11-05 23:23:31 +01:00
Harald Sitter
54226b1ad4 set +x on btrfs-shrink 2024-11-05 22:17:45 +01:00
Harald Sitter
ed8a9f87c1 trun the efi match pattern around
it looks to me like the last argument may be the one that gets created.
possibly explaining why boot counting doesn't work
2024-11-05 21:42:39 +01:00
Harald Sitter
70299f7b5d update hostname 2024-11-05 21:41:16 +01:00
Harald Sitter
3a4572584b call btrfs shrink in correct path
we are inside the dissect, but we need to call the program in our source
dir
2024-11-05 21:41:07 +01:00
Harald Sitter
a005d14005 enable boot counting out of the box 2024-11-05 21:28:24 +01:00
Harald Sitter
ecb14adf55 give btrfs enough space to dedupe and balance 2024-11-05 21:10:45 +01:00
Harald Sitter
5adeefbeec rebuild the image with final sizes
after shrinking the btrfs it will have dangling space at the end. we'll
split the image into its partitions with actual filesystem sizes (i.e.
the new btrfs size) and then re-assmble the image using the partitions
as copyblocks.
this gives us a tightly packed image
2024-11-05 21:10:06 +01:00
Harald Sitter
a133acf128 verbosity-- 2024-11-05 21:00:05 +01:00
Harald Sitter
26f6290687 shrink btrfs to the actually necessary size
this doesn't yet give any gains because partition resizing also needs
figuring out, but it's a start.

specifically we now deduplicate files, balance extents and shrink the
actual filesystem as much as possible.

in a second step we could then chop off the end of the partition table
to actually reduce the image size. except it is more difficult than one
might think. perhaps we should construct a new table entirely instead.
something to figure out
2024-11-05 19:46:58 +01:00
Harald Sitter
b8ff7638b5 enable avahi-daemon.service 2024-11-05 15:54:29 +01:00
Harald Sitter
c816ae1a2e print size for the correct dirs 2024-11-05 14:46:22 +01:00
Harald Sitter
16e7295772 drop unnecessary defrags
they only slow down the build now

we apply compression via the host system
2024-11-05 03:54:49 +01:00
Harald Sitter
ca9e93066c verbosity-- 2024-11-05 03:49:13 +01:00
Harald Sitter
1fdb021f51 print compression diagnostics
something isn't quite right with the compression coverage, technically
we should have force compression, but effectively we have a large chunk
of data that is not compressed
2024-11-05 03:37:38 +01:00
Harald Sitter
867fe8e7df remove lldb
we have gdb through drkonqi
2024-11-05 03:33:22 +01:00
Harald Sitter
e03ea07848 install toolbox as a container solution 2024-11-05 03:33:12 +01:00
Harald Sitter
a0f7202ad5 throw away x11
technically we only need xwayland. we'll see how it goes
2024-11-05 03:32:51 +01:00
Harald Sitter
7270678c63 clean up after install-distro-packages a bit 2024-11-05 02:46:43 +01:00
Harald Sitter
52e3f57631 calamares: bring back localecfg
but remove locale-gen so calamares doesn't try to generate locales
2024-11-05 02:30:18 +01:00
Harald Sitter
e1173612cb more aggressively set a timezone for live
timedatectl doesn't really want to write the file if it thinks the tz is
already UTC
2024-11-05 02:27:51 +01:00
Harald Sitter
b190b8dd16 move calamares script to live directory
not really needed after installation
2024-11-05 02:12:34 +01:00
Harald Sitter
8b3d72c0a1 comment++ 2024-11-05 02:10:25 +01:00
Harald Sitter
6b144bad73 move cleanup to finalize
qt cleanup was already done here
2024-11-05 02:00:22 +01:00
Harald Sitter
c1a8c0592c only unpack en locale for flatpaks
our locale story is a bit undefined right now anyway
2024-11-05 01:59:53 +01:00
Harald Sitter
3b2cb02e82 line wrap the flatpak install
easier on the eyes
2024-11-05 01:59:23 +01:00
Harald Sitter
ed8ff5c416 force mighty compression
not that I think it makes a difference. we already receive the
compressed data
2024-11-05 01:17:37 +01:00
Harald Sitter
836a916d6d check for file rather than directory
the directory gets created here...
2024-11-05 01:17:18 +01:00
Harald Sitter
2236368db9 only load gpg key if it's actually available 2024-11-05 00:36:53 +01:00
Harald Sitter
45d3933a45 fix permissions before build 2024-11-05 00:31:23 +01:00
Harald Sitter
c850a863e9 vacuum: generate sha256sums from server data
that way we don't have extranous entries in there
2024-11-05 00:19:32 +01:00
Harald Sitter
6afc5dde36 vacuum: close connection
not that it matters much since this process doesn't stick around
2024-11-05 00:18:35 +01:00
Harald Sitter
5da433f90b refine comment a tad 2024-11-03 19:30:05 +01:00