Generic cloud images with the linux-generic kernel are not able to
boot without an initramfs. Previously, these images attempted to boot
without an initramfs, would fail, and then retry with an initramfs.
This slows the boot and is confusing behavior.
It was reported and confirmed in LP bug #1875400
(https://bugs.launchpad.net/cloud-images/+bug/1875400) that on the public
KVM cloud image there exists a large list of packages marked for auto-removal.
This should never be the case on a released cloud image.
These packages are marked for auto-removal because in the KVM image binary hook
we removed both initramfs-tools and busybox-initramfs packages. Due to package
dependencies this also removed:
busybox-initramfs* cloud-initramfs-copymods* cloud-initramfs-dyn-netconf*
cryptsetup-initramfs* initramfs-tools* initramfs-tools-core* multipath-tools*
overlayroot* sg3-utils-udev* ubuntu-server*
But it did not remove all the packages that the above list depended on.
This resulted in all those packages being marked for auto-removal because they
were not manually installed nor did they have any manually installed packages
that depended on them.
The removal of initramfs-tools and busybox-initramfs was to avoid the
generation of initramfs in images that should boot initramfsless.
This requirement is obsolete now because the initramfsless boot handling
is now handled via setting GRUB_FORCE_PARTUUID in /etc/default/grub.d/40-force-partuuid.cfg.
In test images I have verified that GRUB_FORCE_PARTUUID is set and that
boot speeds have not regressed.
LP: #1875400
Vagrant images were previously put at 10G, but this was a regression
from Trusty, in which they were 40G. This made it a tough sell for
users to upgrade if they were using a Ubuntu desktop experience.
This change does not impact disk usage as Vagrant with the virtualbox
provider dynamically allocates space with the VMDK. On a test system,
the VMDK took up 1.1G of disk space according to df, and after
creating a 2G file in Vagrant, the VMDK grew to 3.1G.
Therefore, users who are running on a system with little free space will
not see adverse effects if they upgrade to a new vagrant image
Seeing any snap via snap_preseed will evaluate the base for each snap
and seed the appropriate base. There should be no reason to explicitly
seed the 'core' snap and with snaps moving to 'core18' this will add
'core' without need.
The _snap_post_process function is meant to install snapd if core18 is the
only core snap installed or removed snapd if core is installed and snapd
was not explicitly installed. But the current logic in _snap_preseed
will never call _snap_post_process. $core_name will never be empty
with the existing logic, but even if it were that would only be for the
'core' snap and we'd miss using the 'core18' logic that pulls in snapd.
Given the case statement in _snap_post_process can handle doing the
right thing given any snap we can just call it unconditionally.
In the buildd image chroot, /etc/resolv.conf is a symbolic link to
a configuration file in the /run directory. A call to truncate will
modify that file, which we should not do. Instead, we want to remove
the symbolic link and replace it with an empty file.
Back in 2017 some code was added to ignore failures tearing down loop
devices. But debugging that growpart race on cloud images made me (very)
aware of a potential cause of the race: doing something like zerofree on
a device will cause udev scripts to run, and if they are still running
by the time kpartx is called, you would expect the kpartx -d to fail. So
lets see if a udevadm settle helps, and get rid of one of the "sometimes
this fails but we don't know why" comments...
With the removal of snap-tool failures are seen in image builds that do
not have the 'core' snap included by the seed. This is the case for the
minimized subproject of the ubuntu-cpc project where lxd/core is removed.
In that subproject, any binary hook which adds a snap that is based
on 'core' will not add 'core' and fail 'snap debug validate-seed'.
snap-tool included the following logic in the 'snap-tool info' when
determining snap bases:
# Have "base" initialized to something meaningful.
if self.is_core_snap():
snap_data["snap"]["base"] = ""
elif snap_data["snap"].get("base") is None:
snap_data["snap"]["base"] = "core"
The snap store does not return a base if the base is core which makes
this necessary. This patch looks for the base in 'snap info' output
and if none is found (and the snap is not snapd or core) it assumes the
base is 'core' and installs it. This restores the behavior lost in the
migration from snap-tool to snap cli.
snap-tool was added to support a deprecate cohort-key feature of the snap store.
Recent changes in snap assertions have added additional fields which snap-tool
is not retrieving. This resulted in snap install failures on first boot.
This patch removes snap-tool and returns to using the snap cli. This ensures
snap downloads will function without odd incompatibilities.
- drop ds-identify policy, not needed with improved cloud config
- drop disabling network, doesn't work with ip=
- fixup setting up the INSTALLER_ROOT mountpoint
- enable cloud-init
- make cloud-init handle the default/baked in networking configuration
- install and enable openssh-server for the installation only
- provide cloud.cfg that generates random installer user password
- disable subiquity on sclp_line0 line based console
Instead of injecting an empty resolv.conf with an includes.chroot, we'll
inject it in late with a hook. The empty resolv.conf breaks DNS early in
the build, and causes some binary hooks to fail.
Subiquity images mount /usr/lib/modules from a squashfs, which systemd
tries to unmount on shutdown, whilst they are still being in use. As
systemd-udevd kmod built-in's libkmod has modules.* files
memorymapped. This produces an warning on shutdown, flooding the
screen with messages as systemd switches to a more verbose
output. Specify LazyUnmount=yes on that mount, such that unmount call
succeeds without flooding the shutdown log.
When we encounter a failure in 'lb binary' the launchpad builders can
only surface the build output from stdout. If the binary hook failure
implicates the archive we can not determine fault without the apt
proxy log. This patch will dump the proxy log to stdout to aid in
debugging these failures.
As discussed and explained to Jibel and Didier on 17th of June, based
on canary image build logs, it showed clearly that lb controlled
initrd was only built once at the very first layer by
chroot_hacks. However, that is the wrong layer to build the final
initrd at, as at this point casper is not present yet and is not part
of the build.
Thus insure that chroot_hacks only runs at the live layer.
Ideally a subset of chroot_hacks should run on every layer, as each
layer should be squeaky clean, and most of layers without
initrds. However, jibel & didrocks are still implementing requested
patches to unbreak layer images and make each layer smaller. Hence
this minimal portion of the overall required work.
Signed-off-by: Dimitri John Ledkov <xnox@ubuntu.com>
David Krauser
Dimitri John Ledkov
Steve Langasek
Balint Reczey
Łukasz 'sil2100' Zemczak
Philip Roche
Jinming Wu, Patrick
Pat Viafore
Robert C Jennings
Michael Hudson-Doyle
Iain Lane
Adam Conrad
Colin Watson