When the initramfs gave a persistent name to
a network interface, renaming it via Name is
not working with the v238 naming scheme even
if NamePolicy is unset.
Switch to the newest network interface naming
scheme to resolve this issue.
https://github.com/flatcar-linux/Flatcar/issues/36
We need to make emerge-gitclone check out its exact remote branch
name, like `refs/remotes/origin/x`, in case of the branch name being
specified as a revision.
See also https://github.com/flatcar-linux/dev-util/pull/2
On DigitalOcean eth0 is still used, so we can just continue
to match for it. Maybe 'e*0' would be possible, too.
There is a problem when IPv6 is activated currently, thus,
just revert the patch.
See also https://github.com/flatcar-linux/bootengine/pull/9
The torcx profile docker-17.03 start not to work since it started
pulling in docker-runc 1.0-rc10, with an error message 'flag provided
but not defined: -console'. That's because Docker 17.03 is incompatible
with runc 1.0-rc4 or newer. [1] So the docker-17.03 profile needs to
pull in docker-runc 1.0-rc2 like before. Bring back docker-runc 1.0-rc2
and its related patches that were removed.
[1] 244c9fc426
In the initramfs persistent ifnames were disabled.
This caused problems because when the renaming was
done after the initramfs, a race made it fail, as
originally reported in
https://github.com/coreos/bugs/issues/1767
Reverts the booteninge commit
"systemd: add module to disable network device renaming"
and widens the networkd match rule for DigitalOcean.
Since sudo 1.8.28, every sudo started printing out a warning like
`/etc/environment: No such file or directory`, when `/etc/environment`
does not exist.
Also sudo <= 1.8.30 is affected by a pwfeedback vulnerability,
CVE-2019-18634. https://seclists.org/oss-sec/2020/q1/48
Update sudo to 1.8.31 from upstream Gentoo, to resolve the issues.
See also https://bugs.gentoo.org/698946.
When /etc/flatcar/docker-1.12 contains "no"
the profile for 18.06 was searched for instead
of 19.03. Keep docker-1.12-no.json in sync
with the latest version in app-torcx/docker/files/.
The host's /etc/nsswitch.conf is a symlink to
/usr/share/google-oslogin/nsswitch.conf
but that is not present in the rkt container.
Do not only bind-mount /etc but also the target
of the symlink. With a broken nsswitch.conf
any entries in /etc/hosts are not considered
which makes problems when a custom DNS server
is used.
When a custom DNS server is used
coreos-metadata-sshkeys@core.service
fails to resolve metadata.google.internal
because only "metadata" is specified in /etc/hosts.
