Important to note that this bug is a sure way to corrupt user data in
production environments.
E.g. databases running in Docker containers have no clue that they're
being asked to shut down gracefully, because of blocked signals. When
under load, a database would be forcefully killed with SIGKILL
'docker wait' might be handy, it prints the container's exit code.
$ CT=$(docker run -d nginx); docker stop $CT > /dev/null && docker wait
$CT; docker rm $CT > /dev/null
Would print 0 (graceful exit) when the signals are coming through, and
137 (killed with SIGKILL) on a buggy distro
root@noble
Important to note that this bug is a sure way to corrupt user data in
production environments.
E.g. databases running in Docker containers have no clue that they're
being asked to shut down gracefully, because of blocked signals. When
under load, a database would be forcefully killed with SIGKILL
Hello. I've tested the noble-proposed packages on 24.04 amd64 and
aarch64, signals are fixed after a reboot as expected, containers are
being stopped normally.
Upgrading docker.io to 26.1.3-0ubuntu1 does the job, while containerd
version doesn't seem to be a factor.
** Tags removed: verification-
Given the verification is complete with the verification-done-noble tag
set, what is the expected timeline for the fixed packages to transition
from noble-proposed to noble-updates so they are available to all Ubuntu
24.04 users?
--
You received this bug notification because you are a member of U
