Bug#975289: Acknowledgement (systemd: always reporting unit file has "changed on disk" when override.conf is present)

Richard van den Berg richard at vdberg.org
Wed Nov 25 11:56:33 GMT 2020


On 25-11-2020 02:11, Michael Biebl wrote:
> Can you break this down to a more minimal test case which would make
> this easily reproducible with systemd from unstable?

This reproduces the issue with 246.6-2~bpo10+1:

1) Create a normal unit file /etc/systemd/system/test-start.service with 
contents

[Unit]
Description=Test start

[Service]
Type=oneshot
ExecStart=/usr/bin/mailx -s "message from test-start.service" root < 
/dev/null

2) Create an instantiated unit file 
/etc/systemd/system/test-warning at .service with contents

[Unit]
Description=test warning

[Service]
Type=oneshot
ExecStart=/usr/bin/mailx -s "service %i at %H has failed" root < /dev/null

3) Create a system wide drop-in 
/etc/systemd/system/service.d/test-warning.conf with contents

[Unit]
OnFailure=test-warning@%p.service

4) Create drop-in /etc/systemd/system/test-start.service.d/override.conf 
(by executing "systemctl edit test-start") with contents

[Service]
ExecStartPost=/usr/bin/mailx -s "message from override.conf" root < 
/dev/null

5) systemctl daemon-reload

6) systemctl status test-start

7) goto 5

The "Run 'systemctl daemon-reload' to reload units." warning persists as 
long as /etc/systemd/system/test-start.service.d/override.conf is newer 
than /etc/systemd/system/service.d/test-warning.conf and can be cleared by:

touch /etc/systemd/system/service.d/test-warning.conf
systemctl daemon-reload



More information about the Pkg-systemd-maintainers mailing list