race: systemctl --wait start UNIT followed by journalctl -u UNIT: output sometimes missing

[edsantiago]

systemd version the issue has been seen with

systemd-253.7-1.fc38.x86_64

Used distribution

Fedora 38

Linux kernel version used

6.4.6-200.fc38.x86_64

CPU architectures issue was seen on

x86_64

Component

journalctl

Expected behaviour you didn't see

[ context: SERVICE emits one line of output and exits immediately ]

# systemctl --wait start SERVICE
# journalctl -u SERVICE
...date..host... one line of output!

Unexpected behaviour you saw

Sometimes fail to see output.

It is possible that this is the same as #2913 or #3767 or perhaps another issue I couldn't find. Or perhaps there's a --flush-output option to systemd start. If so I apologize.

Steps to reproduce the problem

journald-race.sh

Additional program output to the terminal or log subsystem illustrating the issue

...........................................................................
**FAILED foo-ZHFwFXrHH5PcmrEuilyD.service in 170 seconds

-- No entries --
retrying after 2 seconds:
Aug 02 14:28:15 host bash[144555]: GOT HERE
Aug 02 14:28:15 host systemd[1]: Started foo-ZHFwFXrHH5PcmrEuilyD.service - foo.
Aug 02 14:28:15 host systemd[1]: foo-ZHFwFXrHH5PcmrEuilyD.service: Deactivated successfully.

[yuwata]

That's expected. Please specify Type=oneshot for such testcase. Please see systemd.service(5) for more details.

[edsantiago]

@yuwata thank you for your quick response. I've modified the reproducer script as follows and I still see the same symptom:

@@ -15,6 +15,7 @@
 
 [Service]
 ExecStart=/bin/bash -c "echo GOT HERE"
+Type=oneshot
 EOF
 
     systemctl daemon-reload

[yuwata]

Then, such 'race' is still expected. Simply, journald has not processed the output.