SuccessExitStatus= confusingly does not affect the symbolic name mapping of exit statuses in "systemctl status" output

[joshuamiller01]

systemd version the issue has been seen with

244

Used distribution

CentOS7

Expected behaviour you didn't see

Main PID (as reported by systemctl status <service> exit shows SUCCESS when exit code of ExecStart matches SuccessExitStatus

Unexpected behaviour you saw

Main PID shows status=1/FAILURE, even though unit isn't marked as failed (per 'Active: inactive' from systemctl status <service>

Steps to reproduce the problem
If I create a timer, which runs a oneshot service, and ExecStart returns 1, I expect it to fail, which it does:

[root@somehost ~]# cat /etc/systemd/timers/test.timer
[Unit]
Description=test timer
Requires=test.service

[Install]
WantedBy=timers.target

[Timer]
OnCalendar=*:0/10:0
Unit=test.service

[root@somehost ~]# cat /etc/systemd/timers/test.service
[Unit]
Description=test service
After=network.target

[Service]
Type=oneshot
ExecStart=/bin/false
#SuccessExitStatus=1

[root@somehost ~]# systemctl link /etc/systemd/timers/test.timer
Created symlink /etc/systemd/system/test.timer → /etc/systemd/timers/test.timer.
[root@somehost ~]# systemctl link /etc/systemd/timers/test.service
Created symlink /etc/systemd/system/test.service → /etc/systemd/timers/test.service.
[root@somehost ~]#
[root@somehost ~]# systemctl start test.timer
[root@somehost ~]# systemctl status test.service
● test.service - test service
     Loaded: loaded (/etc/systemd/timers/test.service; linked; vendor preset: enabled)
     Active: failed (Result: exit-code) since Fri 2020-05-08 10:12:47 PDT; 11s ago
TriggeredBy: ● test.timer
    Process: 3377705 ExecStart=/bin/false (code=exited, status=1/FAILURE)
   Main PID: 3377705 (code=exited, status=1/FAILURE)
        CPU: 1ms

May 08 10:12:47 somehost systemd[1]: Starting test service...
May 08 10:12:47 somehost systemd[1]: test.service: Main process exited, code=exited, status=1/FAILURE
May 08 10:12:47 somehost systemd[1]: test.service: Failed with result 'exit-code'.
May 08 10:12:47 somehost systemd[1]: Failed to start test service.

The problem is that, if I fix the service unit to have SuccessExitStatus=1, and reload, subsequent runs of the service show it as partially failing; the unit is correctly marked as 'inactive (dead)', but MainPID reports '(code=exited, status=1/FAILURE)'.

[root@somehost ~]# vi /etc/systemd/timers/test.service  # Uncomment SuccessExitStatus
[root@somehost ~]# systemctl daemon-reload
[root@somehost ~]# systemctl start test.service
[root@somehost ~]# systemctl status test.service
● test.service - test service
     Loaded: loaded (/etc/systemd/timers/test.service; linked; vendor preset: enabled)
     Active: inactive (dead) since Fri 2020-05-08 10:13:30 PDT; 3s ago
TriggeredBy: ● test.timer
    Process: 3380053 ExecStart=/bin/false (code=exited, status=1/FAILURE)
   Main PID: 3380053 (code=exited, status=1/FAILURE)
        CPU: 1ms

May 08 10:13:30 somehost systemd[1]: Starting test service...
May 08 10:13:30 somehost systemd[1]: test.service: Succeeded.
May 08 10:13:30 somehost systemd[1]: Started test service.
[root@somehost ~]#

[joshuamiller01]

Hmm, actually, even if you start with the service having SuccessExitStatus properly set, you still get the partial failure reported: "(code=exited, status=1/FAILURE)" even though the unit isn't marked as failed.

[poettering]

Ah, this is a misunderstanding. The "1/FAILURE" display just shows the numeric code and the textual name of it, i.e. using the table "systemd-analyze exit-status" shows you. SuccessExitStatus= does not affect that textual mapping, it only controls what happens as result of a such an exit status.

I'll prep a PR that tries to clear this up a bit.

[poettering]

documentation fix waiting in #15776