[systemd-devel] One shot service failure on Fedora 37

[Bill Steinberg]

> "Directly" means *not using wrapper scripts.* You can put command line arguments in ExecStart.

Is there a way to run a shell script in a systemd which runs this type of command: "eval $DCC_LIBEXEC/dccifd $DCCIFD_ARGS”. The Distributed Checksum Clearinghouses shell scripts end up running this command. I’d rather use their scripts directly instead of having to extract the necessary parts required for a systemd forking type.

> On Apr 18, 2023, at 12:02 AM, Mantas Mikulėnas <grawity at gmail.com> wrote:
> 
> On Tue, Apr 18, 2023, 02:59 Bill Steinberg <bill at ponusridge.com <mailto:bill at ponusridge.com>> wrote:
>> 
>> Hi Barry,
>> 
>> Thanks for the response. Answers inline below.
>> 
>>> On Apr 17, 2023, at 5:09 PM, Barry <barry at barrys-emacs.org <mailto:barry at barrys-emacs.org>> wrote:
>>> 
>>> 
>>> 
>>>> On 17 Apr 2023, at 19:05, Bill Steinberg <bill at ponusridge.com <mailto:bill at ponusridge.com>> wrote:
>>>> 
>>>> Hello systemd devel,
>>>> 
>>>> I have a systemd service that I’ve run on prior versions of fedora which fails to start via systemd on Fedora release 37. It is a oneshot service that starts the distributed checksum clearing house’s dccifd service via a shell script. Here is the definition of the service:
>>>> 
>>>> [Unit]
>>>> Description=Distributed Checksum Clearinghouses dccifd daemon
>>>> After=syslog.target network.target
>>>> 
>>>> [Service]
>>>> Type=oneshot
>>> 
>>> Oneshot seems wrong.
>>> 
>>>> RemainAfterExit=yes
>>>> ExecStart=/var/dcc/libexec/rcDCC -m dccifd start
>>> Does this run a background daemon?
>> 
>> Yes, the rcDCC shell script starts and runs a linux executable, a background daemon as you call it.
> 
> 
> A "background daemon" is just Type=forking.
> 
>> 
>>> Can you just run that daemon directly?
>> 
>> I could run the shell script directly to start the dccifd executable however if the fedora linux server is rebooted I would need to remember to run the shell script manually. I’d like the dccifd daemon to start automatically when the fedora linux server is started. Isn’t systemd meant for this?
> 
> 
> "Directly" means *not using wrapper scripts.* You can put command line arguments in ExecStart.
> 
>> 
>>> Hopefully that program can be run without demonising.
>>> 
>>> 
>>>> ExecStop=/var/dcc/libexec/rcDCC -m dccifd stop
>>> If it is oneshot it does not need a stop
>> 
>> Is there another type that should be used besides oneshot? I may want to run systemctl stop dccifd.service, for example when dccifd is being upgraded to a new version.
>> 
>> The dccifd executable is started and stopped using a shell script. It is not run directly. One reason is that the shell script contains the arguments that are passed to the dccifd linux executable.
> 
> 
> That's still just Type=forking.
> 
> Make sure the script `exec`s the main process rather than spawning it underneath as usual.
> 
> But why put those arguments in a shell script? Isn't systemd meant for this?
> 
>> 
>>> 
>>>> Restart=no
>>>> 
>>>> [Install]
>>>> WantedBy=multi-user.target
>>>> 
>>>> 
>>>> The error in the journalctl log is:
>>>> 
>>>> systemd[1]: Starting dccifd.service - Distributed Checksum Clearinghouses dccifddaemon…
>>>> systemd[1]: dccifd.service: Main process exited, code=killed, status=11/SEGV
>>>> systemd[1]: dccifd.service: Failed with result 'signal’.
>>>> systemd[1]: Failed to start dccifd.service - Distributed Checksum Clearinghouses dccifddaemon.
>>>> 
>>>> The two scripts in ExecStart and ExecStop run successfully outside of systemd. Any info as to why systemd fails when executing these scripts would be appreciated.
>>>> 
>>>> Best,
>>>> Bill Steinberg

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/systemd-devel/attachments/20230418/b5cc4a23/attachment.htm>