Is there a way to make systemctl start synchronous

Question

I'm developing/have developed a unit file for systemd

[Unit]
Description=FreeRADIUS multi-protocol policy server
After=syslog.target network.target
Documentation=man:radiusd(8) man:radiusd.conf(5) http://wiki.freeradius.org/ http://networkradius.com/doc/

[Service]
EnvironmentFile=-/etc/sysconfig/radiusd
ExecStartPre=/usr/sbin/radiusd $FREERADIUS_OPTIONS -Cxm -lstdout
ExecStart=/usr/sbin/radiusd $FREERADIUS_OPTIONS -fm
Restart=on-abnormal

[Install]
WantedBy=multi-user.target

Experiencing an issue where calling systemctl radiusd start does not return an error code, even if /usr/sbin/radiusd -fm exits with one.

Is there a way to make systemctl act synchronously? i.e. waiting for a set period before returning and indicating the service was successfully/unsuccessfully started.

I'm fine with changing to Type=forking or any of the other options, like dbus (and writing the code to integrate with dbus) if that means systemctl will exit with an error indicating that the service failed to start.

To pre-empt the obvious question, yes, after start systemd does see the unit as failed.

bash-4.2# systemctl status radiusd
radiusd.service - FreeRADIUS multi-protocol policy server
   Loaded: loaded (/usr/lib/systemd/system/radiusd.service; enabled)
   Active: failed (Result: start-limit) since Wed 2015-08-12 12:26:18 EDT; 19s ago
     Docs: man:radiusd(8)
           man:radiusd.conf(5)
           http://wiki.freeradius.org/
           http://networkradius.com/doc/
  Process: 10610 ExecStart=/usr/sbin/radiusd $FREERADIUS_OPTIONS -fm (code=exited, status=1/FAILURE)
 Main PID: 10610 (code=exited, status=1/FAILURE)

and yes systemctl status radiusd does return a non 0 exit code. Just annoying to integrate this with salt stack. Currently, without synchronous start, the bundled saltstack pkg module reports the service as running, after applying config/code updates that cause service failures.

I meant the unit file that came with the freeradius package. — Michael Hampton, Aug 12 '15 at 17:55
I wrote the unit file that came with the freeradius package. — Arran Cudbard-Bell, Aug 12 '15 at 17:55
https://github.com/FreeRADIUS/freeradius-server/blob/v3.1.x/redhat/radiusd.service — Arran Cudbard-Bell, Aug 12 '15 at 17:58
You'll want `Type=forking` then. This is also what downstream distributions are using. — Michael Hampton, Aug 12 '15 at 18:02

score 2 · Accepted Answer · answered Aug 12 '15 at 17:52

I think you have already answered your own question.

If you want to fork a daemon as well as want systemctl to return the status code, Type=forking is the way to go. There's Type=oneshot as well. You should use this only when you expect your script/program to exit, as opposed to running as daemon. systemctl actually waits for the ExecStart= program to finish.

From my CentOS 7.1 machine:

[unixguy@infra01 system]$ pwd
/usr/lib/systemd/system
[unixguy@infra01 system]$ find ./ -type f | xargs grep 'Type='  | awk -F: '{print $2}' | sort | uniq -c | sort -nr | head -5
     64 Type=oneshot
     37 Type=forking
     11 Type=notify
      6 Type=idle
      6 Type=dbus
[unixguy@infra01 system]$ find ./ -type f | xargs grep 'Type=forking' | awk -F: '{print $1}' | head
./rc-local.service
./rdisc.service
./tcsd.service
./plymouth-kexec.service
./plymouth-halt.service
./plymouth-poweroff.service
./plymouth-reboot.service
./plymouth-start.service
./rpc-statd.service
./systemd-cfengine-bootstrap.service
[unixguy@infra01 system]$ find ./ -type f | xargs grep 'Type=oneshot' | awk -F: '{print $1}' | head
./systemd-kexec.service
./quotaon.service
./halt-local.service
./initrd-cleanup.service
./initrd-parse-etc.service
./initrd-switch-root.service
./initrd-udevadm-cleanup-db.service
./kmod-static-nodes.service
./systemd-binfmt.service
./systemd-backlight@.service

As you can see the daemon services are using Type=forking and one time execution services/scripts are using Type=oneshot

Is there a way to make systemctl start synchronous

1 Answers1