finit equivalent of oneshot (RemainAfterExit) services

[aanderse]

consider the following systemd service for nft:

[Unit]
Description=nftables static rule set
Documentation=man:nftables.service(8)
Wants=network-pre.target
Before=network-pre.target shutdown.target
Conflicts=shutdown.target
DefaultDependencies=no
ConditionPathExists=@pkgsysconfdir@/rules/main.nft

[Service]
Type=oneshot
RemainAfterExit=yes
StandardInput=null
ProtectSystem=full
ProtectHome=true
ExecStart=@sbindir@/nft 'flush ruleset; include "@pkgsysconfdir@/rules/main.nft"'
ExecReload=@sbindir@/nft 'flush ruleset; include "@pkgsysconfdir@/rules/main.nft"'
ExecStop=@sbindir@/nft flush ruleset

[Install]
WantedBy=sysinit.target

as far as i understand there isn't a direct equivalent in finit, even if you use the sysv compatibility layer because of the way runelevls work (example sysvinit file for reference)

so this is what i came up with:

task [S234] name:nftables :start <service/syslogd/ready> nft -f /etc/nftables.conf
task [06]   name:nftables :stop  <service/syslogd/ready> nft flush ruleset

does this seem like the "right" answer?

[troglobit]

Despite being an init geek I'm actually not that well versed in systemd. In this case it seem like the real magic happens because of oneshot + RemainAfterExit=true, right? Because that means the 'service' remains active even though the call to nft exits.

Finit doesn't have a RemainAfterExit=yes but maybe we should. I have a feeling that could be quite useful!

For now, I think I would go with a wrapper daemon, like this nft-helper I made a while ago. Maybe even simpler, a sleep inf "daemon" complemented with pre:/post:/reload: scripts ...

[aanderse]

I'm actually not that well versed in systemd. In this case it seem like the real magic happens because of oneshot + RemainAfterExit=true, right? Because that means the 'service' remains active even though the call to nft exits.

ah... sorry, i'll keep that in mind! but yes, you got it, spot on 👍

Finit doesn't have a RemainAfterExit=yes but maybe we should. I have a feeling that could be quite useful!

definitely, this would be very useful! would you consider this a task still, or move it into the service category?

For now, I think I would go with a wrapper daemon, like this nft-helper I made a while ago. Maybe even simpler, a sleep inf "daemon" complemented with pre:/post:/reload: scripts ...

very cool! thanks for linking!

[troglobit]

[snip]

Finit doesn't have a RemainAfterExit=yes but maybe we should. I have a feeling that could be quite useful!
definitely, this would be very useful! would you consider this a task still, or move it into the service category?

Not sure, but it feels more and more that it's time to dust off the plans for #148 and systemd has this clever notion of just calling everything services and use Type= instead, so that's still on the table.

Side node, barely a day goes by when my colleague @wkz admits to me that he's still not learned the difference between run and task, so maybe Type=oneshot and BlockEverythingUntilDone=True would more user friendly as well?

[snip]

[aanderse]

he's still not learned the difference between run and task

whoa! i find that very shocking because there is such a large conceptual difference between the two... i thought the documentation was very clear ind distinguishing them... to me they aren't the same thing at all

but if other people see it that way then maybe it makes sense 🤷‍♂️

[troglobit]

TBH he probably just confuses the keywords. From their names it's not really obvious what they do 😅

I think clarity and "conciseness" (is that a word?) wins over brevity. But naming things will forever be the biggest problem in this space.