systemd 254 With New Soft Reboots Feature, systemd-battery-check

[phoronix]

Phoronix: systemd 254 With New Soft Reboots Feature, systemd-battery-check

Systemd 254 is out today in time for appearing in the late-2023 Linux distribution releases...

https://www.phoronix.com/news/systemd-254-Released

[sarfarazahmad]

hmm, quite a few fancy things for memory bits.

[RealNC]

The soft reboot thing immediately reminded me of the quick reboot feature of qemm386 back in the DOS days.

[SpyroRyder]

Can someone tell me what the use case for a soft reboot is? When is it really useful ovee just hard rebooting?

[PedroHLC]

Can someone tell me what the use case for a soft reboot is? When is it really useful ovee just hard rebooting?

Besides being very usecase-less for non-WSL setups, a portion of it generated some stress inside the team as they pushed changes Poettering disapproved, during his vacations: https://github.com/systemd/systemd/p...ment-164617339. And then the author ended up being very passive-aggressive in the revert PR: https://github.com/systemd/systemd/p...ent-1648643043

[mether]

Can someone tell me what the use case for a soft reboot is? When is it really useful ovee just hard rebooting?

If you have a large server for instance, a hard reboot requires initializing all the hardware again and that can take quite a long time (my personal experience is with a large multi terabyte database server taking over half an hour). A soft reboot is considerably quicker. Image updates are another use case when combined with kernel live patching with something like kpatch

[CommunityMember]

Can someone tell me what the use case for a soft reboot is? When is it really useful ovee just hard rebooting?

Especially on server class systems bring up the hardware can take quite some time (identifying the hardware, loading firmware, etc. can take minutes). Just restarting all of user space (with the ability to pass along certain file descriptors such that services can run at least partially uninterrupted) can result in a system that is ready to function much faster (sometimes).

[Britoid]

I can imagine this being useful for ostree updates. If the kernel doesn't change, there's really no reason to reboot the hardware if you can reboot the userland.

[Vistaus]

I can imagine this being useful for ostree updates. If the kernel doesn't change, there's really no reason to reboot the hardware if you can reboot the userland.

Exactly. People are talking about servers and they're not wrong, but I don't constantly reboot either. I only reboot when there's something important to reboot for. So if I can reduce that number even further by soft rebooting, then I'm even more happy!

[caligula]

Especially on server class systems bring up the hardware can take quite some time (identifying the hardware, loading firmware, etc. can take minutes). Just restarting all of user space (with the ability to pass along certain file descriptors such that services can run at least partially uninterrupted) can result in a system that is ready to function much faster (sometimes).

Not only server class! E.g. Asrock Rack workstation/server ITX motherboards with no PCIe cards & one memory module may spend 2-3 minutes booting the firmware whereas an full size ATX board & Haswell Core i7 with 20 drives and PCIe cards + 4 memory modules might boot the firmware + systemd-boot + kernel + window manager in 8 seconds. The server class mini ITX hardware is so much better that it spends 50 times more booting the firmware!