[systemd-devel] _netdev for system root mount?
Thomas Blume
Thomas.Blume at suse.com
Mon Mar 16 10:46:15 UTC 2020
On Mon, 16 Mar 2020, Mantas Mikulėnas wrote:
> On Mon, Mar 16, 2020 at 11:52 AM Thomas Blume <Thomas.Blume at suse.com> wrote:
> On Fri, 13 Mar 2020, Alexander E. Patrakov wrote:
>
> > On Fri, Mar 13, 2020 at 7:07 PM Andrei Borzenkov <arvidjaar at gmail.com> wrote:
> >
> >> And what is the "official" way to prevent various units required by root
> >> mount from being stopped during shutdown? There could be arbitrarily
> >> deep stack (NIC - iSCSI - multipath - raid - lvm - crypto - ...).
> >
> > https://systemd.io/ROOT_STORAGE_DAEMONS/
>
> So, that means that the iscsi unit files in the running system are not
> designated and supported for system root, right?
>
>
> I've only used iSCSI for data volumes, but... how could the rootfs possibly be dependent on a process running *from the same rootfs*? I mean, the iSCSI or NBD daemons have to start from
> somewhere else *before* the rootfs is set up and mounted, don't they?
>
> If the rootfs is iSCSI-based or NBD-based, then I would expect the corresponding daemons to be started from the *initramfs*, meaning they wouldn't be managed as rootfs .service units in
> the first place -- and they wouldn't be stopped along with other .service units either.
>
> (Note that if your initramfs itself is systemd-based, then it has a completely separate set of units, with its own boot order and everything.)
>
>
> What about the network.service?
> I guess this should be also unsupported for the network device providing system
> root?
>
>
> network.service is a distro-specific addition. I don't know what it supports on your system.
>
> But in general, network configuration tools often have an option to leave an interface configured upon exit. For example systemd-network has KeepConfiguration=.
>
>
> Finally, can I also conclude that the _netdev parameter as an ordering
> constraint for the network block device is also not supported for system root?
>
>
> Same comment as above... how is systemd supposed to put other units before the rootfs, if they're started *from* the rootfs?
I see the logic, thanks.
Still, I think ths should better be documented, at least for the _netdev parameter.
Best regards
Thomas Blume
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5
90409 Nürnberg
Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Felix Imendörffer
More information about the systemd-devel
mailing list