Comment 7 for bug 1428486

Revision history for this message
Steve Langasek (vorlon) wrote : Re: [Bug 1428486] Re: Only start rpc.statd if $NEED_STATD

On Thu, Mar 12, 2015 at 07:54:05AM -0000, Martin Pitt wrote:
> An empty value is "auto", and both the upstart job (shell commands in
> gssd-mounting.conf) and rpc-gssd.service
> (ConditionPathExists=/etc/krb5.keytab) do that "auto" detection. But
> indeed we don't currently check this value in the systemd jobs; I wonder
> if we need to? (e. g. RHEL/Fedora just use these systemd units). I. e.
> isn't the job of an init system to figure that out for you?

> Steve, do you have an opinion on this? Do we need the ability to
> forcefully start gssd without krb, or forcefully suppress it with krb?

In the previous sysvinit/upstart detection, the reason to support force
starting of gssd is to enable running gssd for an NFS mount that isn't
configured in /etc/fstab.

In the sysvinit/upstart detection - which was based on fstab entries and not
on the presence/absence of /etc/krb5.keytab - I can think of no reason to
want to suppress gssd from starting when it's started by autodetection. And
if someone wants to suppress gssd start under systemd, I think they should
use the 'update-rc.d' interface instead.

So I think these are both corner cases, and migrating the gssd setting from
/etc/default/nfs-common is a low priority.