possibly problem with rawhide (systemd-217?): "Found ordering cycle on basic.target/start"

Lennart Poettering mzerqung at 0pointer.de
Fri Oct 31 15:30:31 UTC 2014


On Fri, 31.10.14 16:20, Lennart Poettering (mzerqung at 0pointer.de) wrote:

> On Fri, 31.10.14 16:13, Michal Schmidt (mschmidt at redhat.com) wrote:
> 
> > On 10/31/2014 03:42 PM, Kevin Fenzi wrote:
> > > On Fri, 31 Oct 2014 14:01:12 +0100
> > > Lennart Poettering <mzerqung at 0pointer.de> wrote:
> > > 
> > >> So the problem appears to be that gssproxy.service been ordered before
> > >> remote-fs-pre.target. That target is ordered before
> > >> basic.target. However gssproxy.service also is ordered after
> > >> basic.target (simply because all services by default are ordered
> > >> before basic.target, unless they explicitly specify
> > >> DefaultDependencies=no), hence there's an ordering cycle.
> > >>
> > >> Most likely some NFS maintainers tried to move gss-proxy.service into
> > >> the early boot, and didn't set DefaultDependencies=no.
> > >>
> > >> That said, services running in early boot must be written in a
> > >> specific style (i.e. not assume /var to be around, and suchlike), I
> > >> do wonder if gssproxy is ready for that.
> > >>
> > >> Anyway, long story short: file a bug against the gssproxy package.
> > > 
> > > I don't think this explains all the problems folks are having with
> > > systemd-217. 
> > 
> > I wonder if the new ordering dependency between
> > systemd-journal-flush.service and systemd-tmpfiles-setup.service
> > (added in 74055aa76 "journalctl: add new --flush command and make use
> > of it in systemd-journal-flush.service") participates in the ordering
> > cycles.
> 
> Ahh, indeed. It moves remote-fs.target into the early-boot where it
> doesn't belong.
> 
> My fault.
> 
> Will drop the remote-fs.target dep from the flush service.
> 
> Thanks for tracking this down.

Would be good if somebody who ran into this problem could check if
this change fixes it:

http://cgit.freedesktop.org/systemd/systemd/commit/units/systemd-journal-flush.service.in?id=919699ec301ea507edce4a619141ed22e789ac0d

(the actual commit unfortunately contains an unrelated change to
nspawn, I fucked that up, sorry. Ignore everything but the change to
systemd-journal-flush.service)

Thanks,

Lennart

-- 
Lennart Poettering, Red Hat


More information about the devel mailing list