Re: portmap/networking: shutdown script links screwed-up?
Jose Marin <jose@ma.hw.ac.uk> writes:
> I am running several boxes with unstable(woody), and the sutdown process
> on some of them hangs near the end. They are all NFS clients to other
> Linux and Solaris machines. The ones that hang do it when trying to
> access NFS (I think trying to umount the nfs mounts I have in /etc/fstab)
> *after* the portmapper and the networking is gone. They're hung sending
> RPC timeout messages (sendmsg error 101, I recall).
Yep found that too...
There are two bugs open against netbase (#59282, #59286).
> Since the other machines that do not hang are almost identical, I traced
> the difference to this: the rc6.d/rc0.d *links* to the "portmap" and
> "network" scripts in /etc/init.d. The scripts themselves are the same.
>
> The machines that hang on shutdown/halt have:
>
> rc0.d/S10portmap -> ../init.d/portmap
> rc0.d/S15networking-> ../init.d/networking
>
> The machines that do not hang have:
>
> rc0.d/K35portmap -> ../init.d/portmap
> (and do not have any link to "networking")
>
>
> I didn't configure this, the difference is due to the fact that I
> installed woody on the machines that have the problem some weeks later
> than the ones that work. This must have changed in netbase, but I
> couldn't find it in the changelog.
>
> Anyone found this problem/bug? What fix would you suggest?
>
> Incidentally, why isn't there a script "umount-nfs" separated from
> "umountfs"???
That's something I've been pondering about. It's a good idea. Maybe a
wishlist bug ? :-)
And also, the boxes which have this problem cannnot shutdown to single
user mode and come back to multiuser, the portmapper is not
restarted...
Phil.
Reply to: