> Folks
>
> My experiments with installing ZFS-FUSE in a Centos 6 system reveal
> behavior different from that observed in Centos 5.6.
>
> The version of ZFS-FUSE is that provided on the EPEL repository, and
> was installed on a 32-bit machine. I use it because of its
> deduplication facility.
>
> In Centos 5.6, the facility worked as advertised.
>
> In Centos 6, however, the system "hung" in initialization upon
the
> first reboot after installing ZFS-FUSE and creating a pool. By
> "hung", I mean that the cute sliding blue bar that progresses
during
You may want to disable the cute sliding blue crap by editing the grub
options and remove "rhgb quiet" from it. That's what I do most of
the
times because I want to see what's going on.
> bootstrap got to nearly complete and then stood there for at least 20
> minutes. The identical computer, configured without ZFS-FUSE,
> completed the system initialization within normal time.
>
> Following some clues found by googling the bug, I suspected that
> AUDITD and ZFS-FUSE had a conflict. After several experiments, I
> found what appears to work.
>
> REMEDY:
>
> When you install ZFS-FUSE, it will start the zfs-fuse service, and
> will also automatically issue a "chkconfig zfs-fuse on" command.
The
> remedy is to prevent zfs-fuse from running during initializtion by issuing
> chkconfig zfs-fuse off
> and starting the service after the initialization is complete.
Hm, I've just started running a server with SELinux enabled and already
learned why it was a good thing to stay away from it until now. I guess
that's also where your problems come from: the selinux-policy does not
have support for zfs-fuse, or for the way you are using it. you can check
audit.log to see if SELinux prevents zfs-fuse from starting up. I may be
wrong of course.
Regards,
Simon
>
> I personally chose to start it via a crontab entry as root:
> @reboot <something>
> where "something" could be the direct service call, or (as was in
my
> case) the starting of a script to do a bunch of startup things
> including starting zfs-fuse service.
>
>
> IS THIS THE DESIRED BEHAVIOR
> I sort of doubt it, and I'm sure it will eventually be sorted out by
> the zfs-fuse developers.
>
> David Kurn
>
>
>
>
> _______________________________________________
> CentOS mailing list
> CentOS at centos.org
> http://lists.centos.org/mailman/listinfo/centos
>