On 01/30/2011 08:40 PM, JB wrote:
Terry Barnaby<terry1<at> beam.ltd.uk> writes:
> ...
Your analysis is very plausible.
I remember from Slackware (many years ago ...) - it took explicit steps to
TERM active processes, reasonably waited for them, and then killed them.
I tried to follow the selinux line as well.
The support for nfs home dirs caused problems in the past.
You have a mix of nfs3 and nfs4, and the nfs4 may be buggy (some selinux and
'mount' related features are scheduled to be ironed out in F15).
I would dive in, just for kicks, and tried both cases:
- switch selinux to permissive mode ; this may not be enough, so ...
- disable selinux entirely
You can do it on the kernel command line or /etc/sysconfig/selinux - but you
have to shutdown twice in order to test the halt script.
JB
Hi,
Thanks for the info. selinux is actually disabled on all of these
systems. I'm not sure why /home uses NFS 3 while the others use NFS4.
They are from the same server and there is no specific config
for 3 or 4, so on Fedora 14 I would have expected them to be 4.
The server is Fedora 14 as well.
I think it is the /home mount that is likely to be causing the
problem (as the GUI programs are probably accessing files in the users
directory) and this uses NFS V3. So I wouldn't have expected the
NFS V4 code to be much involved here.
Note this is a Fedora 14 issue. Fedora 12 has been running in this environment
for more than a year with the same setup without this issue.
I could add a delay before the unmount the NFS file systems as see if this
reduces the problem.
Terry