[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [Linux-cluster] NFS configuration question



Eric Kerin wrote:

On Tue, 2005-07-12 at 03:09 +0200, Birger Wathne wrote:

I did a make install and service rgmanager stop/start.

Now, status checks seem to work. After exportfs -r all file systems eventually reappear.



Good to hear that fixed the monitoring for you.


Do you want the bad news as well?

The > 14 character long export path bug is still there in the FC4 branch.
If the exported path is > 14 characters it will always fail status check and get reexported. This is because exportfs breaks lines after the exported path if it is long. I think exportfs shouldn't do this, but I guess the cluster software will have to work around it.


The line that fails is:
exportfs | grep -q "^${OCF_RESKEY_path}\ .*${OCF_RESKEY_target}"

A proper fix would be a filter to remove the offending line breaks. My quick and dirty (and far less portable) fix is to use this instead:

grep -q "^${OCF_RESKEY_path}[   ]*${OCF_RESKEY_target}(" /var/lib/nfs/etab

Note that my code is a little bit more restrictive. It makes shure the target matches exactly, while the existing code will match if ${OFC_RESKEY_target} is a prefix, won't it? In other words, if you look for an unqualified target and there is an export to a fully qualified host with the same hostname (but possibly in a different domain... hard to tell) the existing code will match. I guess there should be a '$' at the end of the regexp.

Should I bugzilla this, or will you take it? I don't remember my user, but I think I created one. Perhaps it's simply birger. :-)

--
birger


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]