[Linux-cluster] Problem with nfsclient.sh status checking
Birger Wathne
birger at uib.no
Tue Apr 26 11:22:21 UTC 2005
birger wrote:
> I have now looked into the error messages I get in /var/log/messages
> saying things like
> Apr 26 13:03:08 server1 clurgmgrd[4969]: <notice> status on nfsclient
> "nis-hosts" returned 1 (generic error)
>
> The problem is in nfsclient.sh. It uses the output from exportfs to
> check that file systems are still exported. When the mount point name
> is longer than 14 characters exportfs breaks the line. This breaks
> nfsclient.sh.
>
A little more info... Whenever the status check failed the client got
stale NFS handle for a little while. I guess the entry got unexported
and reexported. Stale NFS handles are a bad thing. Hanging NFS service
is better from a data integrity point of view.
The very quick and dirty fix follows. Edit status check in
/usr/share/cluster/nfsclient.sh to look like this
exportfs -v | tr -d "\n" | sed -e 's/([^)]*)/\n/g'| grep -q
"^${OCF_RESKEY_path}[\t ]*.*${OCF_RESKEY_target}"
the -v option to exportfs is used to give an easy way to put line breaks
in just where they are supposed to be.
--
birger
More information about the Linux-cluster
mailing list