I have a nfs caching problem between a server and a client.
Here's a simple shell script which can reproduce it when the system
is in this state, which fortunately is not all the time.
[timm@fngp-osg ~/test]$ cat ../test4.sh
#!/bin/bash
i=0
while true
do
if [ ! -s a ]
then
cp /home/timm/a ./a
fi
ln a b
cat b
rm b
i=`expr $i + 1`
echo $i
done
[timm@fngp-osg ~/test]$ pwd
/home/timm/test
fngp-osg is the client.
fnpcsrv1 is the server.
[root@fnpcsrv1 test]# grep fngp-osg /var/lib/nfs/xtab
/export/lsi_home fngp-osg.fnal.gov(rw,sync,wdelay,hide,nocrossmnt,insecure,root_squash,no_all_squash,no_subtree_check,insecure_locks,acl,mapping=identity,anonuid=-2,anongid=-2)
fnpcsrv1:/export/lsi_home/timm on /home/timm type nfs
(rw,noatime,nodiratime,hard,tcp,timeo=600,retrans=8,intr,rsize=32768,wsize=32768,0,0,addr=131.225.167.44)
Part of the output from the script is as follows:
160
ln: `b': File exists
one
two
three
four
five
six
seven
eight
nine
ten
When it is in the wierd state:
On the server, file "b" does not exist in /home/timm/test
[root@fnpcsrv1 test]# ls -l /home/timm/test
total 4
-rw-r--r-- 1 timm oss 49 Mar 3 10:03 a
[root@fnpcsrv1 test]#
On the client
[timm@fngp-osg ~/test]$ ls -l /home/timm/test
total 4
-rw-r--r-- 1 timm oss 49 Mar 3 10:03 a
[timm@fngp-osg ~/test]$ \
timm@fngp-osg ~/test]$ ls -l /home/timm/test/b
-rw-r--r-- 1 timm oss 49 Mar 3 10:03 /home/timm/test/b
[timm@fngp-osg ~/test]$ [timm@fngp-osg ~/test]$ cat /home/timm/test/b
one
two
three
four
five
six
seven
eight
nine
ten
[timm@fngp-osg ~/test]$ rm b
rm: cannot remove `b': No such file or directory
If the file is recreated on the server, the script goes
on its merry way for awhilen, then gets into the loop.
Every once in a while either the ln or the rm process of the script
hangs in state "D"
Neither client or server are particuarly busy right now.
client load is 2, server load is 0.16.
Any idea what might be happening here?
These are both kernels more or less as shipped with red hat enterprise 3
update 5, 2.4.21-37. the server kernel has XFS extensions compiled in
and the server file system is XFS.
Any idea what might be going on here? Anyone ever seen this before?
You might ask--the above script is a stupid thing to be doing, why do it?
There's a third party app that is using this hard link across nfs in
place of file locking for some of the stuff it is doing. That's where
it comes from.
Sometimes this is accompanied by errors on the server
Apr 28 16:39:08 fnpcsrv1 kernel: rpc-srv/tcp: nfsd: sent only -107 bytes
of 132 - shutting down socket
Apr 28 16:39:10 fnpcsrv1 kernel: rpc-srv/tcp: nfsd: sent only -107 bytes
of 132 - shutting down socket
Apr 28 16:39:10 fnpcsrv1 kernel: rpc-srv/tcp: nfsd: sent only -107 bytes
of 264 - shutting down socket
but the above errors also happen even when I'm not running this test
script.
Any help is appreciated.
Steve Timm
--
------------------------------------------------------------------
Steven C. Timm, Ph.D (630) 840-8525 [email protected] http://home.fnal.gov/~timm/
Fermilab Computing Div/Core Support Services Dept./Scientific Computing Section
Assistant Group Leader, Farms and Clustered Systems Group
Lead of Computing Farms Team
-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs