[Gluster-devel] Stale NFS file handle, then EINVAL

[Emmanuel Dreyfus]

Another bug: first find complains of a stale NFS filehandle
find: /gfs/usr/src/tools/binutils/../../gnu/dist/binutils/cpu/sh.opc: Stale
NFS file handle

Then ls -l or cat gets EINVAL
# ls -l  /gfs/usr/src/tools/binutils/../../gnu/dist/binutils/cpu/sh.opc
ls: /gfs/usr/src/tools/binutils/../../gnu/dist/binutils/cpu/sh.opc:
Invalid argument
# ls -l  /gfs/usr/src/gnu/dist/binutils/cpu/sh.opc                     
ls: /gfs/usr/src/gnu/dist/binutils/cpu/sh.opc: Invalid argument 


The file exists on the backends (both get the same results)
server1# ls -l /export/*/usr/src/gnu/dist/binutils/cpu/sh.opc
---------T  1 root  wheel     0 Jul 19 09:47
/export/wd1a/usr/src/gnu/dist/binutils/cpu/sh.opc
-rw-r--r--  1 root  wheel  2485 Feb  2  2006
/export/wd3a/usr/src/gnu/dist/binutils/cpu/sh.opc

server1# lsextattr /export/*/usr/src/gnu/dist/binutils/cpu/sh.opc
/export/wd1a/usr/src/gnu/dist/binutils/cpu/sh.opc
       trusted.glusterfs.dht.linkto   
       trusted.gfid
/export/wd3a/usr/src/gnu/dist/binutils/cpu/sh.opc
       trusted.afr.gfs-client-3    
       trusted.afr.gfs-client-2         
       trusted.gfid

server1# getextattr -x trusted.glusterfs.dht.linkto
/export/wd1a/usr/src/gnu/dist/binutils/cpu/sh.opc
/export/wd1a/usr/src/gnu/dist/binutils/cpu/sh.opc       
   000   67 66 73 2d 72 65 70 6c 69 63 61 74 65 2d 31 00    gfs-replicate-1.

Server logs say nothing. Client log says this:

[2011-07-19 19:45:03.186431] W [stat-prefetch.c:3627:sp_stat_helper]
0-gfs-stat-prefetch: lookup-behind has failed for path
(/usr/src/gnu/dist/binutils/cpu/sh.opc)(Stale NFS file handle), unwinding stat
call waiting on it 
[2011-07-19 19:45:38.422606] W [fuse-bridge.c:413:fuse_attr_cbk]
0-glusterfs-fuse: 1378: STAT() /usr/src/gnu/dist/binutils/cpu/sh.opc => -1
(Invalid argument)

I have to unmount and remount the filesystem on the client to fix the problem.
What happens? 

-- 
Emmanuel Dreyfus
http://hcpnet.free.fr/pubz
manu at netbsd.org