[vox-tech] NFS locking problem after upgrade

[Chris Jenks]

   Dear List,

   I have exported /var/cache/apt and /var/lib/apt/lists with NFS to allow 
several PCs on my home network to share the same deb package lists and 
archives. This system was working fine until about a month ago, perhaps 
due to a routine update/upgrade. Since then, when I try apt-get or 
aptitude update on either of two of the clients (spongebob and pinky), 
they hang at "Reading Package Lists... 0%" and require a reboot to break 
the spell. /var/log/messages on the clients complain that the server is 
not responding. Two other clients - a workstation (calvin) and a router 
(brain) - both work properly, as does the server (cruncher). All five 
machines use differently customized 2.6.8 Debian-patched kernels, each 
configured to the machine's hardware and function. All five machines run 
Debian testing. The two clients that fail are desktop PCs, while the two 
that work are Thinkpad 701C 486s.
   The mount command (e.g., mount -t nfs -o rw,tcp,intr 
cruncher:/var/cache/apt /mnt/apt) returns instantly without error from all 
clients, and the mounted directories can be entered, viewed and written to 
without a problem. Perhaps the problem is with file locking. rpcinfo -p 
lists portmapper, nlockmgr, nfs, status and mountd on cruncher or 
spongebob (spongebob is also an nfs server), and portmapper, status and 
nlockmgr on brain, calvin or pinky.
   The only other new symptom is that ssh-ing from one workstation to 
another sometimes gives an "unknown host" error, which goes away if I ping 
the remote workstation from the router (brain).
   Since I suspected an upgrade may have led to this problem, I tried 
looking for old versions of mount, nfs-common and liblockfile1 in 
/var/cache/apt/archives. Only nfs-common had an old verison, and 
reinstalling that on pinky had no effect.
   Any and all suggestions what to try next would be appreciated.

   Yours,

     Chris