John Kohl writes:
 > I was using NFS V3 to update a private CVS tree from my own private
 > repository, and I kept getting weird directory errors.  The first time
 > it complained that some directory in the repository didn't exist
 > (although I could see it when I ran 'ls' just after cvs bailed out), and
 > then it complained about failing to create a read lock.
 > 
 > When I remounted the repository using NFS V2, things worked just fine.
 > 
 > NFS Client: i386 -current (pcmcia ne2000 clone ethernet, 75MHz pentium CPU)
 > NFS server: i386 -current (SMC 8432 (dc21040) ethernet, 100MHz pentium CPU)

	Last Friday we saw something similar. While doing some
compilation on machine hola, on a NFS mounted directory belonging to machine
raistlin, the disk of raistlin became full. Some space was released in 
raistlin (by deleteting things in other directories). Since then, things 
in hola began to work in strange ways.
An "ls" in hola, of the directory where the compilation had taken 
place showed only part of the directories and files which were 
actually present there (according to an "ls" in raistlin). Killing the
nfsd in raistlin was enough to get a consistent view of the files.

	Bot raistlin and hola are running NetBSD/i386-1.2. raistlin
is a 486 machine, while hola is a Pentium. Both machines use a 3C509
Eth. card.

		Jesus.

-- 
Jesus M. Gonzalez Barahona         | addr.:  c/ Butarque, 15
Grupo de Sistemas y Comunicaciones |         28911 Leganes, Spain
Departamento de Informatica        | tel: +34 1 624 94 58
Universidad Carlos III de Madrid   | fax: +34 1 624 94 30
e-mail: jgb@gsyc.inf.uc3m.es       | www: http://ordago.gsyc.inf.uc3m.es/~jgb