I was prompted by trawling over Bugs Online to open a call with NetApp about the following problem, which I've known about for a year or more. The call isn't making much progress (we won't go into whose fault that is ...) and as the whole thing was really a request for information ("did you know about this infelicity? is it going to be fixed some year?") I can't conscientiously exert much pressure. But I can waste your time asking whether you've observed the same effect, can't I? :)
As I am sure you are all aware, although ".snapshot" can be looked up in any directory, it is only visible to readdir() in NFS mount points (5.3 System Administrator's Guide, p.406: "Actually, the .snapshot in the mount point is "real" to make the pwd command work ...").
The problem is that sometimes ".snapshot" isn't visible to readdir() *even* in mount point directories. This seems to happen only in directories with more than one 4K block (it's notable that "ls -f" shows .snapshot coming after the first block's worth in such directories, if it shows up at all), and rather rarely even then (2 out of over 150 user home directories of >4K that I looked at today), but as long as the directory isn't tinkered with the state remains the same.
I've observed this with several OnTap 5.2.x & 5.3.x releases, and several different Solaris releases as clients. It *could* be an NFS client implementation problem, I suppose, but I don't think I believe that.
Does this sound familiar to anyone?
Chris Thompson University of Cambridge Computing Service, Email: cet1@ucs.cam.ac.uk New Museums Site, Cambridge CB2 3QG, Phone: +44 1223 334715 United Kingdom.