3

I have my home directory on an NFS file server, our IT team has moved to another server one week ago. Before the move, I killed all my processes with killall -u my_user_name. After the move, I was able to start working normally and resumed work for 2 days. Now, suddenly, all my work done after the move has just disappeared in thin air!

The IT teams claims that I didn't kill all the processes, which left my home directory as stale file handle after the move.

Is it possible that a stale file hand persist for 2 days without any errors? I was working normally saving and creating lots of files and directories without any problem/error message what so ever (except for things being slightly slower, which I just assumed due to NFS cashes not yet updated or something).

4
  • Endless probabilities here.
    – gena2x
    Commented Jun 3, 2015 at 15:25
  • Ok, I can provide any details needed. But is the claim true that one can have a stale file handle for 2 days and keep writing to a "phantom" directory for 2 complete days?
    – Bichoy
    Commented Jun 3, 2015 at 15:28
  • It depends what they did to move it. If they just changed an automount setting, and if your directory on the old server was still mounted on the client, and they left the NFS service on the old file server running and did not remove the old directory (e.g. they just renamed it to user.old), you could have continued to use and modify the old files. Commented Jun 4, 2015 at 8:34
  • To phrase it in the form of a question: can you find out the steps the IT team took to move your home directory? Commented Jun 6, 2015 at 18:43

2 Answers 2

2

Endless probabilities here. Stale file will last a long as there is a program using that file/directory. Depends on how you 'start' work every day, how network is organized.

One issue I see directly in your question is that 'killall -u user' wouldn't certainly kill all user processes. You should use (in bash) 'killall -9 -u user & disown' and check that shell you used really exit. There might be other users using your files (you should have used 'lsof' to figure that out).

5
  • But one thing I fail to understand. Whenever I lost network connection, all my processes freezed immediately (which I believe due to NFS). How can it be that for 2 days, there was no NFS server response and I can just work peacefully?
    – Bichoy
    Commented Jun 3, 2015 at 15:45
  • Also, if this happens, would I be able to use the home directory regularly without any problem/error message for 2 days?
    – Bichoy
    Commented Jun 3, 2015 at 15:50
  • If NFS server is down you can't do anything with nfs share exported from that nfs server.
    – gena2x
    Commented Jun 3, 2015 at 15:59
  • In this case processes freeze on any disk access.
    – gena2x
    Commented Jun 3, 2015 at 16:00
  • Exactly! Then if the home directory is moved to another server, and I have a stale file handle as they claim, this means my home directory was not remounted from the new location. How can all other processes (vnc server and sophisticated cad tools ) could access my files without any errors for two days? I am not convinced with their claim...
    – Bichoy
    Commented Jun 3, 2015 at 16:04
0

The automountd is a root process, so you could not have killed it.

Now if you continued to work, there is a good chance that the old NFS-server was still up and running for a while.

Ask your IT-team whether they can rsync newer files from the old server to the new server for your home-directory.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .