Oooops ... I just see that I still have several unsummarized Requests.
Here we go.
> I had set up a cron Job for this Weekend to move all the /home/TI Stuff to
> a bigger Partition. Basically, the Job did this:
> - Mount big Part (sd5c) on /mnt
> - tar | tar over all Data
> - Umount /mnt and /home/TI
> - Mount big Part on /home/TI
> - Check for some Test File, if found then exit
> - Umount /home/TI
> - Mount original Part (sd2h) on /home/TI
> I read my Mail this Morning, and it said "/home/TI busy". A User stayed
> logged in over the Weekend. OK, I thought, so nothing changed. And all
> of a sudden, I see ...
> kleopatra:/home/TI/bern% mount | grep TI
> /dev/sd2h on /home/TI type 4.2 (rw,quota)
> /dev/sd5c on /home/TI type 4.2 (rw,noquota)
> ARGHH! I never would have thought that this could happen. I just imagine
> that my Users write to sd2h and then get sd5c's Info "no such File" ... <sweat>
> I'll go and find out which Computer accesses which Partition right now, but
> could someone enlighten me on how the %*@!!! this can happen and what the
> Effects {sh|c}ould theoretically be? BTW, this is a NIS+NFS Cluster under
> 4.1.2, and df reports both FSes with the Size of the original Partition.
^^^^^^^^ Typo; should read
NEW Partition
Most People suggested that this is considered legal Behaviour, and that the
last-mounted Partition should be the one in Use. I would have liked this,
but it was only true for the File Server. The NFS Clients happily continued
to access the old Partition (probably NFS Handles point to Inodes instead
of the explicit Path you initially specified). Thus, I was correct to fear
that my Users might produce several "latest Versions" of whatever File; In
Fact, one did. Repeat:
WARNING: NEVER REPLACE A FS MOUNTED VIA NFS WITH ANOTHER FS UNLESS YOU CAN
======== REBOOT OR OTHERWISE REINITIALIZE THE NFS CLIENTS, OR CAN RELY ON THE
FACT THAT YOUR USERS WILL ACTUALLY STUMBLE OVER NON-EXISTING FILES
OR DIRECTORIES!
I was lucky because there had been only one User having written to both FSes,
and only one other User (me) having written onto the new FS, so I had very
little Work sorting Things out.
Regards,
J. Bern
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:08:33 CDT