Hi all,
we're currently running a file based backup of our AFS cell using ADSM. In the
future we will switch over to a volume based approach ("vos dump", etc.), but
for the moment we're forced to do it file based. This is because the volume
based approach would require a full backup of all volumes approx. every one or
two weeks (for restore considerations), which is almost impossible for us to do,
as we're restricted by a few environmental limitations (ADSM server performance,
network bandwith, etc.). Thus the whole cell cannot be backed up in one day.
Our current approach looks like this: A script descends the directory tree,
checks whether a directory entry is an AFS mountpoint and, if so, hands this
entry over to ADSM. ADSM then checks all of these subdirectory trees for files
which have changed since the last backup and backs them up if necessary. The
main reason for doing it this way is that we can make sure that portions of max.
2 GB (the max. volume size) get scanned for files to be backed up (or will be
backed up completely if a new volume) and we can do that a reasonable time (see
performance considerations above). Also true in case of restore.
This approach has one severe shortcoming (as you can already imagine ... ;-) ):
AFS provides no command that provides the opposite functionality of "fs
lsmount", i.e. a command like e.g. "fs lsvolmount", which lists the
mountpoint(s), where a volume is mounted. So our script that scans the directory
for mount points has to make a "fs lsmount" call for each and every subdirectory
- and this takes long!
Now my question is: Has anybody similar problems (and maybe created a reasonable
solution or workaround for this) or would a command like that one described
above be of any use to others? Let me know what you think about it; I will also
contact Transarc and ask them if such a command would be possible to implement.
TIA for your answers and have a nice weekend!
Markus
---
Markus Alt
IBM Laboratory Boeblingen, Germany
