Solaris

We are not a huge shop -- 1 production V440 and one devel 450 box. Our backup strategy in a nutshell is if the system goes completely down, we reinstall the OS and then restore from tape the folders we need. I'm using Unix agent with BackupExec 10. It allows me to pick and choose down to a folder-level or even file-level backup.

I'm still learning this Unix admin stuff, so I'd appreciate some feedback. I'm attempting to identify which system folders on a Solaris 9 box, should be backed up regularly? I'm attempting to filter down my M-R backups to only do critical system files as well as my user-defined mount points. On Friday evenings, I would do a full system -- backup everything I can.

Out of the following system mount points, what really needs to be put to tape? And is it possible that only certain files or certain subfolders need to be backed up? I'd appreciate any feedback. Thanks.

DeadLetters/
devices/
export/
etc/
home/
kernel/
mnt/
net/
nsr/
opt/
platform/
proc/
sbin/
TT_DB/
usr/
var/
vol/
xfn/

See the device and files doc - note the backup/restore sections for a general overview.

As far as what to backup - everything. If you leave something out, that's when you'll need it after a disaster.

Each server/site is different - so there really isn't a list that will cover your site/servers. You have to know what's important. Search for disaster and/or recovery - these questions have been asked before
Example: disaster recovery

"On Friday evenings, I would do a full system -- backup everything I can" What does that mean? Is there stuff you can't backup? Why?

After you back up everthing on Friday, your Daily backups should be some form of incremental. One scheme is to backup everything that changed since Friday's backup. The other scheme is to backup everything that changed since the last backup of any kind.

We do use exclude lists...stuff to ignore. I assume that your backup system is smart enough to ignore /proc, /fd. (And it will stat special files rather than read them.) If not, you should explicitly exclude these. Anything NFS mounted? Exckude that. The files get backed up once on the box where they reside. Database files *.dbf or whatever. The database should have its own backup system. The .dbf files are huge and unless you shutdown Oracle, there is no point in backing them up. They change too fast. Some people exclude /tmp. We don't. It doesn't save enough to make it worth it.