Veritas snapshot + rsync vs. rsync alone

I may be doing overkill in my current backup scheme. I am making
Veritas snapshots of my "active" volumes - not real busy, maybe a dozen
files written to every 10 minutes or so - and then doing rsync to make
remote copies of these volumes.

Do you think this is being too cautious? I ran an experiment of using
rsync on a text file that was currently being written to, and the only
bad result was the last line was left in a partial state. Is this
about what I can expect if I skip the snapshot step? The only type
files that would be changing during backups would be text files.

The snapshots are fast, but the "SnapBack" steps take about 40 minutes
to complete, so I would like to eliminate that step if it safe to do
so.

Boyd

I did a test yesterday in which I used a perl script to write to 10
files - five text files, five "binary" files (random numbers with no
line-feeds - I did put in 'EOL' text at the end of the print statements
so I could see where the break-off points were. I wrote random amount
of data to all 10 in a loop, repeatedly.

Then, while the above script was running, I did a SnapShot of the
volume (a mirrored volume under Veritas VM on a Solaris 9 Ultra II +
A5100 StorEdge), and then did rsync to the remote host. I then
examined the 10 files, and saw that all of the files had a nice cut off
at the end of the line (or "EOL" on the "binary" files) on the last
line of the file.

Next, I repeated the process, skipping the SnapShot step. Instead, I
just did rsync -a on the active volume to the remote host. When I
examined the 10 files, again they were all cleanly terminated at the
end of a line (or "EOL" on the "binary" files).

I was happily surprised at this result.
Boyd