VanLUG Email Archive
Ted Powell (ted@eslvcr.fireplug.net)
Fri, 26 Feb 1999 16:08:15 -0800
On Fri, Feb 26, 1999 at 03:35:42PM -0800, Harondel J. Sibble wrote:
>
> On 24 Feb 99, at 20:44, Ted Powell wrote:
> > If you are backing up an online medium, then by the time you have rewound
> > your tape and started to compare it bit-for-bit with the original data,
> > the original data may well have changed. [...]
>
> true, but that's pretty much true for any os with maybe the exception of
> dos....
Well, I would phrase that as, "true, *AND* that's pretty much true
for any os; therefore I would not consider doing it under any normal
circumstances."
In other words, I don't agree that the fact that it's pretty much true
for any os weakens what I said in any way.
In the last backup system I spent serious time on, the data was
checksummed and byte-counted on its way to the tape. One tape file was
written for each file system that was being backed up. During the verify
phase, each tape file was read back and checksummed and byte-counted.
These values were compared against the saved values that were calculated
during the write phase (which of course would not have changed). The
backup system produced a report that for each tape file showed what
file system was backed up to it, whether it was 100% or incremental,
and gave the size and checksum.
It was only necessary to confirm in the morning that there were zero
error messages--well, not even that; when the report was emailed the
subject line indicated success or failure. I consider this to be far
less risky than:
> Plus as you do it a few times, you begin to notice what the norm is
> changewise and can take that into consideration, just like one peruses the
> security logs for out of ordinary events.
Personally, I find one error message versus zero a lot easier to detect
than 473 versus 472.
With a saved backup report containing checksums, one can remount the
tape at any later time and checksum it again, if one needs reassurance
that the tape is still valid. If you remount a tape and compare it with
disk again, and they differ, you don't know for sure which has changed.
I only used a simple 32-bit sum of the data, but it could have just as
well been a 32K-bit hash value, or anything that wasn't onerously large
to store.
-- ted@psg.com http://psg.com/~ted/ (Ted Powell) If your hard drive crashes, perhaps you have a recent backup. If Earth crashes, what then? We need off-site backup: Luna, L5, Mars, wherever.
This archive was generated by hypermail 2.0b3 on Fri 26 Feb 1999 - 16:11:59 PST