On Sat, Feb 13, 2010 at 6:40 PM, Jeffrey J. Kosowsky
<backuppc AT kosowsky DOT org> wrote:
John Hudak wrote at about 12:06:29 -0500 on Saturday, February 13, 2010:
> Hello:
> I am considering using an external USB drive as the storage for my backups.
> I am running backup pc under Debian 5.0.
External USB drives are a *BAD* idea for multiple reasons:
- Slow
- Unreliable
- Subject to being disconnected
etc.
Yes, I know 1, and 2 is 'it depends', and 3 is exactly the reason why I want to use usb drives
> Part 1
> What do I need to do to configure the USB disk as the target? (e.g. how do I
> do it?)
> The USB disk is currently formatted as a NTFS file system. Do I *need* to
> reformat it to ext3? or other?
- NTFS is not usually used - need to check whether it supports the
types of hard links required for BackupPC
>
> Part 2
> Assume I am crazy paranoid about preserving backup data and I get a second
> USB drive to serve as a backup to the first USB drive.
> Also assume that I am not concerned about the bandwidth across the network
> or the various buses.
>
> >From a data reliability standpoint, is it better to run a backup session to
> USB drive 1, and then repeat the backup to USB drive 2? OR
> run a backup session to USB drive 1, and then copy the backup directories to
> USB drive 2???
Look at the archives and FAQ - this has been discussed *many* times so
no point in wasting peoples time in rehashing.
I did a quick search of the archives before asking - I did not find a definitive answer...
> The first approach could have errors in different backed up files on disk 1
> or 2 but given the odds, very unlikely that the same exact error would show
> up
> in the same exact way in the same file across both USB disks.
> OTOH, the second approach would allow the exact error in the backup on USB
> disk 1 to be copied to USB disk 2.
>
> I am leaning towards repeating the backup on two drives.
>
> My understanding is that files that are backed up (using either rsync or
> smb) are 'encrypted' (for lack of a better word), and to view them I need to
> use zcat.-True?
There is a better word -- *compressed*
So that is the word that is not clearly used in the documentation. There are many ways that backups can be manipulated: stored in a completely nonstandard/proprietary file system and protocol such as z-san, they can be encrypted, and they can be compressed. The backup PC doc talks about using compression, but does not state if any compression is used in the default configuration. Compression is often configuration parameter. It does not make sense to compress many audio and video formats. If the data to be backed up consists predominantly of these types of files, then it makes no sense to waste CPU cycles applying compression to get < 5% compression.
> Also, can the backup profile be specified to perform complete data copies
> periodically, as opposed to a baseline and then periodic incrementals?
Read the documentation and FAQ.
I have read the doc, (where IMHO) this should have been clearly stated. Instead the doc frequently introduces a topic with 1-2 sentences, then goes off on a tangent for 1-3 paragraphs about how things were done in a previous verson (completely irrevelant), or talks about what will be comming (again, irrelevant), or points one to another section of the doc, in the middle of some other thread that is related to the topic but does not address the topic at hand, or, introduces a topic, then talks about 3-4 other ways to accomplish the same thing, without telling the reader exactly how to do the initial topic to begin with. So my fault...I need to also read FAQs.
> Lastly, does anyone have a statistical number that represents the
> probability of a backup file (e.g. on the target backup disk) containing an
> error introduced
> by the backup procedure? I know there are error probabilities for both disk
> and tape reads/writes failures, but am wondering if anything like that
> exists for the backup software. (A group I used to work with did this sort
> of testing, and actually had some statistics on the reliability of backup
> programs, wrt types of files, sizes, w/wo compression, and the types of
> compression. Not sure the open source community would go through this type
> of assessment - but thought I'd ask.
The probability is either 0 if no bugs in the software (or your
configuration of it) or 100% if bugs in the software and your dataset
triggers the bug. Your question is not very well-framed and pretty
meaningless. I suggest you learn a bit more about backup in general
and backuppc in particular. There is a lot of good documentation on
BackupPC in the Wikki and in the archives, I suggest you reference it...
Well, in the extremely simplistic and ideal case, it is 0 or 1. In the real world, where algorithms are badly designed, or implemented, or both, then problems arise due to things not originally considered in the development which cause errors to arise in the backup procedure that is not about the conversion of 0/1's in computer memory or the transferring of them to tape, disk, etc.
The question to you may be meaningless, but it is not to me, and it wasn't to an R&D group that I worked in to look at this precise issue.
Again, the documentation (e.g. manual) for backup PC, leaves a lot to be desired. While I was reading through the issues in the forum, it struck me the number of times I came across the phrase "I followed the directions and it did not work/having trouble," etc. I guess I have to spend a lot of time searching archives to find this info, and then figure out if it applies to this version, or a previous version.
-J
