Re: Backing up

Tim Streater <timstreater@xxxxxxxxxxxx> wrote:

real-address-in-sig@xxxxxxxxxxxxxxx (Rowland McDonnell) wrote:

Tim Streater <tim.streater@xxxxxxxxxxxx> wrote:

They should have tested it to see if they could retrieve an arbitrary
file. But then, perhaps they did, your summary didn't mention that.

My information source - a bloke who works in MS's MBU - did not mention
that one way or another.

This
is just the point that was made by others, that any backup system is
reliant on a piece of software, you have to hope that the software is
quite reliable.

But I won't do that - I'm not just going to `run the software and pray'.
I've met far too much unreliable software in my time to be that stupid.
No sensible person would, if you ask me. You need verification.

Fine, but to what level?

<sigh> I don't bloody know! `As much as I can be arsed doing, such
that my checks keep my mind happy'. That's what I do, anyway.

I see two approaches to backup:

1) Use cp on files/folders/disks that matter and copy them to some other
medium. That's what I do at home. I don't bother to verify this.

2) Use some software/scripts to implement a strategy based on some
number of (say) tapes, some period of time, and some set of requirements
about for how long into the past you want to retrieve not just today's
files, but the state of the file at some previous time. This is what I
have implemented myself in the past. So I could give you any day during
the last fortnight and one back up 12, 11, ... months ago but no better
granularity than that. You use many tapes to guard against failure of
one tape. You restore an arbitrary file from time to time to ensure the
tape has something on it and you can read the catalogue (e.g. if for
some reason the backup writes files not with their real names (in which
case you better be able to read the catalogue - write it to another tape
etc etc).

This approach can be more or less complex depending what you want.

Uhuh.

I expect that, much like them testing for quality of naval shells during
the First World War (quantity of cordite in each), you better think
carefully about how you test the backup tapes.

My usual apprach to that sort of `issue' is multiple copies. Of course,
for that to be really useful, you'd have to make sure your backup tapes
were all from different manufacturing batches, wouldn't you?

[snip]

So? One can - if one has the storage available - just maintain
multi-gen copies of everything, using a script if you like. Then
verify. There's no need to have a fancy `diff-style' incremental
storage setup all the time, and if you *can* avoid such a setup, you've
got something more reliable.

It's why I'd quite like tape.

Tape is good but be prepared for more complexity.

Pfft. Depends how you do it - why not just dump everything to a fresh
tape each day? I knew a chap who maintained a useful multiple gen
backup that way (university lab) going back a year - every day this
week, then he pulled tapes out so by the time you got to a year back,
there was, oh, I think one tape a month by that time or something.

[snip]

I've no idea how you came to that conclusion.

Well, if you do some variant of (2) above, you have some software (other
than cp) involved. I thought you didn't like that.

<sigh> again. I've no idea where you're getting these ideas from. I do
some of my backups using Disk Utility to create compressed disc images.
I verify them, of course. Checksums are handy, as is copying everything
off it in cases where I'm really paranoid.

There's all sorts of silly things I do like that. But the idea is to
make sure I've got another copy, just in case, somewhere.

<yawn>. Anyway, back to the day job. I'll leave you to present us with
the world's most amazing, all-singing, all-dancing, proposal for a
backup system.

Keep on waiting. Any proposal I might come up with would involve a
multiple generation backup done the most reliable way I can think of
*and then verified*. It's what I've been doing for a long time.

I've been bitten in the past by a diff-style incremental backup system.
I'm not going to rely on software that produces a backup that I'm not
confident will be usable in the future on a different OS, basically.

So, in fact, we are probably in agreement.

Probably.

Nighty night.

Good morning.

Rowland.


--
Remove the animal for email address: rowland.mcdonnell@xxxxxxxxxxxxxxx
Sorry - the spam got to me
http://www.mag-uk.org http://www.bmf.co.uk
UK biker? Join MAG and the BMF and stop the Eurocrats banning biking
.

Relevant Pages

  • Re: NTBackup
    ... the good part of that is 'I use a cleaning tape', ... recalibration' is a one shot affair for DDS, ... Backup Status ... Operation: Verify After Backup ...
    (microsoft.public.windows.server.sbs)
  • Re: Whats the easiest way to do a backup and verify?
    ... >> Is there an easy way to combine a backup and verify when doing backups ... >> tape is readable. ... I'm not very familiar with tapes, but I think that the dump is written ...
    (freebsd-questions)
  • Re: VTL vs. Backup to Disk
    ... Disk is designated as DASD. ... emulated to provide a complete tape emulation solution. ... The question boils down to backup application support. ... so a good VTL may provide for a better solutions and better ...
    (comp.arch.storage)
  • RE: backup not starting
    ... "The requested media failed to mount" error message when you try to perform ... Backups on Windows Small Business Server 2003 may use more backup media ... If the issue persists, please manually clean the tape driver, and then ... by specifying the same media label name. ...
    (microsoft.public.windows.server.sbs)
  • RE: Backup failed event:
    ... 1.There is actually no tape in the tape device and it is not correctly ... Please checked the HCL list for DLT backup device from our HCL ... your backup media. ... This operation writes a "free media" label on each tape. ...
    (microsoft.public.windows.server.sbs)