In article <Q86dnS4Ukr8BMsrZnZ2dnUVZ_vSdnZ2d (AT) giganews (DOT) com>,
no$pamwebmaster@no$pamforallspec.com (Glen B) wrote:


Under normal circumstances I would probably do so, but as I am intending
to retire in July, it hardly seems worth it.

Alan Pritchard

Please reply to: alan.pritchard (AT) gmail (DOT) com

Thanks for all the helpful comments over my problem. Although I can find
no trace of problems in logs, I am going to replace both the disks and the
SCSI controller, as some suggestions had been made that this might have
caused the problem.

Also (in spite of what I said to Glen), upgrade Linux & d3.

Thanks again, especially to Tom deLombarde who emailed me, with extremely
helpful pointers. Especially the following:

1. Make ***SURE*** that you are getting good D3 backups.
1a. Make ***SURE*** that you are getting good D3 backups

and more importantly

Make ***SURE*** that you are getting good D3 backups

!!!

For once I was doing this.

Alan Pritchard

Please reply to: alan.pritchard (AT) gmail (DOT) com

Corollary: Once in a while, check your backups.

Shell Oil of Canada had an IBM/RT and OA and did filesaves religiously every
night. They never verified that any of them worked. A broken wire that
went un-noticed for 6 months and they had a near-fatal disaster.

"Save and Save Often"

Mark Brown


"Alan Pritchard" <alan.pritchard (AT) gmail (DOT) com> wrote

<ramblin>
Not that anyone cares but at RD I spent a lot of time to get a small
change in upgrade documentation from "do a file-save" to "get at least
two complete and verified backups". Based on experience the
distinctions were important to me. A generic file-save doesn't
include accounts that are temporarily DX'd - get complete saves.
Saying "get at least two" is a message that one backup simply doesn't
provide enough ass coverage. Emphasizing "verified" is a clue that
just doing a backup isn't enough. It took too long to get that one
sentence changed amidst claims that just telling someone to backup was
enough, so I wasn't going to push for more text about exactly how to
verify a backup or explain why a site should go through the
time/expense of getting two. I just checked and I'm happy to see
similar text is still in some install/upgrade docs.

I think at some point I even got in a recommendation that the same
disk not be used for upgrades (AP/Pro, D3/Linux, D3/ProPlus, etc) but
that note was removed later. I'm sure the recommendation would be
appreciated by anyone who has gone through the broken wire syndrome
that Mark mentions, but it's lost on people who believe documentation
should be as brief as possible. From a Support perspective it's a lot
easier to tell someone to put their old hard drive back in than it is
to say "sorry your tape drive doesn't work, and by the way, you just
blew away your hard drive which was the only source for all of your
data". Yeah, been there, done that. Poor bastards.
</>

T


"Mark Brown" wrote:

On Wed, 03 May 2006 21:27:29 +0000, Mark Brown wrote:

And actually test restore from them once in a while. I knew a company
that also did backups religiously. What they didn't know was their
biggest file, for some reason, had a "1" for the resize modulo. It took 3
days to crawl thru that file when replacing a broke drive! (Also on an RT)

Art

Not to mention the tape drive with a broken write head... It whizzed
through the backup curiously without a fault, and then verified it back
without a problem... Of course the backup on the tape was 6 months old when
the write head was last working!

We verify the backup by checking the time stamp on the header and also by
reading back through the whole file list - comparing the filelist to backup
with the filelist on tape to make sure they are identical. (This is on
jBASE not D3 but I guess the same can be done on D3).

Regards
Simon

--
================================
Simon Verona
Dealer Management Service Ltd
Stewart House
Centurion Business Park
Julian Way
Sheffield
S9 1GD

Tel: 0870 080 2300
Fax: 0870 735 0011

"Art" <artmartz (AT) triad (DOT) rr.com> wrote

Hi Art,

Whoa!!! And no one suggested breaking out and restarting the restore
with then (N option?

This is yet another reason why I keep emphasizing that the D3 dm,bp
source code is merely a 'suggestion', rather than a 'one size fits
all' set of sysadmin resources. This includes file-save, which a good
base, but some value-add resellers might prefer to add more rigorous
verification code and other features on behalf of their clients.

"Simon Verona" <nomail (AT) nomail (DOT) zzz> wrote:

On Thu, 04 May 2006 07:54:27 -0700, Tom deL wrote:

This goes back a ways, I'm not sure if OA even had the "N" option then.

Art