Subject: Re: SCSI problems or DUMP ?
To: Brett Lymn <blymn@baea.com.au>
From: Sean Witham <Sean.Witham@asa.co.uk>
List: netbsd-help
Date: 07/09/1997 15:19:12
On Wed, 9 Jul 1997, Brett Lymn wrote:

> 
> Are these dumps done on mounted file systems?  Are these file systems
> very active during the time of the backup?  The problem with dump is

No-one is using the files at this time, unless they set something to
run over night VERY rare.

> that it takes a snapshot of the file system at the inode level, it
> works out what to backup first and then proceeds with the write to
> tape.  If your file system is changing at the time then there is some
> chance that you will screw up the dump.  If the file system is
> reasonably quiet then you have a good chance of getting valid backups,
> if the file system is being thrashed then the backup will not be good.
>

true but surely that leads to curropted files not checksum errors such
that you cant even retreive any of the directory structure level data
from the dump with the initial "restore vi". I did a quick check with
tar and similar problems could be produced. So I now want to verify
the integrety of the SCSI bus and the termination. I need some advice
on how to detect, messaure, reproduce and even delibrately cause
termination problems etc so I can confirm what is happening with some
control cases as well as active tests. The troible is I don't know how
to go about this and wondered if there were any tools to test the scsi
bus with NetBSD. I bit like tcpdump but for SCSI activity instead of
network activity *8-)

Any help or advice on varioous hardware checks and software checks
would be of use. I'm goign to get some passive and active through
terminators as try them instead of the drive termination. Thing seem
to work when I turn the termiantion off on segate drive. The dat
drive has no termination ability so I can't use that instead.

--Sean

 
> >Any advice on how to pin down this fault would be helpful. I have the
> >suspecion that the SCSI bus is not properly terminated 
> 
> I would have expected reams of scsi errors in that case, just
> corrupting data is an unlikely event.
> 
> >or there is a
> >bug in dump .
> >
> 
> Not very likely, dump has been around a long long time.
> 
> -- 
> Brett Lymn, Computer Systems Administrator, British Aerospace Australia
> ===============================================================================
>   What do you get when you cross a cantaloup with a dog?        Melancholy :-P
> 
>