On Thu, 21 Oct 1999, Brian D. Haymore wrote:
Hi,
> > I'm sorry this is a bit off-topic, but large raid sets need big backups ;-)
Hehe, i'll take the oportunity also to ask a couple things ;)
> > My HP SureStore DAT24 stops after writing 12 GB of data to a DDS-3 tape. I
> > suspect there must be something wrong with the compression. To verify this I
> > need some additional information:
a DDS-3 tape stores right that, 12Gb of data. A different matter
is if one can achieve higher storage capacity by compressing it, either by
software or hardware... to reach 24Gb.
After testing all the backup tools I found, I use (and prefer)
dump/restore -i (Or amanda on more complex scenarios). At first, dealing
with tape capacity was a little confusing, but I ended up learning that
the best way to handle that is to tell dump something like this (for a
DDS-2 tape)
dump 0uBf 8192000 /dev/nst0
rather than specifying tape density/length, etc. dump is a little
dumb :) with those. This way, all you have to do is specify a larger
capacity that the one the tape has, and dump will detect correctly the end
of tape.
> > - What are the right dip switch settings of the DAT24?
>
> I've found that hardware and software compression equals bloated results.
> The software compression by arkeia wins over the hardware compression for
> our DLT library and our HP library, which is the same one you have.
I prefer hardware compression, though. When things go wrong, and
specially if you use some fancy backup software, chances are it's the
first thing is going to fail. So I rest better knowing I can go to
wherever, take a DDS tape unit and extract the data...
> > - Did 'mt datcomression 2' really switch the compression on??
I found that actually mt, from mt-st, doesn't do some things. I
ended up usinf mt-dds, a little but helpful utility that comes with
several other tools great to deal with DAT... you can find it at FM.
That way, you can check/set compression on your backup scripts
like:
/usr/local/bin/mt-dds -f /dev/nst0 comp-on
if [ ${?} -eq 0 ]; then
....
And also set buffering, tape partitioning, and such.
> > - What is the best block size for the dump command???
Dunno... experiment...
> > - Any other topics I have to look for????
Check to restore random files to test the backups are good? :)
Beware that some mt binaries of 'old' RH distros were broken, and
you could become mad isuing commands like 'mt asf X' and such...
Now my questions:
1.- I have a /dev/sdaX mounted /home, and a /dev/md0 mounted
/home/samba. If I try to dump /dev/md0 everything goes fine. But if I try
to dump /dev/sdaX, dump hangs indefinitely... Is this due to the raid
being mounted "inside"?
(happens with 2.2.7 and 2.2.12 kernels with latest raid patches applied)
2.- I dunno what happen with DDS tapes and ncr53c8xx
(using the great driver from Gerard Roudier) controllers, but they use to
have a somewhat odd interaction; either with HP or Seagate Units. Which is
the best setting for those? Asynchronous? sync at 10 Mb/sec? Disable
disconnection? I have found that one can for example dump data sync at 10,
but cannot restore it although is set to asynchronous and disabling
disconnections... Happens with every ncr53c8xx based controller I've found
(810, 815a, 875).
Any hints? What's the best setup for those?
Cheers,
*****---(*)---**********************************************---------->
Francisco J. Montilla System & Network administrator
[EMAIL PROTECTED] irc: pukka Seville Spain
INSFLUG (LiNUX) Coordinator: www.insflug.org - ftp.insflug.org
