--On Saturday, November 08, 2003 12:43:12 -0800 Dana Bourgeois <em-lists AT 
netgods DOT us> wrote:

> I am building a new Amanda server and have to use GNUTAR to split up some
> very large DLEs.  The result is only 128G but 62 DLEs.  This is not my
> question.
> 
> I am running all of these for the first time.  Amanda is scheduling 35G (one
> tape) but only about 3G is actually getting written for one reason or
> another (some is compression, some is weird tar errors).  At this rate it
> will take weeks to have this running.  What's the fastest way to get this
> going?

First, you need to fix whatever it is that causes your backups to stop
after 3 GB when you have a 35GB tape.  Once you solve that Amanda will
sort out the rest of it for you relatively quickly (although with  four
tapes you won't have much room for incrementals [4*35=140 - 128 = 12GB,
or about 10% of your data can change each week, but that might be OK
for your case.

>    Should I run amdump a couple of times per day adding clients each
> run?  

No, planner gets confused when run more than once a day.

> Perhaps run a different set of clients each night?  That would take 4
> nights since my cycle is 4 tapes.  Problem is the cycle is also one week so
> it will take several weeks to get this going.  Should I tell Amanda it has 3
> or 4 tapes to write to at first so it will schedule more stuff?

It sounds like Amanda is already scheduling 35GB worth but it is failing
for some reason.
> 
> Half of the clients are on a slow hub system with the rest on a fast switch.
> So I have potential timeout problems as well.

Fiddling with spindle numbers in the disklist might help here, but I
wouldn't worry about it until after you get everything up and running.
You can also increase the timeouts in amanda.conf, but is not generally
needed.  If you're not seeing timeout errors don't worry about it.

> What is the general rule of thumb in getting a relatively large disklist
> going quickly? Planner is not queuing up enough DLEs and I'm getting only
> about 10% of the data I should in one run.  

Some people add a few DLEs per day (useful if you want to make sure
certain ones get backed up first), others just toss in the whole batch
and ignore the 'no space' errors while planner sorts things out for
a few days.

> At this rate, it'll take a
> months to get everything right.  Since I'm using ftape, I'm going to
> increase the size of my tapes so planner will get more scheduled up.  I
> figure later, I'll drop the tape size down when planner has better history.

At the start of your email you said Amanda scheduled 35GB but only
wrote 3GB, so increasing your tape size won't help, only confuse the
planner and increase the chance of it trying to write something too
big at the end of your tape and wasting that space.

Post your entire daily report (change the hostnames if you want) and
this group can give you more detailed help.  We might then want you
you to look at and/or post some of your debug files and config files
depending on the errors and warnings in your daily report.  Your first
problem to solve is why only 3GB is being written and the daily report
will give some clues.

Frank

> 
> 
> Dana Bourgeois
> 

-- 
Frank Smith                                      fsmith AT hoovers DOT com
Systems Administrator                           Voice: 512-374-4673
Hoover's Online                                   Fax: 512-374-4501