Amanda-Users

Re: vtapes and runtapes > 1

2005-12-30 06:58:03
Subject: Re: vtapes and runtapes > 1
From: Gene Heskett <gene.heskett AT verizon DOT net>
To: amanda-users AT amanda DOT org
Date: Fri, 30 Dec 2005 06:40:40 -0500
On Friday 30 December 2005 04:12, Geert Uytterhoeven wrote:
>On Thu, 29 Dec 2005, Gene Heskett wrote:
>> On Thursday 29 December 2005 05:03, Geert Uytterhoeven wrote:
>> >On Wed, 28 Dec 2005, Gene Heskett wrote:
>> >> On Wednesday 28 December 2005 11:25, Geert Uytterhoeven wrote:
>> >> >Is anyone else using vtapes with runtapes > 1?
>> >> >
>> >> >Recently I decreased tapelength and increased runtapes from 1
>> >> > to 2. On most days 1 vtape is sufficient. But every time
>> >> > Amanda hits (artificial as specified by tapetype) end of tape
>> >> > on a vtape and retries on the second tape,
>> >> >
>> >> >she fails with:
>> >> >|*** THE DUMPS DID NOT FINISH PROPERLY!
>> >> >|
>> >> >|These dumps were to tapes DAILY14, DAILY15.
>> >> >|The next 2 tapes Amanda expects to used are: DAILY16, DAILY17.
>> >> >|
>> >> >|FAILURE AND STRANGE DUMP SUMMARY:
>> >> >|  anakin     /usr/share lev 3 FAILED ["data write: Connection
>> >> >| reset by peer"]
>> >>
>> >> ISTR I had a similar problem Geert, but that was because I'd
>> >> somehow disabled my holding disk when I converted to vtapes.
>> >> Fixing the
>> >
>> >I'm indeed not using a holding disk.
>>
>> That begs the question then Geert, are you going to have one
>> tonight?
>
>No ;-)
>
>> You see, the reason it failed after it brought in the next 'tape'
>> is that once it had done that, it then discovered that since there
>> was no holdng disk image to work from, it had no copy of the src
>> tarball to retry a save of.  Since amanda isn't wired to repeat the
>> whole disklist entry in the event of a failure, it has no choice
>> but to send you a fail message.
>
>IC...
>
>> Some may say that this is a shortcoming of amanda, but I'm inclined
>> to agree with the amanda design in this respect, as it can go on
>> and finish the rest of the DLE list on the second (or more) tape
>> its allowed to use by the runtapes setting.  If the DLE was truely
>> too big for the medium, then it would sit there and use new medium
>> each time until it had run out of runtapes, or usable tapes.  So if
>> its a fail, there really isn't any use of a retry if the holding
>> disk image is missing.
>
>Unfortunately she never continued with the remaining DLEs, they are
> just marked as `RESULTS MISSING' in the report.

Is this on the same machine, or on another client machine?  In the 
latter case, the connection may have timed out during the failure 
recovery.

If the former, then I don't have a good idea what happened, and the 
logfiles for that session should be perused for the reason.  Its in 
there someplace.  Its even possible the connection timed out there 
too, since the server backs itself up using the client/server 
connection just as if the client, in this case itself, was another 
seperate machine.

What are the dtimeout and etimeouts set for in the amanda.conf?  The 
defaults there tend to be somewhat shorter than is sometimes needed.  
This is a moderately fast x86 box, with two clients, itself and my 
firewall box.  But it needs more than the default 300 seconds for some 
things:
--------
# grep timeout /usr/local/etc/amanda/Daily/amanda.conf
etimeout 900         # number of seconds per filesystem for estimates.
dtimeout 1800        # number of idle seconds before a dump is aborted.
ctimeout 8           # maximum number of seconds that amcheck waits
-------
I might add that the last ctimeout setting appears to have no effect, 
depending on the phase of the asteroid eros, it takes either less than 
200 milliseconds per client, or about 10.5 seconds to do an amcheck.

The firewall box is a bit slower, its only a 500mhz k6-iii with half a 
gig of ram, so thats the one that lollygags around and needs the 
longer timeouts.  Its doing its own gzipping too.                               
                      

>Hence runtapes = 2 behaves exactly the same as runtapes = 1, exact
> that if it hits end-of-tape on the first one, it wastes the second
> tape (by just storing the end-of-tape marker there).
>
>Gr{oetje,eeting}s,
>
>      Geert
>
>--
>Geert Uytterhoeven -- There's lots of Linux beyond ia32 --
> geert AT linux-m68k DOT org
>
>In personal conversations with technical people, I call myself a
> hacker. But when I'm talking to journalists I just say "programmer"
> or something like that. -- Linus Torvalds

-- 
Cheers, Gene
People having trouble with vz bouncing email to me should add the word
'online' between the 'verizon', and the dot which bypasses vz's
stupid bounce rules.  I do use spamassassin too. :-)
Yahoo.com and AOL/TW attorneys please note, additions to the above
message by Gene Heskett are:
Copyright 2005 by Maurice Eugene Heskett, all rights reserved.

<Prev in Thread] Current Thread [Next in Thread>