> On 2006-03-09 18:14, Iulian Topliceanu wrote:
>>>> Paul Bijnens wrote:
>>> Do you have runtapes > 1 ?
>>> Or do you run amdump multiple times a day?
>>
>> Amanda uses sometime more than 1 tape per run. That's why vtape-7 was
>> again next. So even if I would have defined runtapes 1, Amanda would
>> have
>> used more.
>
> AFAI have seen Amanda never uses more (v)tapes than the number specified
> by "runtapes" (which defaults to 1).
>
> I think something else is going wrong here...
> Can you send the complete amanda.conf to me?
> Are you sure there is only one config running?

I did some thinking and the only explanation for amadmin to report wrong
information on vtape-7 is the lack holdingdisk space.

Paul, as you can see in my amanda.conf, no holdingdisk is defined. I
hadn't that much space for the vtapes, and considered it to be useless
since I'm using vtapes and they're as fast as the HDD.

Amanda is writing the information directly to tape, and when it comes to
the end of the tape, it sees that there is no more space left on the
device so the DLE can't be dumped on tape entirely but it has already
overwritten the information on vtape-7.

The reason why Amanda estimate didn't notice that the dump is too big to
fit on the tape is that the vtape is 30 GB big, and /data/data0/share1 is
33 GB big.

Is this a valid explanation?

I still have no clue about the missing files in the incremental dump. How
could that be possible? Is it because of network problems? (knowing that
Amanda is using an UDP stream)

The level 0 dump from a single machine (should be a cluster bun one
machine is almost always not working) ~300 GB.

Iulian Topliceanu