* Gerhard den Hollander <[EMAIL PROTECTED]> (Fri, Apr 20, 2001 at 09:54:04AM +0200)
> I've set amanda.conf to use all of the holding disk, minus 1Gb
> 
> holdingdisk hd2 {
>     comment "smaller holding disk"
>     directory "/eighteen1/amanda/lto"
>     use -1Gb
>     chunksize 250Mb
> }
> 
> However, I am getting diskfull errors on that disk when amanda is backing
> up .. 
> 
> Now I know there is another job running on that disk, that uses up quite
> some diskspace, but that job stops when the freediskspace is less than 500M
> and continues after that.
> 
> (i checked ;) )
> 
> However, amanda happily continues dumping until the disk is full, thena
> borts the dump, and retries later on.
> 
> Any clue ?

To be slightly more exact.
the sequence of events is:

amdump starts dumping   [disk space 3G ]
        [dump is estimated to be 2G]
amdump dumps some parts [disk space 2G ]
other program starts filling the disk, while amanda is dumping
        [diskspace is now 500M, other program waits till enough free
        diskspace]
amdump continues dumping
disk full ..


So it looks like amdump checks if there is enough space on the holding disk
before it starts dumping,
and then simply continues dumping whitout checking if the holding disk
becomes full, until it runs in a diskfull error and aborts the dump.

The right thing (imo) would be to check every time a new chunk is written,
and if the chunk to be written would fail the holding disk free space
criteria, it should abort the dump.

Thoughts ?

Currently listening to: Gone-JJJ live

        Gerhard,  <@jasongeo.com>   == The Acoustic Motorbiker ==       
-- 
   __O  Got no religion, don't need no friends
 =`\<,  Got all I want and I got no need to pretend
(=)/(=) Don't try to reach me, 'cause I'd tear up your mind
        I've seen the future and I leave it all behind

Reply via email to