>----- Original Message ----- 
>From: "Paul Bijnens" <paul.bijnens AT xplanation DOT com>
<snip>
> Good question. Never seen such behaviour before.
> In ~amanda/CONFIG/ there is a file named "amdump.1".
> Maybe that contains some information.
> Can you find out which DLE that dumper is dumping?   (the above file
> can help maybe.)
>
> Also notice that the last stable version is 2.4.4p4, while you're
> running 2.4.3-something.  Maybe upgrade?
>

Hello again.

It did it again this night, but my little perl proggy did catch the event
and killed it.
Here is the very last lines of the file amdump.1 :

    ...
    driver: hdisk-state time 25590.732 hdisk 0: free 2048000 dumpers 0
    driver: QUITTING time 25590.733 telling children to quit
    driver: send-cmd time 25590.733 ignored to down dumper dumper0: QUIT
    driver: send-cmd time 25590.733 to dumper1: QUIT
    driver: send-cmd time 25590.733 to dumper2: QUIT
    driver: send-cmd time 25590.733 to dumper3: QUIT
    driver: send-cmd time 25590.733 to taper: QUIT
    taper: DONE [idle wait: 7805.047 secs]
    taper: writing end marker. [DMP014 OK kb 3392160 fm 1]
    driver: dumper0 exited with signal 15
    driver: FINISHED time 25591.483
    amdump: end at Wed Nov 10 10:06:32 CET 2004
    <end of file>


I guess the signal 15 appears when dumper0 has been detected as a
memory-eater,
but at this point, all operations are finished, aren't they ?
What means "ignored to down dumper dumper0" ?

May be this ring somebody's bell ?

ps: at any rate, my problem is fixed - it's dirty, but it doesn't crash the
system
anymore, and the backups seem all right, which is the purpose of it, so
my questions above are just academic :)

Regards,

Jean Flinois <tech AT vtech DOT fr>
V-Technologies, Savennières