how to tell amdump to forget dumps

hello,

I still have problems with certain large partition backups, amdumpsends this way:sendbackup: time 33685.307: 87: normal(|): DUMP: 60.89% done at 2453kB/s, finished in 5:59

sendbackup: time 33753.599: index tee cannot write [Broken pipe]

Although I tweak etimeout/dtimeout, firewalls setting, tcp_keepalives(http://wiki.zmanda.com/index.php/Results_missing)it still freezes on level 0 large dumps :-( . (even larges level 1 inthis example)

apart from a solution to that problem (which I'll appreciate above all;-) ) ,how can I tell a running amdump to forget theese "[Broken pipe]" DLEand pass to others "wait for dumping"

here are "frozen" dumps (since 4Hours although dtimeout is 2400 !?)
Using /var/lib/amanda/int/amdump from lun jun  2 15:12:49 CEST 2008
$ amstatus int --dumping

cobra3:/p1v2f1 0 9681m dumping 6916m( 71.44%) (15:27:54)cobra3:/p1v4f1 0 15496m dumping 7336m( 47.35%) (15:21:44)pasargades:/var/spool/imap2 1 7854m dumping 4263m( 54.28%) (15:25:58)pasargades:/var/spool/imap3 1 9394m dumping 5355m( 57.01%) (15:21:44)


and here is a sample of others DLE waiting for dumps
$ amstatus int --waitdumping
Using /var/lib/amanda/int/amdump from lun jun  2 15:12:49 CEST 2008
cobra3:/p2v5f1                             1       59m wait for dumping
cobra3:/p2v5f2                             1      144m wait for dumping
cobra3:/usr                                1       27m wait for dumping
colibri:/data1                             1      521m wait for dumping
colibri:/data2                             0      215m wait for dumping
...

I would really apreciate to let theese dumps start on that amdumpinstead of doing an amcleanup which will use an other tape (disk vtapein my case) and start again an other amdump .


thanks for any advice .