I am working on setting up Amanda to back up machines in the DMZ to an
Amanda host on the internal network with a large raid array rather than a
tape drive.

I am running Amanda 2.4.4 compiled from source.  My installation was
compiled with the --with-portrange=10085,10090
and --with-udpportrange=850,859

The dump of all volumes that use dump succeeds, both locally and on the
remote server (Amanda client), but the remote dump that uses tar fails while
a local tar dump succeeds.

The Router for the internal network is allowing all TCP and UDP traffic on
ports 10080-10090 and 850-859 through to the Amanda server (this router does
not filter outbound traffic).  For testing, the internal iptables firewall
on the Amanda server is off, and the iptables firewall on the DMZ host
(Amanda client) is set to allow all traffic on all ports from the Amanda
server.

Can somebody please give me some pointers on what can cause the connection
reset error?  I can't find anything like this via Google or on the Amanda
Faq-o-matic.

Thank you,

   - Brian Peterson

FAILURE AND STRANGE DUMP SUMMARY:
  ethos.brav hdb1 lev 0 FAILED [mesg read: Connection reset by peer]

FAILED AND STRANGE DUMP DETAILS:

/-- ethos.brav hdb1 lev 0 FAILED [mesg read: Connection reset by peer]
sendbackup: start [ethos.braverock.com:hdb1 level 0]
sendbackup: info BACKUP=/bin/tar
sendbackup: info RECOVER_CMD=/bin/tar -f... -
sendbackup: info end
? sed: couldn't write 76 items to {standard output}: Connection timed out
\--------

Monitoring the dump shows that it gets 55-85% of the way through the dump
before hanging and eventually failing.

Things likely to be relevant from amanda.conf:

tapetype HARD-DISK     # what kind of tape it is (see tapetypes below)
tapedev "file:/home/raid/backup/amanda-backup/"

holdingdisk hd1 {
    comment "main holding disk"
    directory "/home/raid/backup/amanda-backup/holding"
}

reserve 20 # percent

define tapetype HARD-DISK {
    comment "hard drive raid array tape simulator"
    length 15000 mbytes
}

define dumptype hard-disk-dump {
    global
    comment "Hard disk backup to RAID array"
    holdingdisk yes
    priority high
    index yes
}
#failed with holdingdisk set to no as well

define dumptype hard-disk-dump-tar {
    hard-disk-dump
    comment "Hard disk backup to RAID array using gnu tar"
    program "GNUTAR"
    compress server best
    exclude list "/etc/amanda/excludelist"
}
# of note here is that the dump failed with
# compress on the client as well

The disklist looks like this so far:
localhost hdb2 hard-disk-dump
localhost hdb1 hard-disk-dump-tar
ethos.braverock.com hdb1 hard-disk-dump-tar
ethos.braverock.com hda3 hard-disk-dump
ethos.braverock.com hda6 hard-disk-dump