This one time, at band camp, Jon LaBadie wrote:
>As I noted, I was uncertain of the details, someone understanding more
>about the networking aspects can comment more.  But just for clarification
>of the way I worded that:
>
>- I did not say anything about the backup data stream over udp
>- Your logs showed it was the index backup stream having problem,
>  not the backup stream (they are separate)
>- I said the timeout setting "affected" the stream, I wasn't sure
>  if it was the stream itself or not
>- amanda does use some udp connections, perhaps for control communication

Thanks.

The only things I can find in google relating to the index stream are
conntrack-related timeouts; there is no explicit firewall running on the
server or client on their backup network interfaces (amanda is running over
a private segment) let alone NAT, so the amanda conntrack module isn't being
used.

Another data point though; this backup today started at 1:30am (it's now
10:15am) and is only 35% through -- about 20GiB.  This is pretty abnormal
(well, compared back to when it used to run -- it would on a bad day be
finishing about now).