View Full Version : clients fail with "Transport already connected" to 2.5.2p1 server

October 23rd, 2007, 03:10 PM
I'm trying to upgrade RHEL 4.3 tape servers from 2.5.0p2 to 2.5.2p2. I have a mix of Linux and Solaris clients running 2.4.2, 2.4.4 and 2.5.0. The existing clients have been configured to use TCP ports 50000-50100. After upgrading the server to 2.5.2p1 about half of the 100 DLE's fail in sendbackup. So far I haven't been able to find any pattern of amanda client version, OS or network location to explain which ones fail, but it's pretty much the same set every time. One system consistently fails when backing up one DLE but succeeds with another. These were all working when the tape server was running 2.5.0p2. I changed amanda.conf to add "auth=bsd" to global, and added

reserved-udp-port 512,520
unreserved-tcp-port 50000,50100

For the failed DLE's I get
FAILED [too many dumper retry: "[request failed: error sending REQ: send REQ to xx.xx.com failed: Transport endpoint is already connected]"]
/ lev 0 FAILED [cannot read header: got 0 instead of 32768]

October 23rd, 2007, 10:57 PM
Please post the amanda xinetd files on server and clients.


October 24th, 2007, 08:57 AM
service amanda
socket_type = dgram
protocol = udp
wait = yes
user = dat
group = disk
server = /usr/lib/amanda/amandad
disable = no

I'm not sure why this is relevant - there's no problem with starting amandad; this is a dumper issue.

November 6th, 2007, 11:58 AM
Check the configure options


Of if you don't used them, then have a look on


Maybe you need to increase them.