Hi,
Your command line attempts to do a striped transfer from oliver to
poseidon.
In this case, oliver will not have any incoming data channel
connections. So,
the IP address reported in the error should really belong to one of
the poseidon
nodes, as those are the ones that will be listening for the data
channel connections.
It looks like there is a firewall on the machine with IP 208.100.85.21
that does not
allow a data channel connection to be formed.
Also, I recommend that you move to GT5 GridFTP as it has a number of
new features
and bug fixes when compared to GT4.0.8
Raj
On Feb 4, 2010, at 2:22 PM, Esma Y?ld?r?m wrote:
Hi,
My research involves development of models for optimal striping
levels. I need to configure a striped GridFTP server over the LONI
network clusters. However I only have user account and user
certificate. I managed to run gridftp servers over two nodes of two
clusters Oliver and Poseidon as follows:
[e...@oliver1 ~]$ /usr/local/packages/globus/globus-4.0.8/sbin/
globus-gridftp-server -p 6000 -dn -logfile /home/esma/Dn1.txt -log-
transfer /home/esma/Tr1.txt -ia none
[e...@oliver2 ~]$ /usr/local/packages/globus/globus-4.0.8/sbin/
globus-gridftp-server -p 7000 -dn -logfile /home/esma/Dn2.txt -log-
transfer /home/esma/Tr2.txt -ia none
[e...@oliver1 ~]$ /usr/local/packages/globus/globus-4.0.8/sbin/
globus-gridftp-server -p 5000 -r oliver1.loni.org:
6000,oliver2.loni.org:7000 -logfile /home/esma/Cn.txt -log-transfer /
home/esma/Trh.txt -ia none
I also have .gridmap file on my home directory and a copy of the
certificates directory in my .globus folder for user configuration
of the server.
The same process is completed over Poseidon cluster. Then I ran the
globus-url-copy from Poseidon as follows:
[e...@poseidon1 ~]$ time globus-url-copy -vb -stripe -p 4 -dbg -s
"`grid-cert-info -subject`" gsiftp://oliver1.loni.org:5000/work/esma/5G
gsiftp://poseidon1.loni.org:5000/work/esma/
After waiting a long time after the following debug log (enough to
transfer the file) :
debug: sending command:
RETR /work/esma/5G
I get the following error:
error: globus_ftp_client: the server responded with an error
500 500-Command failed. : callback failed.
500-globus_xio: Unable to connect to 208.100.85.21:50383
500-globus_xio: System error in connect: Connection refused
500-globus_xio: A system call failed: Connection refused
500-
500 End.
(The IP in the error belongs to oliver1)
However the file seems to be transferred although a couple of MBs
are missing and only one of the data nodes (first node) is able to
make a log of transfer. The other logs are empty on the other data
node. Tr1.txt:
DATE=20100204190901.611649 HOST=poseidon1.loni.org PROG=globus-
gridftp-server NL.EVNT=FTP_INFO START=20100204190537.544388
USER=esma FILE=/work/esma/5G BUFFER=0 BLOCK=262144 NBYTES=2684172517
VOLUME=/ STREAMS=4 STRIPES=2 DEST=[208.100.81.21] TYPE=STOR CODE=226
on the output logs I usually get the following error:
[11737] Thu Feb 4 13:09:01 2010 :: IPC ERROR:
globus_xio: An end of file occurred
I can not figure out what exactly the problems is. Why is the file
transferred but a very small portion is missing although I get an
error. Also Why only one of the data nodes is able to make a log
while the others can not? Also the time difference between the first
transfer and second round of transfer of file is huge. Is that
possible? I really appreciate if anyone could give me an idea.
Thanks
--
Esma Yildirim
Distributed Systems Laboratory, CS/CCT @ LSU
PetaShare Project (www.petashare.org)
Stork Project (www.storkproject.org)
