hi again, the os is sl303, kernel 2.4.21.-20.ELsmp. i've read some info given in the links, and will look into the firewall issue more closely, but since i don't have direct access to that firewall, it could take a while. btw, how long should it take to complete the loop (100 times globus-url-copy)? can somebody benchmark this, here it takes a while. stijn > On Wed, Mar 02, 2005 at 10:16:25AM +0000 or thereabouts, Stijn De Weirdt wrote: > > hi all, > > > > i have a problem with globus-url-copy when copying to our resource broker > > (part of a belgian test-grid). the globus-url-copy sometimes hangs on the > > client side (for example when fetching job output, so it's very annoying). > > So if you do something like > > for I in `seq 1 100` ; > do > echo $I > globus-url-copy gsiftp://rb.be/etc/group - > /dev/null > done > > it hangs after a few transfers. > > Try unsetting the GLOBUS_TCP_PORT_RANGE at the client end and > see if that helps. And those ports are not open inbound. > > There is also > http://goc.grid.sinica.edu.tw/gocwiki/gridftp_works_only_once_within_a_minute_or_so > > which may give you some ideas. Of course this could be a completly > different problem. > > Steve > > > > > from the logs on the resource broker i see this message a lot > > Feb 27 05:11:14 gridy7 kernel: application bug: edg-wl-in.ftpd(2313) has > > SIGCHLD set to SIG_IGN but calls wait(). > > > Feb 27 05:11:14 gridy7 kernel: (see the NOTES section of 'man 2 wait'). > > Workaround activated. > This is new feature of RHEL3 kernels trying to be helpful and talking > rubbish. Ignore it as it happens everywhere. > > > > in a post on 15 september 2004 there was a mention of this problem (i > > think), but the actual solution was not posted. if you need more info, let > > me know > > > > stijn > > -- > Steve Traylen > [log in to unmask] > http://www.gridpp.ac.uk/ >