 |
 |
Hi Fabrizio, Ahhhh, then again, the protocol spec says that the server may return any sized chuck at any time. All it gaurantees is that you'll receive them in order and, eventually, all of the bytes you requested (if that many bytes exist). Andy ----- Original Message ----- From: "Fabrizio Furano" <[log in to unmask]> To: "Wilko Kroeger" <[log in to unmask]> Cc: "Peter Elmer" <[log in to unmask]>; <[log in to unmask]> Sent: Friday, January 21, 2005 6:30 AM Subject: Re: strange behaviour of xrdcp in 20050119-0021 > Hi, > > i found the problem about the 0-sized file. > The new server async i/o splits the data it sends into chunks of 64K, > smaller than the old value of 2M. The algorithm which got data from the > cache was not able to concatenate multiple blocks smaller than the read > ahead size. This was indicated in the remarks as a future improvement, > now it was urgent, it's there. > > Working under Linux, committed, I only need a little more time to test > it under solaris. > > Fabrizio > > Wilko Kroeger wrote: > > Hello Pete > > > > > > On Thu, 20 Jan 2005, Peter Elmer wrote: > > > > > >> Hi Wilko, > >> > >> Ok, thanks. Once I turn off the async I/O on the server the following > >>works: > >> > >>[csfd] ~/work > xrootd/bin/xrdcp root://noric01.slac.stanford.edu:2525//usr/work/elmer/deepcopy16.02E.root . > >> > >>as it should. I still have the problem where: > >> > >>[csfd] ~/work > xrootd/bin/xrdcp root://noric01.slac.stanford.edu:2525//usr/work/elmer/deepcopy16.02E.root bogus.root > >>Error Is a directory creating bogus.root > >> > >>creates a directory. (Clearly a separate problem.) > > > > > > If you run a second time it will work and the file is copied to > > bogus.root/bogus.root > > > > This behaviour is already present in previous versions. If you just > > specify a name xrdcp uses it as a directory. If the target name has a '/' > > it works fine, e.g.: > > > > xrdcp root://...////..deepcopy16.02E.root ./deepcopy16.02E.root > > > > I guess we should think about how the syntax should work. > > > > Cheers, > > Wilko > > > > > >> Pete > >> > >> > >>On Thu, Jan 20, 2005 at 10:59:06AM -0800, Wilko Kroeger wrote: > >> > >>>Hello Pete > >>> > >>>I just tried the same thing and I also see that the file is empty. > >>>The xrootd by default has now the asynchronous I/O turned on. You > >>>can turn it off using > >>>xrootd.async off > >>>in the config file (You have to create a config file and use the -c > >>>option with xrootd). That should fix your problem. > >>> > >>>There are still problems with the current version that we see on the > >>>bbrprod cluster and Andy is looking into this. > >>> > >>>Cheers, > >>> Wilko > >>> > >>> > >>> > >>> > >>> > >>> > >>>On Thu, 20 Jan 2005, Peter Elmer wrote: > >>> > >>> > >>>> Hi Fabrizio, > >>>> > >>>> I just tried to use xrdcp to copy a file from one place to another and > >>>>ran into a couple of problems. I'm using version xrootd 20050119-0021. The > >>>>client side is running RHEL3 and the server side is running SL3. > >>>> > >>>> On the server side I have: > >>>> > >>>>noric01> ls -l /usr/work/elmer/deepcopy16.02E.root > >>>>-rw-r--r-- 1 elmer ec 7975402 Jan 20 10:15 /usr/work/elmer/deepcopy16.02E.root > >>>> > >>>>and do: > >>>> > >>>>noric01> xrootd/bin/xrootd -p 2525 /usr/work/elmer/ > >>>> > >>>>On the client side I tried: > >>>> > >>>>[csfd] ~/work > xrootd/bin/xrdcp root://noric01.slac.stanford.edu:2525//usr/work/elmer/deepcopy16.02E.root bogus.root > >>>>Error Is a directory creating bogus.root > >>>> > >>>>The strange thing here is that it tried to create a directory: > >>>> > >>>>[csfd] ~/work > ls -ld bogus.root > >>>>drwxr-xr-x 2 elmer bfactory 96 Jan 20 18:40 bogus.root > >>>> > >>>>I then tried to do: > >>>> > >>>>[csfd] ~/work > xrootd/bin/xrdcp root://noric01.slac.stanford.edu:2525//usr/work/elmer/deepcopy16.02E.root . > >>>>[csfd] ~/work > > >>>> > >>>>but the result was an empty file: > >>>> > >>>>[csfd] ~/work > ls -ld deepcopy16.02E.root > >>>>-rw-r--r-- 1 elmer bfactory 0 Jan 20 18:40 deepcopy16.02E.root > >>>> > >>>> Any idea what is going on here? > >>>> > >>>> thanks, > >>>> Pete > >>>> > >>>>------------------------------------------------------------------------ - > >>>>Peter Elmer E-mail: [log in to unmask] Phone: +41 (22) 767-4644 > >>>>Address: CERN Division PPE, Bat. 32 2C-14, CH-1211 Geneva 23, Switzerland > >>>>------------------------------------------------------------------------ - > >>>> > >> > >> > >> > >>------------------------------------------------------------------------- > >>Peter Elmer E-mail: [log in to unmask] Phone: +41 (22) 767-4644 > >>Address: CERN Division PPE, Bat. 32 2C-14, CH-1211 Geneva 23, Switzerland > >>------------------------------------------------------------------------- > >> >