Print

---

Hi Wilko and Derek,

Yes. I will be changing the olbd to infer the filesystems should there be
no cache directive. Clearly, this is a confusing part of the configuration
and the proposed change should simplify things. I'll probably print a
warning message that I'm infering the filesystems.

Andy

On Mon, 4 Apr 2005, Wilko Kroeger wrote:

>
> Hello Derek
>
>
> On Mon, 4 Apr 2005, Derek Feichtinger wrote:
>
> > Hi,
> >
> > thanks for all the feedback, but since I was away last week I was only able to
> > do another test today.
> >
> > I'm getting some unexpected behavior:
> >
> > I set up the system (config files are below) taking suggestions from your
> > example setups.
> >
> > 1. I try to write a new file NEWFILE via the redirector => fails
> >
> >   $> xrdcp  fileA  root://lxn5219:1094//tmp/xrootd/newfile
> >
> >   Error accessing path/file for root://lxn5219:1094//tmp/xrootd/newfile
>
> I think the problem is that the olbd doesn't know which path to use for
> the new file.
>
> The olbd needs the oss.cache (or equivalent the olb.cache) directive to
> place the file.
> I must admit that I don't exactly understand the whole logic but here are
> the snippet from the data server configuration file that is used for
> writing via the redirector.
>
> #  directory /u1 (or different name) must exists but might not be used
>    /tmp doesn't work)
> oss.cache /u1
>
> #   'inplace' means that the file is not put into the cache
> oss.path  / inplace
>
> xrootd.export /tmp
>
> You can find the test configuration files in:
> http://www.slac.stanford.edu/~wilko/kan/cfg/xrootd-ddev_1rdr.cf   (redirector)
> http://www.slac.stanford.edu/~wilko/kan/cfg/xrootd-ddev_1.cf  (data server)
>
> We will discuss this problem and make the configuration a little bit easier.
>
> >
> > 2. I try to write the same file NEWFILE to a leaf node => succeeds
> >
> >   $> xrdcp  fileA  root://lxn5207:1094//tmp/xrootd/newfile
> >
> > 3.  I try to overwrite NEWFILE with a smaller file via the redirector
> >    =>now it succeeds, but the resulting file is the old file, where the start
> > has been overwritten by the new smaller file.
> >
> >   $> xrdcp  ../fileBshort  root://lxn5219:1094//tmp/xrootd/newfile
>
> The overwrite is a known problem and will be fixed. A wrong options
> (kxr_force) is used for the open request.
>
> Cheers,
>   Wilko
>
> >
> >
> > I'm using the new production version 20050328-0656 with xrdcp as a client.
> > Reading files via the redirector is never a problem, but writing of new files
> > still does not work for me.
> >
> > Thanks for any help,
> > Derek
> >
> >
> > These are the Config files:
> > ==============================================================
> > #
> > # dataserver.cf
> > #
> > # xrootd
> > xrootd.fslib /opt/xrootd/xrootd-20050328-0656/lib/arch/libXrdOfs.so
> > xrootd.export /tmp/xrootd
> >
> > ofs.redirect target
> >
> > odc.manager  lxn5219
> >
> > # olbd
> > olb.port 3121
> > olb.path w /tmp/xrootd
> > olb.subscribe lxn5219 3121
> >
> > oss.nostage
> > oss.nocheck
> > oss.nodread
> > oss.nomig
> > oss.alloc * * 80
> > oss.fdlimit * max
> > oss.path /tmp/xrootd
> > ================================================================
> > #
> > # redirector.cf
> > #
> > # xrootd
> > xrootd.fslib /opt/xrootd/xrootd-20050328-0656/lib/arch/libXrdOfs.so
> > xrootd.export /tmp/xrootd
> >
> > odc.manager lxn5219 3121
> > #odc.trace redirect
> > odc.trace all debug
> >
> > ofs.trace all debug
> > ofs.redirect remote
> >
> > # olbd
> > olb.port 3121
> > olb.path w /tmp/xrootd
> > olb.allow host lxn5219.cern.ch
> > olb.allow host lxn5207.cern.ch
> > olb.allow host lxb1557.cern.ch
> > olb.trace all debug
> > ===================================================================
> >
> >
> >
>