Hi Fons,
On Tue, Aug 17, 2004 at 12:10:16PM +0200, Fons Rademakers wrote:
> unfortunately I play tennis from 19-20 on Tuesday's.
Yes, the time was chosen mostly to accomodate Andy. If there is significant
interest from new people on this side of the Atlantic perhaps we can convince
him to call in earlier (18:00?) on some other day...
> The main issues I've at the moment are:
>
> - build procedure of sec modules separate from xrootd.
> Currently we can build xrootd and therefore the sec modules only
> on Linux and Solaris, however we really require the netx client to be
> able to build on all ROOT supported platforms, incl Win32. Since for
> the time being the building of the client is linked to the building
> of the server we have the client only on Linux and Solaris.
Ok, I'll stop by and talk to Gerri about this after lunch.
> - support for "localhost" in the url (backward compatible with TNetFile)
Ah, yes, I ran into this the other day. In fact I ran into a related feature
the other day (and spent 30 minutes thinking I was going out of mind):
TFile::Open() will force the use of TFile (as opposed to TXNetFile) if it
thinks that the file really is local. With xrootd there are a number of cases
where that will not be the correct behaviour:
o When testing client/server on a single machine (in a "hostile"
environment in some other country, my case from the other day...)
o When testing the client on a server which uses the olbd to redirect
the client to another machine (which is the one which actually has the
file)
o When testing the client on a server which is using the option to stage
the file from mass storage. Accessing it directly often won't work.
o xrootd allows the file namespace to be placed "within" another directory,
e.g. perhaps my server has space in /mnt/somebigdisk and my clients
are all expecting /store/foo/bar.root type files. I can put it in
/mnt/somebigdisk/store/foo/bar.root and with a config option set xrootd
up to access that file even if clients come in with:
root://myserver//store/foo/bar.root
This allows the sysadmins to configure/name filesystems however they
like without being constrained to mount things in an experiment-specific
way (e.g. BaBar's filespace starting with /store requiring a mount point
/store, etc.)
Personally I would be happy to just see this TFile optimization removed
(since if I ask for root://somehost//path/file.root that is probably what I
meant, not "/path/file.root"), but if somebody really depends on that I guess
a TEnv parameter to allow me to force it to use TXNetFile would be fine.
Pete
> On Tue, 2004-08-17 at 09:15, Peter Elmer wrote:
> > Hi All,
> >
> > The next meeting on xrootd development, testing, deployment and monitoring
> > will be Tuesday, 17 August, 2004 at 10:15 CA, 19:15 Europe and 18:15 UK. The
> > location at SLAC is SCS conference room B. At CERN we will meet in my office
> > (Bldg. 32 2C-14). We'll see how many people are around this week.
> >
> > The connection info for this meeting is:
> >
> > phone: +1-510-665-5437
> > passcode: 4465
> >
> > The agenda is:
> >
> > o mailing list/HN
> > o latest fixes - Alvise, Fabrizio, Andy
> > o progress on the problem list
> > o testing asynchronous mode
> > o getting rid of feature in KanAccess.cfg that allows "localroot" to
> > be specified there
> > o New development
> > o posix-compatible client
> > o proxy support
> > o client side monitoring support?
> > o ROOT integration of XTNetFile/xrootd
> > o ./configure issues with kerberos/security stuff reported by Axel?
> > o building on MacOSX
> > o Verbose echo of compilation commands (requested by Fons)
> > o XTNetAdmin
> > o wrapper - Fabrizio
> > o SRT integration - Pete
> > o checksums
> > o Sites
> > o Example configurations
> > o Updates to servers at SLAC on Thursday and during the computing shutdown
> > o moving to latest version (20040810-2328) for both xrootd and olbd
> > o cleanup of config files
> > o making all access to /store readonly
> > o turning off non-xrootd log rotation for olbd
> > o olb03/olb04/kanolb-a renaming
> > o Padova - Fulvio/Guglielmo
> > o problems reported by Antonio:
> > http://babar-hn.slac.stanford.edu:5090/HyperNews/get/KanSOS/393.html
> > o CNAF
> > o problems reported by Enrica
> > o RAL - Chris
> > o outstanding problems after xrootd 20040810-2328?
> > o Is the olbd crash gone?
> > o new disk space?
> > o import of skims
> > o rpm structure (libstdc++, config files)
> > o cleaning up classic Kanga data
> > o In2p3 - Jean-Yves
> > o use and status (disk space)
> > o status of import of kan data (presumably skims at this point)
> > o confusion about what should be made available at In2p3 (ARTF)
> > o plans to move to more recent versions (beginning of September)
> > o xrootd procedures
> > o restart of servers (deployment of new versions)
> > o problems with particular data servers
> > o adding a new server
> > o ganglia/monitoring of kanNNN servers at SLAC - Yemi
> > o metrics for staging?
> > o monitoring of scratch area, etc.
> > o adding back xrootd metrics (after restart on Thursday?)
> > o Updates to presentation
> > o alarms?
> > o monitoring of #file descriptors/TCP connections, etc.
> > o monitoring of xrootd versions and alarms when they fail the "Sesame
> > Street test" (Should be possible after restart on Thursday!)
> > o renaming of servers during/after shutdown
> >
> > I have almost certainly forgotten some things, so please feel free to bring
> > anything important up during the meeting.
> >
> > 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
-------------------------------------------------------------------------
