 |
 |
On Mar 19, 2013, at 15:09 , Andrew Hanushevsky <[log in to unmask]> wrote: > Hi Tommaso, > > I trust that your source modifications are working for you. Just to let you know that git head now contains a stat() plugin mechanism and a sample plugin specific to gpfs+tape. It's configurable and should do all the things you wanted it to do. The plugin shared library is libXrdOssSIgpfsT.so with the corresponding source in XrdOss/XrdOssSIgpfsT.cc cool! I would feel much more relaxed if i could use an official release ;) And yes, all the functionality tests we did seem to say we have the features we needed. > > We will now discuss when it will appear in a mainline release. This brings me to another question regarding gpfs+tape staging. Presumably, when you stage in a file using your interface, the file will appear on disk using exactly the same name it originally had. indeed. For "ls" the two situations are completely equivalent. Only "stat" shows the difference. > Currently, this is not the assumption of frm_xfrd since not too many people run such a tightly configured staging system. So, I will also include an additional option to the copycmd directive indicating that direct staging is the prefered method so that the frm doesn't go through an intermediate file. I suspect you have already hit that little detail. > well, I think this is inside my ad hoc recall script (I copy one in /afs/pi.infn.it/user/boccali/public/xrootd_mycomp/copy.sh … probably not the last one but close) called via copycmd I do not create any tmp file, I just ask GEMSS via a file interface (ou probably are not too much interested in the details) to bring the file online. And I exit==0 if all fine, exit==1 if not. This appears to work fine. I was also worrying a bit about the appearence of .fail files, but they do not seem to happen (which I like, if they would they would even migrate on tape!) Another detail was the one you spotted (!): if recall exits with ==1, I should not unlink the tmp file since there is no tmp file… an Unlink() would delete the file also from tape! that is all I guess … when the code hits a binary release I will be happy to test! I can even use the git code, if that comes able to build rpms via rpmbuild … is that the case? thanks tom > Andy > > ######################################################################## > Use REPLY-ALL to reply to list > > To unsubscribe from the XROOTD-L list, click the following link: > https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1 ######################################################################## Use REPLY-ALL to reply to list To unsubscribe from the XROOTD-L list, click the following link: https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1