ciao Andy! On Mon, Aug 24, 2020 at 9:19 PM Andrew Hanushevsky < [log in to unmask]> wrote: > Hi Tommaso, > > Are we talking about parallel sources or fallback sources? to be exact, none of the two. For me, given two sources A and B: - fallback: try A and if that does not work try B - parallel: read from A and B at the same time what I need instead is - try A or B (with 50% probability) -- so a round robin between sources > TFile::Open > never could handle multiple *parallel* sources. Remember that you always > get multiple fallback sources as long as there is a redirector or the > multiple sources are registered in DNS. You just don't use them in > parallel except for xrdcp because that's where it's most effective. > > That said, you can give TFile::Open a metalink file (one that ends in > "meta4") that lists all the possible sources. that is indeed interesting! I am searching on the web, if you happen to have more resources please let me know... > You just need to construct > one, it's not incredibly difficult and there may be some tools out there > that will do it for you (though I haven't been able to find any that > aren't tied to some larger project). So, if you are looking for "fallback" > sources metalink files are the fastest way to accomplish it today. > > Andy > > Thanks ! tom > On Mon, 24 Aug 2020, Tommaso boccali wrote: > > > ciao Andy, thanks for your answer. > > I understand it will be solved for the xrdcp case. But what about a > > standard TFile::Open? Do I understand correctly that since it accepts an > > URL, it will not be possible to specify multiple sources in round robin? > > > > thanks again > > > > tom > > > > On Mon, Aug 24, 2020 at 6:50 AM Andrew Hanushevsky < > > [log in to unmask]> wrote: > > > >> Hi Tommaso, > >> > >> Ah, you are used to the "old" way of itentifying multiple endpoint > using a > >> comma separated list in he URL. That is no longer supported as it was, > >> rightfully, pointed out that it is a violation of the URL standard. I > >> agree with you tat it would be nice to do it this way but the gaurds of > >> the standard won't allow it. The --sources option is used to add > >> additional sources and additional endpoints come from a redirector or > >> from DNS (or both). That is the only official way of doing it. > >> > >> That said, be out on the lookout on for xrdcp extensions that allow you > to > >> list additional sources via an option. We are in the process of allowing > >> that. > >> > >> Andy > >> > >> > >> On Fri, 21 Aug 2020, Tommaso boccali wrote: > >> > >>> Ciao, in the xrootd code I find (e.g. > >>> https://root.cern.ch/root/html516/src/TXNetFile.cxx.html) > >>> > >>> > >>> // a. open a non-ROOT > >>> <https://root.cern.ch/root/html516/ROOT.html> generic file > >>> // > >> "root://server1:port1[,server2:port2,...]/pathfile?filetype=raw" > >>> > >>> > >>> > >>> as a way to instruct the client to probe more than 1 server. > >>> > >>> but if I try it, it does not seem to work: > >>> > >>> [a07cms04@r113c02s03 CMSSW_11_1_2]$ xrdcp -f 'root:// > >>> r171c01s01.marconi.cineca.it, > >>> > >> > r171c01s02.marconi.cineca.it///store/test/xrootd/T1_IT_CNAF/store/mc/SAM/GenericTTbar/AODSIM/CMSSW_9_2_6_91X_mcRun1_realistic_v2-v1/00000/A64CCCF2-5C76-E711-B359-0CC47A78A3F8.root > >> ' > >>> . > >>> [0B/0B][100%][==================================================][0B/s] > >>> Run: [FATAL] Invalid address > >>> > >>> isn't is supported any more? > >>> > >>> And, assuming it still works, I understand it is not a round robin > >> between > >>> servers, but a failover to the second if the first is not ok. > >>> > >>> Is there a way instead to use a random one between those provided, such > >>> that to implement a "client side" round robin? > >>> > >>> thanks > >>> > >>> tom > >>> > >>> > >>> > >>> -- > >>> Tommaso Boccali > >>> INFN Pisa > >>> > >>> > ######################################################################## > >>> 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 > >> > >> > > > > -- > > Tommaso Boccali > > INFN Pisa > > > > ######################################################################## > > 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 > > -- Tommaso Boccali INFN Pisa ######################################################################## 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