130 likes | 250 Views
FAX Current situation. As shown by hc WAN test. sites. Work-around. Either server does it for you or client has to do it. As HC jobs (clients) access all the servers from the same environment they always apply it. export XrdSecGSIUSERPROXY =${X509_USER_PROXY} _ XrdProxy
E N D
FAX Current situation As shown by hc WAN test
Work-around • Either server does it for you or client has to do it. • As HC jobs (clients) access all the servers from the same environment they always apply it. • export XrdSecGSIUSERPROXY=${X509_USER_PROXY}_XrdProxy • voms-proxy-init -d -voms atlas:/atlas -vomsesvomses -key $X509_USER_PROXY -cert $X509_USER_PROXY -out $XrdSecGSIUSERPROXY -limited; • voms-proxy-info -file $XrdSecGSIUSERPROXY -all; • echo $XrdSecGSIUSERPROXY; ls $XrdSecGSIUSERPROXY
NET2 and HU • Can access data only from sites without authentication requirement: CERN, ALGT2_DXDR, MWT2_DXDR and MWT2_DCAP. Even then some combinations timeout. • From Saul: • I'm surprised that anything works at all. We haven't actually joined the federation except for a *very* old installation. I was waiting to use our current gatekeeper for a federation node, but switching over to a new gatekeeper has to be done carefully. I'm getting tempted to install on a separate 10Gbps machine just to get things going.
CERN • As server: • All clients can copy/read from it. Efficiency 30-60% • As client: • Can copy/read from all the servers. CPU efficiency ~ 20% except in local where ~95%. • Monitoring: • Does not work.
BNL • As server • Can’t do xrdcp to AGLT2 or BNL • xrdcpto SWT2_CPB very slow. • Direct access does not work from AGLT2. • As client • xrdcp only from sites without auth. • Direct access works from all the sites. Efficiency 10-90%. • Monitoring: • Does not work.
OU_OCHEP_SWT2 • As server: • Delivers in both xrdcp and direct access • As client: • Works from all the sites. CPU efficiency 20-95%. • Monitoring: • Works.
SWT2_CPB • As server: • Delivers in both xrdcp and direct access. Eff. 40-95% • As client: • Works from all the sites. CPU efficiency 20-60%. • Monitoring: • Works.
slac • As server: • Delivers in xrdcp and direct access. Eff: 10-70% • As client: • Can’t access Oracle in CERN. • Monitoring: • Works.
MWT2 • As server: • Delivers in both. Eff: 20-95% • As client: • Delivers. Eff: 30 – 90% • Monitoring: • Does not work.
AGLT2 • As server: • Delivers but performance very bad even for itself. Spread all over the place. • Looking at server log file there are authorization problems (certificate chain broken, no certificate, etc. ) • As client: • Delivers. Eff: 30-90% • Monitoring: • Works. Many jobs hang in detailed monitoring.
Mwt2 vs mwt2_Dxrd vs MWT2_DCAP • Xrdcp • Direct access
AGLT2 vs AGLT2_DXRD • Much better AGLT2_DXRD just shows problems in configuration of AGLT2.