 |
 |
Hi Jacek, The collector runs xrootd-20050417-0431 on Scientific Linux CERN release 3.0.3 (SL). The dataserver runs xrootd-20050417-0431 on Slackware 10.1.0. Here is a part of the xrootd log. I don't see anything strange, but maybe it helps you: 050428 00:00:00 17956 (c) 2004 Stanford University/SLAC xrd version 20050417-0431_dbg 050428 00:00:00 17956 XrdSched: scheduling midnight runner in 86400 seconds 050428 00:00:09 18092 XrdSched: running monitor window clock inq=0 050428 00:00:09 18092 XrdSched: scheduling monitor window clock in 15 seconds 050428 00:00:24 17956 XrdSched: running monitor window clock inq=0 050428 00:00:24 17956 XrdSched: scheduling monitor window clock in 15 seconds 050428 00:00:31 001 XrdInet: Accepted connection from pcardaab.cern.ch 050428 00:00:31 18092 XrdSched: running ?:16@pcardaab inq=0 050428 00:00:31 18092 ?:16@pcardaab XrdPoll: FD 16 attached to poller 0; num=1 050428 00:00:31 18092 XrootdXeq: catac.14795:16@pcardaab login 050428 00:00:39 17956 XrdSched: Now have 3 workers 050428 00:00:39 17956 XrdSched: running monitor window clock inq=0 050428 00:00:39 17956 XrdSched: scheduling monitor window clock in 15 seconds 050428 00:00:46 18092 XrootdXeq: catac.14795:16@pcardaab disc 0:00:15 050428 00:00:46 18092 catac.14795:16@pcardaab XrdPoll: sending poller 0 detach for link 16 050428 00:00:46 17957 XrdPoll: Poller 0 detached fd 16 entry 1 now at 1 050428 00:00:46 18092 catac.14795:16@pcardaab XrdPoll: FD 16 detached from poller 0; num=0 050428 00:00:54 19190 XrdSched: running monitor window clock inq=0 050428 00:00:54 19190 XrdSched: scheduling monitor window clock in 15 seconds 050428 00:01:09 17956 XrdSched: running monitor window clock inq=0 050428 00:01:09 17956 XrdSched: scheduling monitor window clock in 15 seconds ... And it's full configuration is: # xrootd xrootd.fslib /home/catac/mywork/xrootd/lib/arch/libXrdOfs.so xrootd.export /tmp xrootd.async off xrootd.monitor all flush 30s window 15s dest files io info user pcardaab.cern.ch:9930 xrd.trace all -debug oss.readonly odc.manager pccil 3121 # olbd olb.port 3121 olb.subscribe pccil 3121 Cheers, Catalin. Jacek Becla wrote: > Hmm, you do get packets with zero length. What version of xrootd are you > running and on what OS? > > Jacek > > > > Catalin Cirstoiu wrote: > >> Hi Jacek, >> >> Thanks for the directions. However, I started the collector and when I >> do a xrdcp I get this: >> >> ... >> RT locking.ok.0.unlocked >> RT locking.ok.0.unlocked >> RT locking.ok.0.unlocked >> Caught exception 130 "Invalid packet length 0" >> Caught exception 130 "Invalid packet length 0" >> RT locking.ok.0.unlocked >> Caught exception 130 "Invalid packet length 0" >> RT locking.ok.0.unlocked >> ... >> >> This is why I mentioned in the previous mail the 0-lenght of the packets. >> >> Cheers, >> Catalin. >> >> Jacek Becla wrote: >> >>> Hi Catalin, >>> >>>> No problem. I'll be looking forward to the binary dump. Jacek, can >>>> you point Catlin to the binary collector and provide simple >>>> instruction on how to capture the information? Thanks. >>> >>> >>> >>> >>> the program for dumping packets is not built by default, you will >>> have to tweak src/XrdMon/GNUmakefile to enable it (uncomment lines >>> 83-85), then build. >>> >>> Run the collector "xrdmonCollector", no arguments are necessary. It >>> will store collector's logs in ./logs/collector/<xrootdServer>/<port>/. >>> >>> You will need to have some activity on the server to trigger at least >>> one flush of the collector's buffers. >>> >>> Most likely you will not want to wait until it fills up the first log >>> file, so create the following link: >>> ln -s logs/collector/<xrootdServer>/<port>/active.rcv >>> 20050427_12:00:00.000_sender:1000.rcv >>> Then run "xrdmonDumpPackets 20050427_12:00:00.000_sender:1000.rcv" >>> >>> This tool will produce files for each packet received and store them >>> in /tmp directory. The files will be called "ap.dump.<sequenceNo>" >>> >>> Let me know if you have any questions (the dump tool is not >>> productized yet, thus all the hassle with symlinks) >>> >>> cheers, >>> Jacek > >