Hi Daniela,
> My WMS (wms01.hep.ph.ic.ac.uk <http://wms01.hep.ph.ic.ac.uk>) currently
> seems to be unable to handle jobs that look up a file location on an
> LFC. It works for a couple of jobs (so in principle it can handle it),
> but fails soon with the following error in the workload_manager_events.log:
>
> 24 Jan, 18:19:44 -W: [Warning]
> resolve_filemapping_info(dli_utils.cpp:200): cannot load DLI_SI helper
> lib libglite_wms_dli.so
> 24 Jan, 18:19:44 -W: [Warning]
> resolve_filemapping_info(dli_utils.cpp:201): dlerror returns:
> libglite_wms_dli.so: failed to map segment from shared object: Cannot
> allocate memory
> 24 Jan, 18:19:44 -W: [Warning]
> resolve_filemapping_info(dli_utils.cpp:335): cannot perform
> lfn:/grid/t2k.org/nd280/raw/ND280/ND280/00005000_00005999/nd280_00005000_0000.daq.mid.gz's
> <http://t2k.org/nd280/raw/ND280/ND280/00005000_00005999/nd280_00005000_0000.daq.mid.gz%27s>
> resolution
>
> The memory usage of the workload manager is about 18% at the time (the
> machine has 16GB). Restarting it (I just restart everything
> /etc/init.d/gLite restart) helps briefly, but it's a matter of minutes
> rather than hours before it fails again.
That could be due to a network issue between the WMS and the DLI:
can you check e.g. if there is a non-trivial QoS applied to traffic
for the DLI web service? Some intermediate router/switch/... might
have a configuration that puts limitations on such traffic.
Of course there also could be a bug in the affected components...
|