Developer report: 6-17-2005

This week:
1) Updated the 'encp history page' help information.  Now includes details 
   of the rates reported by encp.
2) Investigated "CRC DCACHE MISMATCH" errors from CMS. 
   * The CRC that dCache was writting into layer 2 and that ecrc was 
     calculating matched.  encp and the mover were calculating
     a different CRC for the error files.  
   * Determined that there is a kernel/filesystem bug that silently causes 
     file coruption for encp when reading from local disk when using XFS,
     reading the file with direct i/o and the file is larger than 4GB.
   * We've stopped using direct i/o on all dCache nodes.