HP OpenVMS Systems

ask the wizard
Content starts here

Recovering from RMS Bucket Check Errors?

» close window

The Question is:

I am working with a large (10 Million blocks)indexed file (prolog 3) that
 contains 6 indexes.  I can read using the primary index and all but one
 alternate index.  When I attempt to read sequentially using this alternate
 index I get the old "bucket format
 check failed" error. (Bucket #591310). Is there a simple way to repair a
 single alternate index in this file?  Is there a way to simply repair this
 bucket?  I understand that I can rebuild the entire file but time is precious!

The Answer is :

  Ten million blocks is not particularly large for an RMS file.
  The usual tool for recovering from RMS file errors is the CONVERT command,
  and do ensure the OpenVMS system has current mandatory ECO kits installed.
  Reconstruction of specific buckets tends to require the assistance of HP
  Customer Services or another service organization (very) familiar with the
  internal structures of an RMS file, and with the tools used to patch the
  file.  (There are Freeware tools that can be used to patch blocks within
  a file, but any such operations do require knowledge of RMS structures
  and of the particular corruption that has arisen here -- whether this
  is a simple patch to a check byte, or a more extensive rebuild.)
  The OpenVMS Wizard would tend to perform a trial conversion into another
  copy of the file as a test, with the process quotas and the file allocation
  and file extent sizes set appropriately for the conversion operation, and
  targeting a disk that is not heavily fragmented nor seeing I/O contention.
  This to determine the required window for the conversion, of course, and
  to then determine the appropriate course of action -- CONVERT, or direct
  patching of the file.
  The trial conversion can also flag other latent errors, and can also
  potentially improve RMS file performance and potentially even system
  performance.  An effective conversion with appropriate quotas and FDL
  settings, and appropriate input, output, convwork and sort locations,
  should not require that much more time than a BACKUP of the file.
  You can also use this to adjust the bucket sizes, key depths, etc.,
  to meet current and expected requirements.

answer written or last revised on ( 28-JAN-2004 )

» close window