Fallback is very important when a system needs to reconstruct data from fallback copies if a hardware read error occurs when it attempts to read the primary copy of the data. When a read error occurs in this case, the file system reads the fallback copy of the hash index subtable rows and reconstructs a memory-resident image of them on their home AMP. This is referred to as Read From Fallback. See Teradata Vantage™ - Database Design, B035-1094. Without this feature, the file system fault isolation logic would abort the transaction and, depending on the error, possibly mark the index as being down. See SET DOWN and RESET DOWN Options.
- Requests that do not attempt to modify data in the bad data block
- Primary subtable data blocks
- Reading the fallback data in place of the primary data.
In some cases, Active Fallback can repair the damage to the primary data dynamically. In situations where the bad data block cannot be repaired, Read From Fallback substitutes an error-free fallback copy of the corrupt rows each time the read error occurs. To avoid the overhead of this substitution, you should drop and recreate the hash index.
To enable the file system to detect all hardware read errors for hash indexes, you should also set CHECKSUM to ON.