We know that rows distribution takes place based on Row hash value of primary index column of the table. Can anyone tell how the distribution of data amongst the fallback amps take place? Also, when the amp is down, how the data is read from the Fallback amp? Basically, please tell as what governs the distribution and access of Fallback data.
Check the manuals on the HashMaps - there exits two. One primary, one for the fallback. The two hashmaps gurantees that the fallback data will be placed on a different cluster.
In case DB detects that an VPROC is down and table is fallback it will use the fallback hashmap to locate the fallback data.
Der Beginn aller Wissenschaften ist das Erstaunen, dass die Dinge sind, wie sie sind.-Aristoteles
Ulrich, the fallback row is always in the same cluster as the primary.
Jim, yes you are right - I used the wrong term, was thinking of the cabinet - but this depends on your HW...
Section AMP Clustering and Fallback in the Database Administration doc is explaining it good.
The other important considerations are
For AMP failures due to hardware-related problems, to protect the availability of your data,
define AMP clustering assignments as closely as possible to the following fault-tolerant
strategy. No two or more AMPs in a cluster:
• On the same node (for MPP systems).
• In the same node cabinet.
• Serviced by the same disk array cabinet.
• Serviced by the same disk array controller.