G
George Potemkin
Guest
> promon is showing a lot of db reads (I'm guessing because its doing index reads) and that's about it. For the non-unique indexes can-find() always reads exactly index blocks and 1 index key where is the number of the levels in the index tree. This value used to be less or equal 5-6 for very large tables. For the unique indexes can-find() may read abnormally large number of index blocks and index keys because it will read and skip the "index entry locks" left by data purge. So activity caused by can-find() is a good probe test to decide if it's a time to run idxcompact against the unique indexes. All needed data are available from VSTs. Poor utilization of index blocks is NOT a factor that affect can-find().
Continue reading...
Continue reading...