The modified operation always depends on the earlier full update. This works fine when the key exists in the active table of the tiered table. In case it doesn’t, we need to get the key from the member table and insert it into the active table to perform the modification operation on it.

In most of the scenarios, the search is performed before performing the modification, so the key/value along with the time window that is saved in the tiered table cursor could be used to insert it into the active table if the key doesn’t exist and perform the modify operation.

If there was no search performed earlier, the cursor search is performed internally and then the modify operation is performed.

The copy of the update that is moved from the shared table to the active table is performed using the same transaction and its time window. All the conflict checks are carried out using the current transaction id. Copying the same data from the shared table to the active table with a different transaction time window can lead to inserting the same data again into the history store table twice.