Asynchronous index rebuild on first access
Pain point: Index rebuild stops if user terminates the connection accessing the table with an index that is marked inconsistent and it starts over from the beginning again on next table access by another user.
An index object is marked invalid and needs to be rebuilt. In our case, an offline table reorg was canceled while in the offline index rebuild phase. The next access to the table will drive index recreate. If the next access is made by a user application, it is possible that this application might time out. The result is that the work performed by the index rebuild is lost and we need to start from the beginning on next access. If the subsequent accesses to the table are made by user applications, then we might never complete the index build until the database administrator steps in to correct the problem.
Ideally DB2 should rebuild the index object asynchronously and have the agent that first accessed the table wait behind that rebuild. This way if a time out is hit by the user application, the index rebuild can continue uninterrupted. Subsequent access to the table will then wait behind the index rebuild and they are free to time out as well. The index rebuild can complete in the smallest amount of time possible without the intervention of the database administrator.
Please update us on the progress of this request.