When restarting the MINE process, there will be some additional internal work required by the MINE process to ensure all changes are captured and there are no gaps in the replication. This work is done automatically on startup by the MINE process.
To ensure complete transactional integrity, the MINE process will have to go back to the last uncommited SCN for a transaction and start parsing for any changes since that SCN. There may have been a future commit for that transaction and MINE needs to ensure that all the changes have been mined. This may mean that Dbvisit Replicate will go to the archive redo logs to capture the full transaction history. Plogs PLOGs are written again on MINE restarts.
The Apply APPLY process is not impacted when there is a restart on the MINE process. If it already has a plog PLOG delivered, it will not ask MINE for the new copy. It knows that SCN of last applied transaction and can carry on from the last applied SCN.
When restarting MINE, the Dbvisit Replicate console will initially report what appears to be incorrect progress counters:
| No Format | ||
|---|---|---|
| ||
-MINE IS running. Currently at plog 66 (redo sequence 1424 [1] 1831 [2]) and SCN 47476178 (04/13/2012 18:57:13). APPLY IS running. Currently at plog 52 and SCN 0 (01/01/1988 00:00:00). Progress of replication orcl:MINE->APPLY: total/this execution -------------------------------------------------------------------------------------------------------------------------------------------- SCOTT.AVI_OBJECTS: 133% Mine:6000/6000 Unrecov:0/0 Applied:8000/0 Conflicts:0/0 Last:--/-- -------------------------------------------------------------------------------------------------------------------------------------------- |
The Apply APPLY SCN is set to 0 while the MINE process goes back to the last uncommitted SCN. The % progress will also not reflect the true number while this initialization is carried out.
...