Dbvisit Replicate has the following components and concepts. Dbvisit Replicate only has one software component and this is called dbvrep.
The MINE process reads the Oracle online redo logs at the source database in real time. The MINE process converts information from the Oracle redo log into a Dbvisit Replicate internal format called a PLOG (parsed log). When the information required is no longer in the redo logs, Dbvisit Replicate will automatically switch to the archive logs and switch back to the redo logs when it has caught up. Redo and archive logs using filesystem and ASM are supported (note that only ASM 10.2. or higher is supported). There can be more than one MINE process. Each MINE process will have a unique name which is determined by the user. Example: MINE, MINE2, etc.
The APPLY process takes the PLOG (created by the MINE process) and converts this information into SQL which can be run against the target database. There can be more than one APPLY process. Each APPLY process will have a unique name which is determined by the user. Example: APPLY, APPLY2, etc.
FETCHER is an optional component. It can be used to offload the mining of the source database to another server. It reads online and archive redo logs at the source database and ships them to the MINE process. There can be more than one FETCHER process. Each FETCHER process will have a unique name which is determined by the user. Example: FETCHER, FETCHER2, etc.
MINE reads oracle online redo logs and creates PLOGs (parsed logs). These logs contain parsed information and are filtered to contain information about replicated tables only. PLOGs are binary logs specific to Dbvisit Replicate. The PLOGs are transferred to the target server where they are used by the APPLY process. PLOGs are platform-independent.
Dbvisit Replicate uses the same optimistic principle as Oracle applies in transactions. Oracle assumes that most transactions are committed and starts writing the transactions to disk before the commit has been issued. Dbvisit Replicate applies the same optimistic commit principle during replication which means that Dbvisit Replicate will start mining and applying the changes before the actual commit has been issued. For large transaction this can be beneficial as the transactions are not buffered, but are mined and applied as soon as they are written to the Oracle redo logs on the source database. This can reduce the amount of disk and memory required while waiting on commit and it also reduces the gap between MINE and APPLY to a few seconds.Dbvisit Replicate supports full rollback of the transaction but will require more work to rollback the transaction. This is similar to the work required by Oracle incase of a rollback.
Dbvisit Replicate consists of only one executable. This is dbvrep on Linux and Unix and dbvrep.exe on Windows. The dbvrep executable handles the mining, the applying, the console and includes the setup wizard.
Dbvisit Replicate is installed differently depending on the platform:
- RPM or
- TAR file
- TAR file
- Windows installer