There are a few Oracle database replication solutions on the market:
- Oracle Streams (powerful, included with RDBMS license (Oracle SE has trigger-based capture, EE mines redo logs and log buffer), but deprecated - no longer developed, complex to manage)
- Oracle GoldenGate (powerful, but very expensive)
- Tungsten (heterogeneous, but from Oracle side requires deprecated CDC and complex to set up - one interesting feature, you can write data modification plugins before data is applied on target)
- DBVisit (pretty inexpensive compared to GoldenGate, but powerful)
In this blog post I'll give a short overview of DBVisit Replicate, that can be used to replicate data real time between two Oracle databases or from Oracle to MySQL/MSSQL. I am not connected to DBVisit company in any way and I was testing their replication solution for a client of mine.
A few interesting key concepts behind DBVisit Replicate:
- It uses optimistic apply on the target side, meaning that data changes are replicated and applied (but not committed) to the target even before transaction is committed on the source. In case of rollback, the target database needs to roll back all the changes too. Positive side is that committed transactions get replicated to the target faster, even if the transaction is large.
- DBVisit uses its own change capture process to mine online redo logs, so it does not depend on triggers to log the changes and does not impact the end user session. The potential downside - Oracle can change the internal structure of redo logs any time, so before upgrading the database check the DBVisit compatibility first.
- DBVisit can run its CPU intensive processing on a different server, so it does not waste expensive CPU cycles on the Oracle DB server. This is called 3-tier architecture in DBVisit. In this architecture source database only needs to run small FETCHER process, that sends redo log changes over a network to a dedicated MINE process/server that actually does to log processing. MINE filters out the required database changes and sends this processed information over network to APPLY process. APPLY then connects to the target database over OCI (so it does not need to be running on the target database server) and executes the DML statements. (Note: fetcher process is optional, so by default dbvisit runs mine process on the source database server).
DBVisit is very easy to install and it supports RAC and ASM. My setup is done on 22.214.171.124 3-node RAC+ASM running on Oracle Linux 5.8. For Grid Infrastructure (ASM) role separation is in use (GI runs under different OS account than RDBMS). I'm using DBVisit Replicate 2.4.21 (unreleased currently, but it contains a many bug fixes needed for my environment).
In my following easy test setup:
- I'm using the default 2-tier architecture, so no fetcher process. Apply also runs in the same host as the target database.
- I'm using TAR version of the dbvisit software (not RPM), so I could have a single shared copy of the software for all servers in the configuration. If you use RPM, then the same RPM package needs to be installed on all servers (and you need root privileges). In my case I'm using OCFS2 filesystem and dbvisit software is extracted to /u02/app/oracle/dbvisit.
- For processing area for each dbvisit process I'm using /u03/dbvisit/pte in this example. In my current case it is also on an OCFS2 filesystem and shared between all servers, but it does not have to be and when I move this setup to production, I'll also use 3-tier architecture and local disks.
- Grid Infrastructure and ASM run under OS account grid.
- I'm using IP 10.0.0.1 as the server address where MINE is running.
- I'm using IP 10.0.0.2 as the server address where APPLY is running.
First execute the only executable file in dbvisit replicate installation package dbvrep and execute and complete the initialization wizard.
[firstname.lastname@example.org pte]$ /u02/app/oracle/dbvisit/replicate/dbvrep Initializing......done Dbvisit Replicate version 126.96.36.19946 Copyright (C) Dbvisit Software Limited. All rights reserved. No DDC file loaded. Run "setup wizard" to start the configuration wizard or try "help" to see all commands available. dbvrep> setup wizard This wizard configures Dbvisit Replicate to start a replication process. The setup wizard creates configuration scripts, which need to be run after the wizard ends. No changes to the databases are made before that. The progress is saved every time a list of databases, replications, etc. is shown. It will be re-read if wizard is restarted and the same DDC name and script path is selected. Run the wizard now? [yes] yes Accept end-user license agreement? (view/yes/no) [view] yes Before starting the actual configuration, some basic information is needed. The DDC name and script path determines where all files created by the wizard go (and where to reread them if wizard is rerun) and the license key determines which options are available for this configuration. (DDC_NAME) - Please enter a name for this replication (suggestion: use the name of the source database):  pte (LICENSE_KEY) - Please enter your license key (or just enter "(trial)"): [(trial)] trial (SETUP_SCRIPT_PATH) - Please enter a directory for location of configuration scripts on this machine: [/home/oracle/pte] /u03/dbvisit/pte... and so on. In the end the wizard will execute a script on both source and target databases that will create a DBVREP schemas and give it all necessary privileges. If you enabled DDL replication, then it will also enable database wide supplemental logging on the source database side (so check DBA_2PC_PENDING view before doing it, that you don't have any pending 2PC transactions open, otherwise adding supplemental logging will hang until the 2PC transactions are resolved).
MINE (or FETCH in case of 3-tier architecture) process needs to run directly on the source database server (in case of RAC pick any one of the database nodes) and under the same OS account as ASM, so in my case grid. Setup wizard creates a script *-run-10.0.0.1.sh to start MINE.
[grid@xxxxxx pte]$ ./pte-run-10.0.0.1.sh Initializing......done DDC loaded from database (234 variables). Dbvisit Replicate version 188.8.131.5246 Copyright (C) Dbvisit Software Limited. All rights reserved. DDC file /u03/dbvisit/pte/pte-MINE.ddc loaded. Starting process MINE...started
Apply process shouldn't need an installed Oracle client software, because DBvisit Replicate comes with an embedded Oracle Instantclient. In the version I'm currently using this did not work for me, so I needed to add the following line to *-APPLY.ddc file to set the correct ORACLE_HOME. But this bug should be fixed in the next released version.
memory_set ORACLE_HOME /u01/app/oracle/product/184.108.40.206/db
Also open *-run-10.0.0.2.sh (the script that executes APPLY process) and set NLS_LANG on the first line. NLS_LANG needs to be AMERICAN_AMERICA.SOURCE_DB_CHARSET:
Now start apply process:
[oracle@xxxxxxxxxx pte]$ ./pte-run-10.0.0.2.sh Initializing......done DDC loaded from database (234 variables). Dbvisit Replicate version 220.127.116.1146 Copyright (C) Dbvisit Software Limited. All rights reserved. DDC file /u03/dbvisit/pte/pte-APPLY.ddc loaded. Starting process APPLY...started
Monitoring and configuring the replication process is done through the replication console, which can be executed using the start-console.sh script. This will display the status of all dbvisit processes and limited list of tables that have had some changes replicated recently. From this command line you can control the replication process
/MINE IS running. Currently at plog 13 (redo sequence 1201  1395  1086 ) and SCN 96447934933 (04/04/2013 16:49:25). APPLY IS running. Currently at plog 13 and SCN 96447934644 (04/04/2013 16:49:25). Progress of replication pte:MINE->APPLY: total/this execution -------------------------------------------------------------------------------------------------------------------------------------------- DBAUSER.DBVISIT_PING: 100% Mine:21/21 Unrecov:0/0 Applied:21/21 Conflicts:0/0 Last:04/04/2013 18:24:36/OK -------------------------------------------------------------------------------------------------------------------------------------------- 1 tables listed. dbvrep>
Some useful commands: LIST PREPARE, PREPARE SCHEMA, PREPARE TABLE, UNPREPARE SCHEMA, UNPREPARE TABLE, SHUTDOWN MINE, SHUTDOWN APPLY, SHUTDOWN ALL, LIST CONFLICT. Before you add (prepare) new tables/schemas with existing data to replication configuration, take a look at the users guide for a proper procedure. If you just execute PREPARE TABLE/SCHEMA and then export the existing data, you will get ORA-01466.
For my current project, it was very important to find a replication solution that could exclude some transactions from replication, for example when you need to purge data from source database but want to keep the same data on the target DB. It is possible with DBVisit Replicate:
- Partition level DDL is not replicated by default, so on the source database you can just drop/truncate a partition and it will not be replicated by default to the target side.
- If you need to exclude specific transactions from replication, then execute SET TRANSACTION NAME as a first command in that transaction.
SET TRANSACTION NAME 'DBREPL_DB_%s_XID_%s'
The first %s: name of the target database (as configured in the setup wizard).
The second %s: is not relevant.
COMMIT; -- just to be sure that the next command is the first in that transaction SET TRANSACTION NAME 'DBREPL_DB_archpte_XID_XXXXX'; DELETE FROM dbauser.dbvisit_ping; COMMIT;
I think this is enough for a first post. You can do a lot of complex configurations with DBVisit, it is a flexible product. Test your setup properly, there can be issues depending on your database setup, that DBVisit has not yet tested for. If you find an issue, report to DBVisit support (this can also be done if you have a trial license), DBVisit has an excellent and fast support team. So far I have created 7 tickets to DBVisit support and all of them have been resolved within hours or a day.
DBVisit also has some helpful videos in youtube.