In this multi-part post, I’m explaining the basics behind how SAP Replication Server works when replicating from a SAP ASE database to a SAP ASE database as part of an HADR (ASE always-on) setup for a SAP Business Suite system.
The post will be based on SAP ASE (Adaptive Server Enterprise) 16.0 HADR with SAP Replication Server (SRS) 16.0.
A recap of what we discussed in Part 1:
- What is SRS.
- The basic premise of HADR with SRS.
- What a transaction is.
In this part we discuss the ASE transaction log and how important it is for the working of the database then we discuss how transactions are replicated to the active SRS and touch on the internals of the SRS.
What is the ASE Transaction Log?
The ASE transaction log is the most active part of an ASE database.
When we talk about database ACID principles, the transaction log is what helps deliver those things. It is simply a log of transactions applied to a database.
The transaction log is where changes to the database are first made, it is the register, used to ensure that the database can be recovered to a consistent point-in-time.
Some of the seven ASE databases hosted in an ASE 16.0 instance do not need this point-in-time recovery capability, so they have a special setting that simply truncates the records of the transaction log frequently. The setting/option is called “trunc log on chkpt” and should *never* be applied to the SID database (the BusSite database) in production.
The SID database holds our vital business data, and needs to be recoverable right up to the last transaction that had a final state (committed or rolled-back) applied on that database.
On a regular interval, data changes contained in the transaction log that have a final state, are applied to the database data devices (the database main persistence).
This process is known as a checkpoint.
In those ASE databases with option “trunc log on chkpt” turned on, a checkpoint causes the used transaction log entries to be simply truncated (cleared), with no dump required.
In ASE databases with “trunc log on chkpt” turned off (like the SID database), after a checkpoint has occurred, the used transaction log segments are marked as freeable, and these freeable segments can become available for re-use only once a transaction log backup (dump) is performed to a backup disk or device.
This is a cycle that allows the database to continue to be used for making changes to the database data while also protecting the register of transactions in case of a database crash or even complete corruption.
If the transaction log ever becomes full, then the database is unable to allow new changes and no new transactions can be started until space is made available in the transaction log. It is therefore very important for regular database transaction logs dumps to be performed.
Not only to free space in the transaction log, but also to protect the database in the event that the transaction log is lost or corrupted. For added protection, these transaction dump file(s) can also be moved off to a separate location, away from the database server itself. This is a function of almost every backup tool and as you will find a requirement of almost every database that abides by the ACID principles.
Which Databases are Replicated?
Each ASE instance hosts multiple ASE databases.
In a Business Suite system, the ASE 16.0 instance hosts 7 databases (plus 2 temporary databases and a model database):
In a HADR system, we usually only replicate the Master database and the SID database. This is because the master database holds vital information about the user logins, roles and privileges, which are needed to be the same in the target system. The SID database is the database for the SAP Business Suite system, it holds our business data, which we need to protect.
How do Transactions Move to the SRS?
Inside the replicated primary ASE databases, there exists a Replication Agent process that runs in its own process thread.
The Replication Agent scans the transaction log of the database it serves and sends the transactions across to the active SAP Replication Server (SRS).
In the active SRS a matching receiver thread (Rep Agent User component thread) receives the transactions and pushes them into the Simple Persistent Queue (SPQ) on the disk of the active SRS server.
Subsequent components inside the active SRS process the transactions from the SPQ and eventually push them into the secondary database.
Part 3 is now available here.
You may also be interested in: