So right now I’m doing an upgrade of one of our test systems from Agile Advantage SP4 on Windows to Agile 184.108.40.206 on Linux. It’s quite the involved process and I’ll post more at length about it later. However, two things I just finished fixing that saddens me:
Oracle doesn’t follow Oracle’s own best practices with the seeded Agile 9.3 database. I have no idea why they did this. It’s not because this best practice is new – it’s been the recommendation since I learned it back in 1997 with Oracle 7.3.
Agile 220.127.116.11 lays down a database with 4 redo log file groups, each with 1 200MB member. Oracle best practices with redo logs is 3 redo log file groups each with 2 members. Depending on the system I/O characteristics you may add more redo log groups and the size is highly dependent on the I/O characteristics. All Oracle databases should at least have 2 redo log groups each with 2 members. So instead of the minimum 2×2 or best practice 3×2 configuration, Oracle provides one of their products with a 4×1 configuration which is just asking for issues.
On a related note, the Oracle Agile 18.104.22.168 database also came with an UNDO tablespace that’s sized to autoextend automatically until it fills up the file system. Best practice is to NEVER set an UNDO tablespace to autoextend because a poorly written query can possibly use all the disk space on the system.
Seriously Oracle, what were you thinking?