e i g h t - c u b e d . c o m 

Repartitioning the GS160s

This morning, I shut down the two AlphaServer GS160s that make up the data warehouse cluster so I could repartition them. Originally, the systems were purchased to run a large Rdb database in an extremely efficient manner, and they were partitioned to support this.

Unfortunately over the last year, the team that does the application design for the data warehouse system have changed direction and have switched to supporting multiple databases on Oracle 9i. This invalidated the original plan, which was to use Rdb's hot standby feature to supply a "reporting" copy of the database.

The reporting copy was to be accessed via it's own partition, and so the machines where carved up into three: one for batch read/write one for interactive, and one for read only reporting.

Because the evolution this morning was rather hurried, I repartitioned the easy way just by bit twiddling the LP_* console variables, as opposed to moving PCI cards and cables around to support the optimal setup of two 8 CPU nodes, one in each firebox.

That's the next step...