Capacity Forecasting ( CPU/Memory/Disk)

Question

Trying to forecast the above metrics for our postgres database servers. We have nagios and cacti. How have you done this?

Answer 1

Disk:

You can experiment with pg_column_size to figure out how big a value (or a table row) will be on disk.

There will be a lot of overhead: headers, empty space in disk blocks, and of course all the space for indexes you create. Estimate something between three and ten times the space got with pg_column_size.

You will also have to reserve disk space for archived transaction logs (WAL).

Memory:

As much as possible for a busy database.

CPU:

As many cores as you expect concurrent queries.