Forum OpenACS Q&A: Re: How to go about debugging a Postgres performance problem?

I looked into this for Janine and the problem is simply one of the buffer, Linux disk, and disk drive caches being flushed of the "mytable" blocks during the VACUUM FULL (there are many large tables on this box).

And on the production box there's a couple of other sites running for the same client, when the staging site sits idle all its database blocks are flushed from cache due to activity on the live sites and the size of their databases.

Running the update a couple of times shows the effectiveness of the caching, as the second time runs many times faster!