[rosser]

Do you have any numbers on how much extra time (that is, time spent servicing queries above the normal query times when your tables are indexed) application queries take after the loads, but before the index rebuilds are complete?

If so, how does that compare, in aggregate, to the time saved in the loads?

Or are you simply not putting the application back into service until the index rebuilds have finished? How long does that take, compared to the time saved?

EDIT: I'm mostly asking these questions to nudge people to think about them in the course of trying this in their own environments. It's my day job to think about these kinds of things; I've worn the PostgreSQL DBA hat for over a decade now.

[whistlerbrk]

Yes, you would never do this unless writes were completely disallowed to your app. You will save time building the indices in one-shot.

[rosser]

My point is: this is not necessarily true. The Fine Article even mentions that the benefits are situational.