What if you only cut-over to the new table only once the new indexes have completed building?

What's the relationship between the old indexes and the new table? Presumably the new table has new rows? Is there some partial overlap?

On Fri, Nov 20, 2015 at 10:37 AM, Jaime Solano <jdjsolano@gmail.com> wrote:

Hi James, thanks for your quick response.

We have a table that needs to be recreated from scratch, every X days. This table is accesed by a web app, and has around 12 indexes to improve latency.

As part of the recreation process, we clone the table, and make this new, cloned table the active one. Thing is, creating the indexes for this new table, takes a few minutes.

What we wanted to explore the possibility of reusing the existing indexes (of the about-to-be-droped table) and somehow link them to the new, cloned table.

Let me know your thoughts about this.


On Nov 20, 2015 1:19 PM, "James Taylor" <jamestaylor@apache.org> wrote:
Hi Jaime,
Not sure exactly what you mean. Would you mind explaining a bit more what you're trying to do (and why)?

On Fri, Nov 20, 2015 at 9:01 AM, Jaime Solano <jdjsolano@gmail.com> wrote:

Hi guys,

As part of a swapping strategy we're testing, we want to know if it's possible to link the indexes from one table to another one (clone of the first). We've tried a few things on the SYSTEM.CATLOG with no luck. The real question would be....is this even possible? If so, how?