Re: Do any of you have a policy on rebuilding indexes?

Yes that is correct, build the index with the most keys first and subsequent indices will share that one rather than building a new one. You will actually get a message '...shares access path...'

Now if you wanna go completely nutso on creating all your indices in the best possible sequence, follow these three simple steps:

SAVLIB YourLib ACCPTH(*NO)

CLRLIB YourLib

RSTLIB YourLib

IBM i, proving it's as brilliant as we like to believe it is, will build the access paths in the sequence that will produce the fewest access paths by sharing the maximum number of access paths.

Now before anyone goes off running these three 'simple steps,' DO understand that this could take 'Some Minutes' (and 'Some,' in this case could be VERY significant!!) If you've never restored a large library that did not have access paths saved, you may be in for a shock!

If you DO want to try this, doing so during the migration to your new Power10 with NVMe storage might be a good time to test this!!

- DrF

On 4/1/2023 3:08 PM, Patrik Schindler wrote:

Hello Vern,

Am 01.04.2023 um 19:45 schrieb Vern Hamberg via MIDRANGE-L <midrange-l@xxxxxxxxxxxxxxxxxx>:

And if I remember, if you create an index for keys A, B and then create an index for keys A, B, C, there is no sharing at that time - you can prove that easily. It sounds like the system adjusts these relationships in a restore, but not when first created.

An interesting remark! Thanks for sharing!

:wq! PoC