Repository navigation
OpenSearch source for Serverless should default to point-in-time #6335
Description
Is your feature request related to a problem? Please describe.
When the opensearch source was originally created, Amazon OpenSearch Serverless did not support point-in-time. So the default behavior for the search is to use none rather than point-in-time (PIT). For other types, the source detects whether the cluster support PIT and then uses it if it does.
Describe the solution you'd like
Data Prepper's opensearch source should use PIT for Serverless by default.
Describe alternatives you've considered (Optional)
This can be done either as a breaking change in Data Prepper 3.0. Or it could just be done in the next minor version release.
If we change this in a minor version release, there is the possibility that existing pipelines will have different behavior if the pipeline is heavily designed with the expectation that PIT is not in use. But, I'm unsure if there really would be a significant pipeline configuration for such scenarios.
Additional context
N/A
Activity
srikanthpadakanti commentedon Apr 6, 2026
I think this can go into a minor release. PIT gives more consistent pagination and I doubt anyone is relying on the NONE behavior intentionally. That was only the default because Serverless did not support PIT back then. Now that it does, switching the default just makes things work better out of the box. If someone does need the old behavior they can still set search_context_type to none.
dlvenable commentedon Apr 6, 2026
@srikanthpadakanti , I tend to agree with you on this. It is unlikely that this will break functionality for Serverless users.