Skip to content

Only expose safely stored data to reads #52400

@ywelsch

Description

@ywelsch

Dirty reads happen when an operation is returning results based on data that has not yet made it fully into the system (for example has only been indexed on the primary, but hasn't made it yet to the replicas, which means that if the primary fails, the data might be lost and the previously returned result would be incorrect). Eliminating dirty reads means that the user will never see data that has not been safely stored on all in-sync shard copies, i.e. the user will only see data that has made it fully into the system, and that the system will never henceforth silently lose nor revert.

Dirty reads affect the get, search, as well as indexing (CAS & update) APIs, and are documented in the Elasticsearch reference docs.

This issue is a place holder for addressing dirty reads on the three above-mentioned APIs.

Relates #20031

Metadata

Metadata

Assignees

No one assigned

    Labels

    :Distributed Indexing/DistributedA catch all label for anything in the Distributed Indexing Area. Please avoid if you can.Team:Distributed (Obsolete)Meta label for distributed team (obsolete). Replaced by Distributed Indexing/Coordination.resiliency

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions