Skip to content

Synchronized segment allocation #6087

Closed

Description

@gianm

Today, when SegmentAllocateAction is run by multiple threads at once, it can end up doing a lot of internal retries due to its SELECT -> INSERT logic (see allocatePendingSegment in IndexerSQLMetadataStorageCoordinator). There isn't much point in running concurrent allocations for the same datasource anyway, and the contention causes retries which add delay due to sleeps. It might work better to synchronize allocations on a per-datasource basis, so only one can happen at a time per datasource.

Activity

  1. jihoonson commented on Oct 25, 2018

    @jihoonson
    Contributor

    Fixed in #6336.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions