Skip to content

Research issues with data order throughout the system #299

@CGodiksen

Description

@CGodiksen

Due to how ingested data is sorted and how compressed data is sorted both when writing to disk and when reconstructing the data in the query engine, issues related to the order of the data might happen.

One example of this is if out-of-order data points are added to the storage engine which then sorts and compresses them. This could cause overlapping compressed segments which could cause issues during querying. It is not certain that this causes issues but the problem should be researched to ensure this does not cause issues. A discussion on this was started here: #293 (comment).

Another potential issue is the disconnect between what Apache DataFusion expects from the ordering of the output from SortedJoinExec and what we currently provide. We currently only guarantee that the order is the same for each field column. It might be necessary to research what impact setting the sort guarantees and requirements has on the execution in Apache DataFusion and considering whether it would be better to remove the currently set guarantees. A discussion on this was started here: #293 (comment).

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions