Skip to content

Track number of non-user offsets used in chunk header #99

@kjnilsson

Description

@kjnilsson

One requested feature that we cannot currently provide is the total number of messages in a stream (or "user entries"). The reason that simply taking the first offset in a stream and subtracting this from the last offset is that there are non-user offsets used for internal purposes such as offset and writer tracking. To accurately calculate the number of user entries we'd need to know the number of non-user entries in a stream.

The chunk header has 4 bytes of reserved data which could be used for a "non-user offset" counter. If we had this information we could calculate the number of messages in a stream by doing: (LastOffset - FirstOffset) - (LastNonUserCount - FirstNonUserCount) - which only requires us to read the chunk header of the first and last chunks of the stream.

Potential issues:

  • for existing streams the count would not be correct until we'd flushed out all segments that do not have this information
  • The non-user entry counter is limited to 32 bits, we'd have to consider if it is likely to ever reach this value and what to do if it does.

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