Struct mz_storage::source::mysql::replication::partitions::GtidReplicationPartitions

pub(super) struct GtidReplicationPartitions {
    active: BTreeMap<Uuid, GtidPartition>,
    future: Vec<GtidPartition>,
}

Holds the active and future GTID partitions that represent the complete UUID range of all possible GTID source-ids from a MySQL server. The active partitions are all singleton partitions representing a single source-id timestamp, and the future partitions represent the missing UUID ranges that we have not yet seen and are held at timestamp GtidState::Absent.

This is used to keep track of all partitions and is updated as we receive new GTID updates from the server, and is used to create a full ‘frontier’ representing the current state of all GTID partitions that we can use to downgrade capabilities for the source.

We could instead mint capabilities for each individual partition and advance each partition as a frontier separately using its own capabilities, but since this is used inside a fallible operator if we ever hit any errors then all newly minted capabilities would be dropped by the async runtime and we might lose the ability to send any more data. However if we just use the main capabilities provided to the operator, the capabilities externally will be preserved even in the case of an error in the operator, which is why we just manage a single frontier and capability set.

Fields

active: BTreeMap<Uuid, GtidPartition>

future: Vec<GtidPartition>

Implementations

impl GtidReplicationPartitions

pub(super) fn frontier(&self) -> Antichain<GtidPartition>

Return an Antichain for the frontier composed of all the active and future GTID partitions.

pub(super) fn advance_frontier( &mut self, new_part: GtidPartition, ) -> Result<(), DefiniteError>

Given a singleton GTID partition, update the timestamp of the existing active partition with the same UUID or split the future partitions to remove this new partition and then insert the new ‘active’ partition.

This is used whenever we receive a GTID update from the server and need to update our state keeping track of all the active and future GTID partition timestamps. This call should usually be followed up by downgrading capabilities using the frontier returned by self.frontier()

Trait Implementations

impl From<Antichain<Partitioned<Uuid, GtidState>>> for GtidReplicationPartitions

fn from(frontier: Antichain<GtidPartition>) -> Self

Converts to this type from the input type.