|
The changes from SERVER-55305 will address how a retryable write statement may execute multiple times without making atClusterTime reads on the config.transactions collection correct for all readTimestamps. To prevent accidental misuse of atClusterTime reads, they should be rejected on the config.transactions collection.
Note however that resharding depends on performing an atClusterTime read on the config.transactions collection (see SERVER-55214) and will need an undocumented option (or some internal privilege way) of continuing to use atClusterTime reads on the config.transactions collection.
|