Summary

Replication is considering allowing secondaries to hide themselves if their replication lag exceeds a given threshold. This ticket is to verify that an in-flight (read) query on a secondary that hides itself will not be disrupted.

Motivation

Who is the affected end user?

Some customers read from secondaries, but they don't want to read from lagged secondaries.

How does this affect the end user?

Customers don't like reading from lagged secondaries, and read load is extra load that can prolong the period before the secondary catches up.

How likely is it that this problem or use case will occur?

Customers periodically complain about reading from lagged secondaries.

If the problem does occur, what are the consequences and how severe are they?

This prolongs the period of replag in the best case and causes queries to stall in the worst case, such as when the customer specifies afterClusterTime.

Is this issue urgent?

No.

Is this ticket required by a downstream team?

No.

Is this ticket only for tests?

Maybe. It's possible that tests already exist for this case.

Acceptance Criteria

This ticket is specifically to understand the current behavior. If the current behavior would not pair well with auto-hiding secondaries, then we might file a follow-up ticket to change the behavior.