The legacy mongo shell client currently forces listIndexes to run against the primary, even when the session is causally consistent and the configured read preference is secondary. This reduces listIndexes test coverage, since it cannot be properly exercised on causally consistent sessions with a secondary read preference. On the server side, support has always been present: listIndexes accepts afterClusterTime at readConcern level local.

The goal of this task is to allow listIndexes to target secondary nodes in causally consistent sessions with a secondary read preference.

Separate task: verify real-world MongoDB driver behavior

The production drivers live in separate codebases from the legacy mongo shell. We do not yet know whether each driver already attaches afterClusterTime to listIndexes in causally consistent sessions, or whether they hard-code this command to the primary (as the shell does today). We should verify each driver’s behavior and file new Jira tickets where changes are needed.