What process decides when a Repository failover is required?

In a high-availability (HA) Tableau Server setup, the Repository (PostgreSQL) has an active and passive instance. Failover occurs if the active Repository fails. Let’s dive into the process:

HA Setup:

Two Repository instances across nodes (active/passive).

Failover switches to the passive instance if the active one becomes unavailable (e.g., crash, network issue).

Cluster Controller:

Role: Monitors all processes (e.g., Repository, File Store) across nodes, detecting failures via heartbeats and status checks.

Failover Decision: If the active Repository stops responding, Cluster Controller initiates failover, promoting the passive instance to active.

Coordination: Works with Coordination Service (ZooKeeper) to update topology but makes the initial detection call.

Option A (Cluster Controller): Correct.

Why: It’s the watchdog process, constantly monitoring Repository health and triggering failover when needed.

Option B (Coordination Service): Incorrect.

Role: ZooKeeper maintains cluster state and coordinates topology updates post-failover, but doesn’t detect the failure—Cluster Controller does.

Option C (Gateway): Incorrect.

Role: Routes client requests—unrelated to internal process monitoring or failover.

Option D (Backgrounder): Incorrect.

Role: Executes background tasks—no involvement in Repository failover decisions.

Why This Matters: Understanding failover ensures HA reliability—Cluster Controller is the linchpin for resilience.