November 20, 2026 by Rene Cannao · Release

Keeping database failover ready with Orchestrator v4.32.1

Your application can no longer write to the database. The primary is unreachable, but the replicas are still up. Before promoting one, you need to know which replicas are healthy, how far they’ve caught up, and which other servers will need to follow the new primary. Those are questions you’d rather answer before an outage.

Orchestrator is an open-source tool for managing that work. It discovers your replication topology, tracks which servers are healthy and how they’re connected, and gives you a live view of the cluster. You can use it for a planned switchover during maintenance or configure it to recover automatically when a primary fails. During a recovery, Orchestrator can select and promote a replica and reconfigure replication around the new primary. With ProxySQL integration configured, it can also update where the proxy sends writes. Orchestrator started with MySQL, and the version we maintain at ProxySQL also supports PostgreSQL streaming replication. Read about Orchestrator’s move to ProxySQL.

We’ve released Orchestrator v4.32.1. Two fixes in this release make a difference in everyday operations. One helps Orchestrator itself come back after a restart. The other gives operators a more accurate view of cascading MySQL replication.

Who takes over when Orchestrator goes down?

If a single Orchestrator process handles recovery for your databases, its host going down can leave you without automatic failover at the moment you need it. You can run Orchestrator on several hosts, but those instances need a way to agree on which one is in charge. That’s what Raft does. Orchestrator instances use it to elect a leader and share operating state, including instructions about planned downtime and information about ongoing failovers.

In a three-member deployment, two members make a majority, called a quorum. If one host goes down or loses contact with its peers, the two that can still reach each other can elect a leader and keep coordinating database recovery. An isolated member lacks the votes to lead on its own. Read the Orchestrator Raft deployment guide.

That setup depends on members coming back after a reboot or maintenance. In earlier versions, a restarting member could hit a race and crash while loading its saved Raft state. v4.32.1 initializes the caches Orchestrator uses to track database instances before restoring Raft state. If you run Orchestrator with Raft, upgrade so that a routine restart can return the member to service without hitting this bug. Thanks to Dave Nusbaum for reporting the problem and contributing the fix. Read about the startup fix.

Which transactions actually came from the primary?

In a cascading MySQL setup, some replicas receive transactions through an intermediate server. Orchestrator uses global transaction identifiers (GTIDs) to help spot transactions that originated outside the expected replication history. An unexpected transaction can point to a write made directly on a replica, something worth investigating before you promote it.

If Orchestrator could see the intermediate server but couldn’t connect directly to the original source, transactions from that source could look out of place on replicas farther down the chain. v4.32.1 uses the upstream identity reported by the intermediate server to recognize where those transactions came from. If you run cascading replication, the upgrade makes these alerts easier to act on by reducing false positives while still flagging writes made directly on a downstream replica. Read about the cascading replication fix.

Get Orchestrator v4.32.1

The release includes Linux downloads for AMD64 and ARM64. If you use RPMs, you can now install the Orchestrator server and CLI packages on the same host. You’ll find signed RPMs and a signed checksum manifest on the release page. Download Orchestrator v4.32.1 and read the release notes.

If you’re new to Orchestrator, start by discovering a test topology and opening the web UI. Seeing which servers follow the primary and how that changes during a planned switchover is a practical way to get familiar with what Orchestrator will manage during an outage. Read the installation guide.