Announcing High Availability for Render PostgreSQL

Stephen Barlow

February 13, 2024

Stephen Barlow

Render takes your infrastructure problems away and gives you a battle-tested, powerful, and cost-effective cloud with an outstanding developer experience. Focus on building your apps, shipping fast, and delighting your customers, and leave your cloud infrastructure to us.

Try Render Free

The Render platform supports all your apps and all their data—with just a few clicks, you can spin up a PostgreSQL database with at-rest encryption, daily backups, and comprehensive extension support. And because that database runs on the same network as your apps, connections are fast, secure, and reliable. Adding to our existing set of database reliability features (like read replicas and point-in-time recovery), today we're happy to announce the full launch of High Availability (HA) for Render PostgreSQL. With just a few more clicks in the Render Dashboard, teams with a Pro instance type or higher can now activate HA for their database:

Following a restart, your database is protected against most causes of extended interruption, reducing their impact to under sixty seconds. (This one-time restart can take a couple minutes, so we recommend flipping the switch during off hours.)

How it works

When you activate HA, Render spins up a second database instance, called a standby. The standby replicates the state of your existing primary instance (asynchronously, on a delay of no more than a few seconds). The standby runs in the same region and network, but on hardware that's physically isolated from the primary. It doesn't receive any incoming queries—it simply keeps itself synced with the primary, ready to jump into action when needed.

Failing over

Even the best built software has its quirks, and so does the hardware it runs on. At some point in the future, your database's primary instance might become unavailable: the underlying node might go down, or the PostgreSQL process itself might crash. When Render's control plane observes that your primary has been unresponsive for 30 seconds, it automatically fails over to the standby by pointing the primary's URL at the standby instead: This swap takes a few seconds, after which the standby instance becomes the new primary! Whenever the unresponsive instance becomes healthy again, it becomes the new standby. Database access is restored in less than a minute, avoiding what might have been a much longer interruption. Render even triggers a failover on your HA database when performing routine maintenance, reducing the associated interruption to less than a minute. Again, failover happens automatically—no action is required from your team. You can trigger a manual failover from your dashboard, but you should only need to do that for your own testing purposes.

Highly available, now generally available

High availability serves as a powerful line of defense against the vast majority of events that cause database downtime. With both HA and point-in-time recovery, you're fully equipped to keep your mission-critical databases healthy and responsive. Learn all the details about HA in the docs, and enable it for your team's Pro and larger databases from the Render Dashboard. For your smaller databases, check out Render's other PostgreSQL reliability features, such as read replicas and daily restorable backups.