AWS: Engine versions and upgrades

We support the latest patch of the current minor release, and of the previous one for 6 months after a new minor ships.

We support the latest patch of the current minor release, and of the previous one for 6 months after a new minor ships.

Today that is KiviDB 1.2, so KiviDB Cloud offers 1.2.2. 1.1 is no longer supported: it gets no more patches, and it has ACL and JSON issues that are fixed only in 1.2.

What you can choose

A new database starts on the latest patch of a supported minor release. Older patches are not offered once a newer one is out: a patch release only fixes bugs and security issues, so there is nothing to stay behind for. When we release a patch, it replaces the previous one in the list. If we withdraw a release, it comes off the list at once.

When a new minor release ships (1.3, say), both 1.3.x and 1.2.x are offered for 6 months, so you have time to test the minor upgrade, which can change how some commands behave. After that, only 1.3.x is offered.

Release candidates are never offered, except to customers taking part in a named preview. Contact us if you want to test one.

Upgrades we do for you

An upgrade restarts the replicas first and the primary last, once the replicas are back in step with it. A database with replicas stays available for reads throughout, and refuses writes for a few seconds while its primary restarts. A database without replicas refuses connections for a few seconds.

A newer patch of your release. When a patch ships (1.2.0 to 1.2.1), we upgrade your database in its next maintenance window at least 48 hours away. The organization's owners get an email with the date first. Patches are cumulative: if a newer patch ships before that window, the same window moves your database straight to the newer one, and the owners get the notice again with the new release and the same date.

A release that is no longer supported. When your database runs a minor release we no longer support, the organization's owners get an email with a deadline at least 30 days away, what to check before upgrading, and a link to the release notes. Upgrade it yourself any time before the deadline. If it still runs the old release then, we upgrade it in its first maintenance window after the deadline, with 48 hours' notice.

To move an upgrade, change the database's maintenance window before it starts. The upgrade moves to the new window, and you get a new notice.

Upgrade yourself

Version in Configuration, Instance settings lists the releases you can move to. Choose one and apply. Downgrades are not supported: take a snapshot before a minor upgrade if you may need the old data back.

Before a minor upgrade, read the release notes' Before you upgrade list. Moving from 1.2.0 to 1.2.2 includes the 1.2.1 changes. The points most applications meet are:

  • ACL categories are Redis's. A user with +@write but not +@connection can no longer SELECT (add +select), and one with +@read, +@keyspace or +@generic but not +@connection can no longer WAIT or WAITAOF (add +wait +waitaof). Our built-in profiles are not affected; check your own ACL rules.
  • WAIT with no replicas waits out its timeout and answers 0, as Redis does, and WAIT is refused on a replica.
  • HELLO says server: redis and version: 8.10.0. Tell KiviDB apart by kividb_version in INFO.

Terraform

kividb_version in a kividb_instance is the oldest release you accept. When we move a database to a newer patch, Terraform shows no change and asks for no downgrade. Set it to a newer release to upgrade.