The discussion revolves around skipping the Redis version check during a Discourse rebuild. NotAnonymous asks if it’s possible to bypass the requirement for Redis 6.2.0 or above, as their external Redis server is on version 6.0.14. They express confidence that the older version won’t cause functional issues, referencing a prior post suggesting the requirement is based on testing rather than necessity. pfaffman initially questions the rationale, but Falco clarifies that a new Redis command introduced in 6.2.0 is used for improved performance in high-traffic scenarios. This command is not available in older versions, and the requirement was added to ensure stability and performance. While the team is considering whether the requirement is too strict, NotAnonymous appreciates the explanation and acknowledges the need for a newer Redis version, though they suggest flexibility for sites unable to upgrade immediately.
Is it possible to skip the redis version check on a rebuild? Rebuild started failing on the most recent version of Discourse and the only error I’m seeing is that Discourse requires Redis 6.2.0 or above. We’re using an external redis server that’s on 6.0.14. I’d be surprised if this caused an issue with Discourse functionality, so I’d like to be able to still update Discourse until we are able to upgrade Redis.
I did also try rebuilding without using external Redis but got different errors then, depending on whether I included the redis template.
I saw a previous post where it was said/implied that the requirement is just because that’s what they use to dev/test on, so anything less they can’t guarantee will work. I’m willing to take that risk or at least test it out if possible. I’m also looking at speeding up our upgrade though.
Hmm. That does sound familiar. But this requirement was added recently, so I’m thinking that there is likely a reason. But I can’t search commits very well on my phone, so we’ll see what else we hear!
We are using a new Redis command that is only available on Redis >= 6.2.0. We ship that Redis version in our base image for 11 months, and it’s also available on AWS Elasticache service, or as the official Redis docker image.
The change was made so we can get better performance in high traffic situations, where Redis becomes a bottleneck for some large Discourse instances.
That said, we are discussing internally if that is too aggressive of a requirement and may revert that performance patch to accommodate sites stuck on an old version.
