This page https://docs.vertica.com/12.0.x/en/supported-platforms/server-and-mc/ has a VERY clear warning "Vertica does not support or recommend in-place upgrades from one major version to another major version. For example, you cannot perform in-place upgrades from RHEL/CentOS 7.x to RHEL/CentOS 8.x." but then there is a reference to https://www.vertica.com/kb/UpgradingtheOperatingSystem/Content/BestPractices/UpdateOSinVerticaCluster.htm.
This article specifically recommends in-place upgrades! It says "Upgrade the operating system on all nodes." So far as I can tell, this article was published years ago for vertica 8, and it mentions a vertica.local_verify step which fails as it no longer exists.
So there is a conflict in our docs.
I tested internally upgrading the underlying Linux OS on a Vertica 12 3-node cluster by simply replacing the CentOS 7.9 nodes one at a time with fresh RHEL 8.9 nodes following https://docs.vertica.com/12.0.x/en/admin/managing-db/managing-nodes/replacing-nodes (https://docs.vertica.com/12.0.x/en/admin/managing-db/managing-nodes/replacing-nodes/replacing-host-using-same-name-and-ip-address/). This worked great for me.
The question for Vertica experts is this: Why NOT upgrade a k-safe cluster by replacing each node live using what seems to be simpler and more modern instructions at https://docs.vertica.com/12.0.x/en/admin/managing-db/managing-nodes/replacing-nodes/replacing-host-using-same-name-and-ip-address ? The replacement process works - the node recovers and the cluster stays up - so no need for formal downtime. Should you replace the clearly outdated https://www.vertica.com/kb/UpgradingtheOperatingSystem/Content/BestPractices/UpdateOSinVerticaCluster.htm article with new advice?
