Rolling vs. Cluster Upgrades

I'm wanting to better understand when to perform a full-cluster restart upgrade vs. a rolling upgrade of ES. I understand that a rolling upgrade has the benefit of minimizing impact to services during the upgrade (with the exception of disabled allocations and missing shards as each node is upgraded). But what I want to know is whether one of these methods has a lower risk and faster recovery time post-upgrade (getting the cluster back to "green"). Also, is there a best-practice for when to use each method? For instance, should full-cluster upgrades always (or only) be done on major version upgrades, or are rolling upgrades always preferable?

Rolling upgrades are pretty much always preferable, even on major upgrades. They are slower (because you wait for the cluster to stabilise after each step) but they keep the cluster up throughout and you can pause the process at any point to resolve any difficulties you might encounter.

1 Like

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.