Strimzi
-
|
Hi guys, We procrastinated updating the Kafka Strimzi operator on our K8s cluster, so I want to understand how to migrate from an older operator to a newer one (skipping various versions) and then switch to a new Kafka version, considering compatibility loss along the way. Based on the "Supported versions" table from https://strimzi.io/downloads/, let's suppose I need to jump from operator 0.44.0 to the latest 0.47.0. The problem is, as seen in the table, 0.44.0 only supports Kafka versions [3.7.0, 3.7.1, 3.8.0], and 0.47.0 supports [3.9.0, 3.9.1, 4.0.0]. So the highest I can go with operator 0.44.0 is 3.8.0, but the lowest supported in 0.47.0 is 3.9.0, leaving a gap. The question is, in this case, whether it is recommended to go directly to operator 0.47.0 or if it is better to transition through an intermediate step to start on a Kafka-accepted range on both operators, allowing for a smoother transition but with the cost of a longer and more complicated production maintenance window. BTW, in my actual (real) situation, I'm on an even lower version of the operator, so the "smooth" transition option will be much slower. Thanks in advance! |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
|
When you upgrade to a newer version (with a big jump) which doesn't have a Kafka version overlapping with the old one, the drawback you have is that the operator won't operate your cluster until you upgrade the cluster itself to a supported version. It doesn't mean that the Kafka cluster doesn't work. It will still be operational and usable by clients but not operated by Strimzi. I think it depends on your needs. As you said, going through more versions is much slower taking into account that anyway the new operator version is going to roll the Kafka cluster pods anyway (even if the Kafka version is the same, the images will be different). |
Beta Was this translation helpful? Give feedback.
When you upgrade to a newer version (with a big jump) which doesn't have a Kafka version overlapping with the old one, the drawback you have is that the operator won't operate your cluster until you upgrade the cluster itself to a supported version. It doesn't mean that the Kafka cluster doesn't work. It will still be operational and usable by clients but not operated by Strimzi. I think it depends on your needs. As you said, going through more versions is much slower taking into account that anyway the new operator version is going to roll the Kafka cluster pods anyway (even if the Kafka version is the same, the images will be different).