I'm planing resources which will be dedicated to run Kafka Connect Clusters in Strimzi Distribution.
My workloads are pretty small, there will be probably only 1 task per connector needed.
I would go with AKS with 2 nodes in my dedicated nodepool.
In that case should i always deploy Strimzi Kafka Connect cluster with 2 workers (pods) inside my AKS Cluster, just to achieve HA? Or just one Worker will be okay and if the node will fail, this worker (pod) will be recreated on second node with the task? Can you share some good practices of resource planning for Strimzi Kafka Connect Clusters? - When i should go with two workers etc.
One worker would be fine. If that pod terminates, AKS would reschedule it. If a task fails, however, that traditionally is a manual restart process (not sure if Strimzi has a solution to this), and so multiple workers would not address this.
If you start running many connector tasks, you may want to scale out to multiple workers, but unless you setup affinity rules on the pods, the worker pods may be on only one node, so not truly addressing HA