I deployed Weaviate on a Kubernetes cluster using a Helm chart deployment. The cluster has 2 nodes and 6 pods, which allows Weaviate to manage sharding and replication. However, I have a problem: when one of the pods becomes unhealthy (ressources reason for exemple), the connection to Weaviate remains active so that’s good, but data indexing fails on this unhealthy pod. What’s more, the data is not automatically redirected to the other healthy pods. How can I solve this problem to ensure that indexing is redirected to functional pods ?
hi @mathieu ! The number of nodes shouldn’t matter here as it is the number of pods that will make it resilient or not.With that said, and considering that your cluster has 6 pods, are you creating the collection with the correct replication factor?
Check here more about replication, specially this part