the “shard is now inactive” message is printed when there haven’t been any newly written records for specific shards for a certain time. The message itself is not problematic and can occur during regular operations:
Whenever there are any performance issues, it is advisable to check metrics accessible through the Crate JMX Exporter as described below:
From our experience, 64 GB RAM sounds relatively low in relation to the 32 vCPU cores. The JMX Exporter gives access to metrics that help identify bottlenecks. For example, GC rates and heap usage (circuit breaker utilization) can indicate if the cluster is running low on the available heap.
Therefore, if you haven’t done so already, I recommend getting the JMX Exporter set up to gain access to application-level performance metrics.