Reputation: 409
A question to people having a Cassandra Cluster in production: Until when do you treat a Cassandra node being healthy from a CPU usage POV? E.g. some sort of average CPU usage threshold (or Unix load) for a certain time period?
We are currently running a 6-node test cluster and apply some sort of our load patterns and we are seeing e.g. an average CPU usage of ~ 45-60% for the last day on each node. No scheduled/explicit read repair during that time. Is this already asking for extending the cluster with additional nodes or can this be considered as being healthy?
Upvotes: 1
Views: 683
Reputation: 19377
I would start to seriously consider expanding the cluster when my peak cpu (or disk i/o) hits 80% of capacity for more than a few seconds. So if your load is very constant you're probably fine, but looking at just the daily average isn't enough detail to say for sure.
Upvotes: 1