EC2 Couchbase Server 2.0 Instance
This is more of a general question:
We originally set up a cluster using the official CouchBase 1.8.3 image and upgraded all 3 nodes to 2.0.
We have noted high RAM usage on 1 of the 3 nodes we have in a cluster. Manually initiatign a failing-over a node and re-balancing, results in higher RAM usage on another of the 3 nodes.
The biggest difference we have seen thus far is 99.4% RAM usage on one node, whilst the other two are at 50%
After a failover/rebalance i'm seeing a more even RAM usage pattern though 1 node RAM usage is growing much faster than the other two...we were expecting a more even\similar RAM usage among the nodes.
My question, what could cause the RAM usage ratios among the nodes to become skewed, should we be looking more closely at the .net clients that access the cluster?
Some background information:
Each node has 15 GB of RAM, with a quota 10 GB per node for our Couchbase test bucket with 1 replica. Ops per sec on the node with the most RAM usage is ~25 ops per sec spikes, average ops per second on the bucket is 6 roughly 14 K objects in the bucket.