Very heavy Momory usage on Couchbase cluster Node

We are using the CB with Sync_gateway.
We have a CB cluster with 2 nodes, if we have 16 GB RAM for each node and 200 GB disc size on each node.

Mobile devices are adding records to the CB through sync_gateway.
It is observed that RAM usage of one server node (with whome Sync_gateway is connected)
is increasing sharply and RAM usage of another server node is relatively low, say, on node 1 the RAM usage is 80% and on other node it is around 30%.

I have set the mem_high_vat value at 70% on each node but still it is showing more than 80% RAM usage for one of the Nodes, and there is considerable difference between the RAM usage on 2 nodes.

Also not all data is moving to disc even if we have set mem_high_vat value at 70%.

Will it bring down the cluster (or a node with high memory usage)?
How to fix this?
Is there any mechanism in Couchbase with which one can forcefully send data from RAM to disc ?

Couchbase Admin console ‘Server Nodes’ tab it show information about RAM usage, CPU percentage usage etc. for each node.
Here which RAM it is considering for display, the RAM of the VM or RAM assigned to the Couchbase bucket ?


Sounds like your cramming alot into those two nodes.
What is your active resident ratio?
Are you running all the services on each node DATA,QUERY & INDEX?

Thanks @househippo,

The active resident ratio is 100 % i.e. all doc required are in RAM.

I am using DATA,QUERY & INDEX services on both nodes.
What should be the preferred way?
Shall I need to separate it ?
If yes how many nodes should run Data service, how many Index service and how many for Query service.
Is there any guideline documentation for that,
I mean for how much data how many Data nodes required and for how many Data nodes how many index nodes required .

Your cluster configuration really depends on a lot of things

  • budget (as more nodes could mean more licenses for Enterprise)
  • your data replication requirements
  • how much data and expected growth
  • how many indexes and size of them
  • number and types of queries

Would suggest reading these to help you with sizing guidelines