Varying RAM usage and CPU usage

Hi,

We have three nodes in our cluster with the following specs:
8GB RAM (6GB allocated to Couchbase)
Intel Xeon CPU E3-1220 V2 @ 3.1GHz
2 machines with 1TB HDD, 1 machine with 500GB
The 1TB machines have Ubuntu 12.04 and the 500GB has Debian 6.

With all three machines very similar with only Couchbase running on them, I wouldn’t expect such varying levels of RAM and CPU usage.

If we say Node A and Node B are the Ubuntu machines and Node C is the Debian machine this will be easier to explain and follow.

Node A currently has RAM usage of 84.3% and never goes above 90%. Node A does however have spikes of CPU usage and reached 64.4% yesterday.

Node B currently has 91.2% RAM usage and regularly fluctuates and reached 95.1% yesterday. We never see spikes in CPU usage on this node.

Node C currently has 97.7% RAM usage and regularly fluctuates and reached 96.4% yesterday and as you can see is already higher than that now. We never see spikes in CPU usage on this node either.

Looking at the Nodes overview on the console, they’ve all got pretty much the same number of active/replica, and the one with the most is Node A but only marginally.

Our total number of documents across 8 buckets is 178796 but with ops per second currently at less than 100 but with occasional spikes.

I’m not having any difficulties or issues as a result of these things but they’re worrying nevertheless as I don’t understand how they can all be so different and if there is something I can/need to do as a preventative measure so they don’t cause issues in the future then I’d appreciate any advice.

If you need any more information please ask.

Thanks,

Graeme

Hello,

Can you confirm that the cluster is balanced from a data point of view?
(when you look at the different nodes and see the number of items for each node)

Regards
Tug
@tgrall

Hi Tug,

When I click on the node and see the graphs, the items graph shows 36k items for the two Ubuntu machines and 35.7k for the Debian machine.

Thanks,

Graeme

Hello,

Having a cluster made of different OS is not a supported configuration. (as we do not test it)

This could explain the difference.

Regards
Tug
@tgrall