Strange Memory Usage

server

#1

Hello there,
I’m using Couchbase Server 4.1 CE.
I have a 3 machine cluster with 32Gb of ram per machine.
I have 4 buckets with 4M, 3M, 10M, 6M documents.
I have 4 indexes one for each bucket.

I have setup 18Gb ram for Data and 10GB for indexes per node.

On dashboard i see this for each machine (give or take a few MB):

It says that is only using 8Gb and then 20GB is being used for other data.
The server is only running couchbase server, so can couchbase be using the 20Gb of ram to do other processing?
My N1QL queries are running slow when this state is reached and i have do restart the machines.

How can i find out whats happening here.

Thanks


#2

@rafael.felix,
it’s a normal behavior for Linux, see http://www.linuxatemyram.com
Also take a look at Endless compaction process causes memory leak (potential), you can find there an example with 128G-CB-server
P.S. Don’t thank me, thank @drigby, check out the irony and don’t get offended when you see

There are no downsides, except for confusing newbies

while reading http://www.linuxatemyram.com

:wink:


#3

Also, note that there are some specific memory related tuneables recommended for Couchbase on Linux. It’s partially related to your observation of memory usage and could be related to some of the observed slowness.


#4

Thanks for the tips.