High Memory Usage of Netdata

Help
1

Suggested template:

Problem/Question

I am running netdata on several nodes of a docker swarm. On all nodes netdata uses between 60 an 120 MB, but on the manager node I have a consumption of 1.5 GB. After doing a container restart it went down to 500MB, not sure if it will stay there. I wonder what makes this high memory consumption and what to do about it.

Here a graph of the worker node:

photo_2024-01-11_16-31-09
photo_2024-01-11_16-31-091280×720 131 KB

and here the change with the update:

photo_2024-01-11_16-34-37
photo_2024-01-11_16-34-371280×720 140 KB

Environment/Browser/Agent’s version etc

v 1.44.1 installed via docker

2

Hello @Martin_Neumann ,

Is the “manager node” centralizing data from other netdada? If this is the scenario, this means that you are replicating data from other children, and it is using memory to store these data.
Another possibility to have a giant usage would be number of collected metrics to be much higher on this host.
Are all hosts running the same netdata version?
We have more details about the memory usage in our documentation Change how long Netdata stores metrics | Learn Netdata.

Best regards,

Thiago

3

No the manager has this role in the docker swarm, what Netdata is concerned, it is collecting only the metrics of its own, not of the other nodes.

The configuration is default without any specific alterations, so I do not understand the reason why it should be using this exaggerated amount of RAM. I have actually another manager that is running more containers and is using only 169 MB. I wanted to explore a bit more what could be the cause, but what I see in the new version the netdata charts were removed. Maybe that helps to reduce memory consumption, but limits me a little bit here.

Yes I have updated all nodes to the newest version.

4

Hello @Martin_Neumann ,

You still can get the number of charts and metrics using our API, please, access http://localhost:19999/api/v1/charts?all and you will have two JSON fields:

  • charts_count: Shows the number of charts you have.
  • dimension_count: Shows the number of metrics.

While I was getting data to answer you, I also remember that we modified recently the way we are showing data for apps plugin and eBPF plugin, instead of show one chart with a dimension per application, now we are showing a chart with a dimension for every application group.

Best regards!