Kubernetes module 6.2.4 vs 6.4.0 : cpu usage x3


(Hamelg) #1

We have a pod running the module kubernetes in our openshift cluster. We had upgraded metricbeat from 6.2.4 to 6.4.0. We have noticed that metricbeat 6.4.0 consummates 3 times cpu than 6.2.4.
Here is our the metricbeat config. Is there any reason ?

====
- module: kubernetes
enabled: true
metricsets:
- state_node
- state_deployment
- state_replicaset
- state_pod
- event
period: 1m
hosts: ["kube-state-metrics:8080"]


(Carlos Pérez Aradros) #2

Hi @hamelg,

We introduced automatic enriching for all Kubernetes metricsets. That means that every monitored resource events (for instance, replicaset metrics) get enriched with labels and annotations from that resource.

You can disable that behavior by setting add_metadata: false in the module settings.

Best regards


(Hamelg) #3

Thank you.
I tried add_metadata: false, it removes label et annotations but it makes no difference on the cpu usage.


(Willemdh) #4

What about 6.3.2? (which I'm using currently on Openshift)


(Carlos Pérez Aradros) #5

in addition to that, could you provide full configuration + your measures? I've reviewed and we didn't add many things apart from the add_metadata.

Best regards


(Hamelg) #6

Sorry for the delay. I'll give you a complete feedback when we'll upgrade our RUN cluster. At now, we seen that on our BUILD cluster, now in 6.4 and I don't want downgrade in 6.2.

Our BUILD cluster is running 1300 pods.

- module: kubernetes
  enabled: true
  metricsets:
    - state_node
    - state_deployment
    - state_replicaset
    - state_pod
    - event
  period: 1m


(system) #7

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.