Elastic AutoOps is a monitoring tool that collects metrics from your local Elasticsearch installation. This data is sent to your user space within Elastic Cloud and can tell you a lot about your installation, the health of the cluster, and how to improve it.
The AutoOps monitoring tool is implemented as an Elastic agent instance. But a NetEye node usually already has an Elastic agent running, and normally you cannot install two Elastic agents in parallel.
One solution is having Elastic AutoOps in a container, which enables independent deployment, makes upgrades predictable, and reduces complexity with other software on the host.
You could run AutoOps in a Docker environment or other similar methods (see deployment options), but I would suggest using Kubernetes, since it’s already part of NetEye: In fact, NetEye 4.49 was the first version to include Kubernetes as part of the product.
In short: Containerizing AutoOps on NetEye with Kubernetes isolates dependencies, simplifies version changes, and aligns the deployment with NetEye native operations.
This blog post also describes very practical settings regarding:
The main advantage is not only “running AutoOps in Kubernetes,” but doing so in a way that is supportable, versioned, and easy to reproduce on NetEye installations.
If your environment requires proxy access, you first need to configure the RKE2/containerd proxy settings on the target node before deploying AutoOps. Without that, the pod may fail to pull the image, or to reach required endpoints.
Add the proxy configuration to /neteye/local/rke2/conf/sysconfig/rke2-user-customization (and fill in “…” with appropriate values).
CONTAINERD_HTTP_PROXY=...
CONTAINERD_HTTPS_PROXY=...
CONTAINERD_NO_PROXY=...
Then restart the service:
systemctl restart rke2-server.service
helm repo add elastic https://helm.elastic.co
helm repo update
kubectl create namespace autoops-agents
kubectl create secret generic npe-elasticsearch-dev-es-http-certs-public --from-file=ca.crt=/root/security/ca/root-ca.crt -n autoops-agents
See: Connect your ECE, ECK, or self-managed cluster to AutoOps | Elastic Docs
Connect to your self-managed cluster:

…click on “install AutoOps agent”:

… and choose “Kubernetes”:

Now define the Elasticsearch endpoint URL:

And take the information and overwrite it into the code block below (IMPORTANT: don’t close the browser tab)

Using the values provided by Elastic Cloud AutoOps, fill in “…” with appropriate values:
kubectl create secret generic autoops-secret \
--from-literal=autoops-token='...' \
--from-literal=autoops-es-url='https://elasticsearch.neteyelocal:9200' \
--from-literal=temp-resource-id='...' \
--from-literal=autoops-otel-url='https://otel-auto-ops.us-east-2.aws.svc.elastic.cloud' \
--from-literal=cloud-connected-mode-api-key='essu_...' \
--from-literal=es-api-key='...' \
-n autoops-agents
… in order to let the Elastic agent communicate with Elastic Cloud (and fill “…” with appropriate values):
kubectl create secret generic autoops-proxy \
--from-literal=http_proxy='...' \
--from-literal=https_proxy='...' \
--from-literal=no_proxy='...' \
-n autoops-agents
apiVersion: apps/v1
kind: Deployment
metadata:
name: autoops-elastic-agent
namespace: autoops-agents
labels:
app: autoops-elastic-agent
app.kubernetes.io/component: autoops
app.kubernetes.io/name: autoops-elastic-agent
app.kubernetes.io/version: 9.4.5
spec:
progressDeadlineSeconds: 600
replicas: 1
revisionHistoryLimit: 10
selector:
matchLabels:
app: autoops-elastic-agent
app.kubernetes.io/instance: autoops-elastic-agent
app.kubernetes.io/name: autoops-elastic-agent
strategy:
rollingUpdate:
maxSurge: 25%
maxUnavailable: 25%
type: RollingUpdate
template:
metadata:
labels:
app: autoops-elastic-agent
app.kubernetes.io/component: autoops
app.kubernetes.io/instance: autoops-elastic-agent
app.kubernetes.io/name: autoops-elastic-agent
spec:
containers:
- args:
- "--config"
- "otel_samples/autoops_es.yml"
- "--set"
- 'receivers::metricbeatreceiver::metricbeat::modules=[{module: autoops_es, hosts: "${env:AUTOOPS_ES_URL}", period: 10s, metricsets: [cat_shards, cluster_health, cluster_settings, license, node_stats, tasks_management], ssl: { certificate_authorities: ["${env:AUTOOPS_CA_CERT}"] } }, {module: autoops_es, hosts: "${env:AUTOOPS_ES_URL}", period: 24h, metricsets: [cat_template, component_template, index_template], ssl: { certificate_authorities: ["${env:AUTOOPS_CA_CERT}"] } }]'
env:
- name: AUTOOPS_TOKEN
valueFrom:
secretKeyRef:
name: autoops-secret
key: autoops-token
- name: AUTOOPS_ES_URL
valueFrom:
secretKeyRef:
name: autoops-secret
key: autoops-es-url
- name: AUTOOPS_TEMP_RESOURCE_ID
valueFrom:
secretKeyRef:
name: autoops-secret
key: temp-resource-id
- name: AUTOOPS_OTEL_URL
valueFrom:
secretKeyRef:
name: autoops-secret
key: autoops-otel-url
- name: AUTOOPS_ES_USERNAME
valueFrom:
secretKeyRef:
name: autoops-secret
key: es-username
optional: true
- name: AUTOOPS_ES_PASSWORD
valueFrom:
secretKeyRef:
name: autoops-secret
key: es-password
optional: true
- name: AUTOOPS_ES_API_KEY
valueFrom:
secretKeyRef:
name: autoops-secret
key: es-api-key
optional: true
- name: ELASTIC_CLOUD_CONNECTED_MODE_API_KEY
valueFrom:
secretKeyRef:
name: autoops-secret
key: cloud-connected-mode-api-key
optional: true
- name: ELASTIC_CLOUD_CONNECTED_MODE_API_URL
valueFrom:
secretKeyRef:
name: autoops-secret
key: cloud-connected-mode-api-url
optional: true
- name: http_proxy
valueFrom:
secretKeyRef:
name: autoops-proxy
key: http_proxy
- name: https_proxy
valueFrom:
secretKeyRef:
name: autoops-proxy
key: https_proxy
- name: no_proxy
valueFrom:
secretKeyRef:
name: autoops-proxy
key: no_proxy
- name: AUTOOPS_CA_CERT
value: /etc/ssl/certs/es-ca/ca.crt
volumeMounts:
- name: es-ca-certs
mountPath: /etc/ssl/certs/es-ca
readOnly: true
image: >-
docker.elastic.co/elastic-agent/elastic-otel-collector-wolfi:9.4.5
imagePullPolicy: Always
name: autoops-elastic-agent
resources: {}
terminationMessagePath: /dev/termination-log
terminationMessagePolicy: File
volumes:
- name: es-ca-certs
secret:
secretName: npe-elasticsearch-dev-es-http-certs-public
items:
- key: ca.crt
path: ca.crt
dnsPolicy: ClusterFirst
restartPolicy: Always
schedulerName: default-scheduler
securityContext: {}
terminationGracePeriodSeconds: 30
kubectl apply -f enroll-autoops-kubernetes-via-helm.yaml


Another strong reason to run AutoOps in a container is how simple upgrading becomes. Instead of reworking a host installation, you only replace the deployment version and re-apply it.
kubectl delete -f enroll-autoops-kubernetes-via-helm.yaml
Now apply the deployment again:
kubectl apply -f enroll-autoops-kubernetes-via-helm.yaml
That gives consultants and operations teams a low-friction upgrade path with very little procedural overhead.
Here are some practical tips for consultants:
For NetEye installations, this approach strikes the right balance between control and maintainability. It respects (1) network constraints such as proxies and (2) product constraints such as internal trust chains and existing container setup, while still giving teams a modern, Kubernetes-native way to operate AutoOps.
Did you find this article interesting? Does it match your skill set? Our customers often present us with problems that need customized solutions. In fact, we’re currently hiring for roles just like this and others here at Würth IT Italy.