- Install kind in this folder. Tested with kind version 0.11.1 .
- Having Docker. Tested with Docker version 19.03.12->20.10.7.
- Have
crds.yaml
from [https://download.elastic.co/downloads/eck/1.8.0/crds.yaml] andoperator.yaml
from [https://download.elastic.co/downloads/eck/1.8.0/operator.yaml] in this folder ( you can usecurl https://download.elastic.co/downloads/eck/1.8.0/operator.yaml -o operator.yaml
andcurl https://download.elastic.co/downloads/eck/1.8.0/crds.yaml -o crds.yaml
to download them).
Tested on MacOS Catalina 10.15.6->11.6
./up.sh
spin up everything
.downn.sh
destroy everything (in fact only the kind cluster ;-) )
- What is inside the kind cluster do not support docker to be restarted (some network stuff).
The stack-monitoring.yaml
is based on ECK beats stack monitoring doc examples but it has been tuned to use k8s annotations and k8s labels to only monitor the stack elements.
- update the
up.sh
to use the operator 1.3.0 up.sh
- annotate some ressources to exclude them from the upgrade (in fact not managed by the operator) like for example
kubectl annotate --overwrite elasticsearch elasticsearch 'eck.k8s.elastic.co/managed=false'
kubectl apply -f eck-operator-1.5.yaml
=> all the non elasticsearch pods are restarted! some are crashlooping! but they get back after the kibana are restarted?- allow the elasticsearch to be managed by ECK:
kubectl annotate --overwrite elasticsearch elasticsearch 'eck.k8s.elastic.co/managed=true'