Managed Prometheus deployment installs a lot of extra stuff
It looks like we are unintentionally installing a lot of extra items as part of the Prometheus deployment from the cluster page:
- AlertManager
- kube-state-metrics
- Pushgateway
- NodeExporter DaemonSet
The values.yaml
we are passing appears to disable these:
alertmanager: |
enabled: false
kubeStateMetrics: |
enabled: 'false'
nodeExporter: |
enabled: 'false'
pushgateway: |
enabled: 'false'
Also deployed from the Helm chart is a sidecar container running with the Prometheus server, jimmidyson/configmap-reload
.
I don't really mind AlertManager and kube-state-metrics, as we will ultimately want those deployed. So even though we don't make use of them today, we will in the future. Similarly, the configmap-reload
will also be helpful for 10.5/10.6, once we start hopefully setting alerts in the prometheus.yml
.
My main concern is whether we really want to deploy a DaemonSet NodeExporter and the Pushgateway. NodeExporter in particular could be a surprise, as you wouldn't necessarily expect that to be deployed along with Prometheus, and our documentation doesn't indicate that it will be.