Kubernetes Host/Node Requirements

Worker Hosts

Hosts that will be used for Kubernetes can only be used for Kubernetes clusters.

All Kubernetes hosts must conform to the requirements listed in the following:

Furthermore, if you will be running any add-ons, then the Kubernetes hosts must also comply with the cumulative requirements for all of the add-ons you will be running. See Add-ons Overview for a list of current add-ons and links to additional details. For example:

  • If you will not be running any add-ons, then the Kubernetes hosts need only comply with the base requirements.
  • If you will be running Add-on_1, then the Kubernetes hosts must comply with the base requirements plus the requirements for Add-on_1.
  • If you will be running Add-on_2, then the Kubernetes hosts must comply with the base requirements plus the requirements for Add-on_2.
  • If you will be running Add-on_1 and Add-on_3, then the Kubernetes hosts must comply with the base requirements plus the requirements for Add-on_1 plus the requirements for Add-on_3.
  • If you will be running Add-on_1 and Add-on_2 and Add-on_3, then the Kubernetes hosts must comply with the base requirements plus the requirements for Add-on_1 plus the requirements for Add-on_2 plus the requirements for Add-on_3.

If you experience issues with traffic being routed incorrectly because iptables are being bypassed, then ensure that net.bridge.bridge-nf-call-iptables is set to 1 in your sysctl config. For example:

cat <<EOF > /etc/sysctl.d/k8s.conf
net.bridge.bridge-nf-call-ip6tables = 1
net.bridge.bridge-nf-call-iptables = 1
EOF
sysctl --system

Make sure that the br_netfilter module is loaded before this step. You can do this by executing the command lsmod | grep br_netfilter. To load it explicitly, call modprobe br_netfilter.

All Kubernetes host clocks must be synchronized with the HPE Ezmeral Runtime Enterprise clocks.

The HPE online sizer for HPE Ezmeral Runtime Enterprise can provide additional guidance for sizing your deployment. See: HPE Sizing Tool for Ezmeral Container Platform (link opens in a new browser tab/window).

Data Fabric Nodes

Each Data Fabric node must meet the minimum requirements described in Requirements for HPE Ezmeral Data Fabric on Kubernetes (for non-production environments only).

Disk and Network Requirements on Hosts Running etcd Service

Etcd is an integral part of the Kubernetes control plane, and requires low network latency and a sustained high performance disk to run.

Hewlett Packard Enterprise recommends running an etcd performance benchmark on your host to make sure it meets minimum network and disk I/O requirements: etcd benchmark tools (link opens an external site in a new browser tab or window).