How to enable 1GB HugePages on OpenShift 4.x nodes ?

Environment

• Red Hat OpenShift Container Platform (OCP) 4.4

Issue

• Need to enable and use 1GB HugePages on worker nodes

Resolution

For OpenShift 4.5+

A Tuned profile can be used to configure the number of HugePages` in the worker nodes. This is described in the following section of the documentation: Configuring huge pages

For OpenShift 4.4 and lower, the way to configure 1G HugePages is via kernel arguments.

Before proceeding with how to enable 1GB HugePages, please go through the following knowledge base solution How do I enable 1GB size HugePage on X86-64 platform? and make sure the nodes support 1GB HugePages.

• Create a MachineConfig definition file like below to include kernel argument for 1G HugePages:

$ cat > 50-kargs-1g-hugepages.yaml << 'EOF'
apiVersion: machineconfiguration.openshift.io/v1
kind: MachineConfig
metadata:
  labels:
    machineconfiguration.openshift.io/role: worker
  name: 50-karg-1g-hugepages
spec:
  kernelArguments:
    - default_hugepagesz=1G
    - hugepagesz=1G
EOF

• Create the new MachineConfig:

$ oc create -f 50-kargs-1g-hugepages.yaml

• List the MachineConfig objects to see that the new one was added:

[user@upi-0 ~]$ oc get machineconfig
NAME                                                        GENERATEDBYCONTROLLER                      IGNITIONVERSION   AGE
00-master                                                   8af4f709c4ba9c0afff3408ecc99c8fce61dd314   2.2.0             15d
00-worker                                                   8af4f709c4ba9c0afff3408ecc99c8fce61dd314   2.2.0             15d
01-master-container-runtime                                 8af4f709c4ba9c0afff3408ecc99c8fce61dd314   2.2.0             15d
01-master-kubelet                                           8af4f709c4ba9c0afff3408ecc99c8fce61dd314   2.2.0             15d
01-worker-container-runtime                                 8af4f709c4ba9c0afff3408ecc99c8fce61dd314   2.2.0             15d
01-worker-kubelet                                           8af4f709c4ba9c0afff3408ecc99c8fce61dd314   2.2.0             15d
50-karg-1g-hugepages                                                                           3s   <------ 
99-master-0db35ec4-b245-4083-a664-a3321b6f3478-registries   8af4f709c4ba9c0afff3408ecc99c8fce61dd314   2.2.0             15d
99-master-ssh                                                                                          2.2.0             15d
99-worker-2416c803-0c17-4a09-9573-d6249634c023-kubelet      8af4f709c4ba9c0afff3408ecc99c8fce61dd314   2.2.0             8d
99-worker-2416c803-0c17-4a09-9573-d6249634c023-registries   8af4f709c4ba9c0afff3408ecc99c8fce61dd314   2.2.0             15d
99-worker-ssh                                                                                          2.2.0             15d
rendered-master-97018ee7bbf8151a1ad368564e8ecfd8            8af4f709c4ba9c0afff3408ecc99c8fce61dd314   2.2.0             15d
rendered-worker-0252be1f29822e0a1fc7f02fbee550f7            8af4f709c4ba9c0afff3408ecc99c8fce61dd314   2.2.0             8d

• Observe there is a new MachineConfig named 50-karg-1g-hugepages.

• Check the status of nodes now, the node will reboot and configure the kernel argument. It may take around 15-20 minute depending on the environment before the nodes become Ready again.

[quicklab@upi-0 ~]$ oc get nodes
NAME       STATUS   ROLES                  AGE   VERSION
master-0   Ready    master                 15d   v1.17.1
master-1   Ready    master                 15d   v1.17.1
master-2   Ready    master                 15d   v1.17.1
worker-0   Ready    worker                 15d   v1.17.1

• Login in to the node and check the kernel argument:

# ssh core@worker-0
[core@worker-0 ~]$ cat /proc/cmdline 
BOOT_IMAGE=(hd0,gpt1)/ostree/rhcos-e5da40289a28b6f0cc587c812476d495019db137587bc91f6d12faf3df7b6d8c/vmlinuz-4.18.0-147.8.1.el8_1.x86_64 rhcos.root=crypt_rootfs console=tty0 console=ttyS0,115200n8 rd.luks.options=discard ostree=/ostree/boot.1/rhcos/e5da40289a28b6f0cc587c812476d495019db137587bc91f6d12faf3df7b6d8c/0 ignition.platform.id=openstack default_hugepagesz=1G hugepagesz=1G
[core@worker-1 ~]$

• Observe the two arguments in the/proc/cmdline output: default_hugepagesz=1G and hugepagesz=1G.

This solution is part of Red Hat’s fast-track publication program, providing a huge library of solutions that Red Hat engineers have created while supporting our customers. To give you the knowledge you need the instant it becomes available, these articles may be presented in a raw and unedited form.