mirror of
https://github.com/openshift/openshift-docs.git
synced 2026-02-05 12:46:18 +01:00
46 lines
1.6 KiB
Plaintext
46 lines
1.6 KiB
Plaintext
// Module included in the following assemblies:
|
|
//
|
|
// microshift_configuring/microshift_low_latency/microshift-low-latency.adoc
|
|
|
|
:_mod-docs-content-type: PROCEDURE
|
|
[id="microshift-low-latency-llc-enable_{context}"]
|
|
= Enabling last-level cache locality in {microshift-short}
|
|
|
|
[role="_abstract"]
|
|
You can align workloads with CPU cores that share the same last-level cache (LLC) to improve performance for latency-sensitive applications. To apply this alignment, enable the Kubernetes CPU Manager option `prefer-align-cpus-by-uncorecache` in {microshift}.
|
|
|
|
:FeatureName: Last-level cache (LLC) locality
|
|
include::snippets/technology-preview.adoc[]
|
|
:FeatureName!:
|
|
|
|
[WARNING]
|
|
====
|
|
This feature is part of a feature gate. After you enable feature gates, you cannot disable them or update {microshift-short}, and your cluster can become unstable or lose data. Enable feature gates only in non-production environments.
|
|
====
|
|
|
|
.Procedure
|
|
|
|
. Add the following content to `/etc/microshift/config.yaml`. If you already have a `kubelet` section from the earlier section, merge the `cpuManagerPolicyOptions` entry and ensure the feature gate is present.
|
|
+
|
|
[source,yaml]
|
|
----
|
|
apiServer:
|
|
featureGates:
|
|
featureSet: "CustomNoUpgrade"
|
|
customNoUpgrade:
|
|
enabled:
|
|
- "CPUManagerPolicyBetaOptions"
|
|
kubelet:
|
|
reservedSystemCPUs: "0"
|
|
cpuManagerPolicy: static
|
|
cpuManagerPolicyOptions:
|
|
prefer-align-cpus-by-uncorecache: "true"
|
|
----
|
|
|
|
. To apply the configuration, restart {microshift-short} by running the following command:
|
|
+
|
|
[source,terminal,subs="+quotes"]
|
|
----
|
|
$ sudo systemctl restart microshift
|
|
----
|