Description

The basic logic of MTU configuration in Cilium is:

• The full MTU value is always set on pod interfaces.
• In the case of using encapsulation or encryption, an abbreviated MTU value is set on the default route inside the pod.

However, this logic does not work well in the context of virtualization, where internal interfaces are created with the MTU of the pod and there is no way to determine the MTU of routes. To work correctly with virtualization, we have previously made changes to create pod interfaces with a reduced MTU.

With the latest version of Cilium, a mechanism for automatically updating MTUs on all pod interfaces has been added, while still following its basic logic.

In this PR, we will make the same change to the automatic MTU update mechanism.

Why do we need it, and what problem does it solve?

When using custom high MTU values, different calculation methods in two different places can lead to problems.

Why do we need it in the patch release (if we do)?

This has been identified in production clusters, and it has become a problem.

Checklist

• The code is covered by unit tests.
• e2e tests passed.
• Documentation updated according to the changes.
• Changes were tested in the Kubernetes cluster manually.

Changelog entries

section: cni-cilium
type: fix
summary: The MTU configuration has been updated.
impact: The MTU will be updated on all interfaces of all pods.
impact_level: default