From 69a67556c77d9ef2fd9110a8c8e60b834c05a3c7 Mon Sep 17 00:00:00 2001 From: Paco Xu Date: Thu, 8 Aug 2024 21:02:13 +0800 Subject: [PATCH] kubelet: add warning log and events for cgroup v2 running on kernel < 5.8 --- pkg/kubelet/cm/container_manager.go | 3 +++ pkg/kubelet/events/event.go | 1 + pkg/kubelet/kubelet.go | 16 ++++++++++++++-- pkg/util/kernel/constants.go | 5 +++++ 4 files changed, 23 insertions(+), 2 deletions(-) diff --git a/pkg/kubelet/cm/container_manager.go b/pkg/kubelet/cm/container_manager.go index 99fb69c8b98..15215cc86c8 100644 --- a/pkg/kubelet/cm/container_manager.go +++ b/pkg/kubelet/cm/container_manager.go @@ -46,6 +46,9 @@ import ( const ( // Warning message for the users still using cgroup v1 CgroupV1MaintenanceModeWarning = "Cgroup v1 support is in maintenance mode, please migrate to Cgroup v2." + + // Warning message for the users using cgroup v2 on kernel older than 5.8 + CgroupV2KernelWarning = "Cgroup v2 is being used on a kernel older than 5.8, which may have limited functionality." ) type ActivePodsFunc func() []*v1.Pod diff --git a/pkg/kubelet/events/event.go b/pkg/kubelet/events/event.go index a9948ba6a54..f23c68c1726 100644 --- a/pkg/kubelet/events/event.go +++ b/pkg/kubelet/events/event.go @@ -77,6 +77,7 @@ const ( FailedPrepareDynamicResources = "FailedPrepareDynamicResources" PossibleMemoryBackedVolumesOnDisk = "PossibleMemoryBackedVolumesOnDisk" CgroupV1 = "CgroupV1" + CgroupV2 = "CgroupV2" ) // Image manager event reason list diff --git a/pkg/kubelet/kubelet.go b/pkg/kubelet/kubelet.go index 68d790242db..a39ee656b86 100644 --- a/pkg/kubelet/kubelet.go +++ b/pkg/kubelet/kubelet.go @@ -52,6 +52,7 @@ import ( "k8s.io/apimachinery/pkg/types" utilruntime "k8s.io/apimachinery/pkg/util/runtime" "k8s.io/apimachinery/pkg/util/sets" + "k8s.io/apimachinery/pkg/util/version" "k8s.io/apimachinery/pkg/util/wait" utilfeature "k8s.io/apiserver/pkg/util/feature" clientset "k8s.io/client-go/kubernetes" @@ -120,6 +121,7 @@ import ( "k8s.io/kubernetes/pkg/kubelet/volumemanager" httpprobe "k8s.io/kubernetes/pkg/probe/http" "k8s.io/kubernetes/pkg/security/apparmor" + utilkernel "k8s.io/kubernetes/pkg/util/kernel" "k8s.io/kubernetes/pkg/util/oom" "k8s.io/kubernetes/pkg/volume" "k8s.io/kubernetes/pkg/volume/csi" @@ -1645,7 +1647,7 @@ func (kl *Kubelet) Run(updates <-chan kubetypes.PodUpdate) { os.Exit(1) } - kl.warnCgroupV1Usage() + kl.cgroupVersionCheck() // Start volume manager go kl.volumeManager.Run(kl.sourcesReady, wait.NeverStop) @@ -3067,11 +3069,21 @@ func (kl *Kubelet) UnprepareDynamicResources(pod *v1.Pod) error { return kl.containerManager.UnprepareDynamicResources(pod) } -func (kl *Kubelet) warnCgroupV1Usage() { +func (kl *Kubelet) cgroupVersionCheck() { cgroupVersion := kl.containerManager.GetNodeConfig().CgroupVersion if cgroupVersion == 1 { kl.recorder.Eventf(kl.nodeRef, v1.EventTypeWarning, events.CgroupV1, cm.CgroupV1MaintenanceModeWarning) klog.V(2).InfoS("Warning: cgroup v1", "message", cm.CgroupV1MaintenanceModeWarning) + } else { + kernelVersion, err := utilkernel.GetVersion() + if err != nil { + klog.V(2).ErrorS(err, "Warning: cannot determine kernel version, unable to determine if cgroup v2 is supported", "message", cm.CgroupV1MaintenanceModeWarning) + } + + if kernelVersion.LessThan(version.MustParseGeneric(utilkernel.CgroupV2KernelVersion)) { + kl.recorder.Eventf(kl.nodeRef, v1.EventTypeWarning, events.CgroupV2, cm.CgroupV2KernelWarning) + klog.V(2).InfoS("Warning: cgroup v2", "message", cm.CgroupV2KernelWarning) + } } metrics.CgroupVersion.Set(float64(cgroupVersion)) } diff --git a/pkg/util/kernel/constants.go b/pkg/util/kernel/constants.go index 6775027e7a0..2bf27ac0750 100644 --- a/pkg/util/kernel/constants.go +++ b/pkg/util/kernel/constants.go @@ -54,3 +54,8 @@ const TmpfsNoswapSupportKernelVersion = "6.4" // nftables mode with by default. This is not directly related to any specific kernel // commit; see https://issues.k8s.io/122743#issuecomment-1893922424 const NFTablesKubeProxyKernelVersion = "5.13" + +// CgroupV2KernelVersion is the minimal kernel version required for cgroup v2 support. +// `cpu.stat` file was added to root cgroup. +// (ref: https://github.com/torvalds/linux/commit/936f2a70f2077f64fab1dcb3eca71879e82ecd3f) +const CgroupV2KernelVersion = "5.8"