From c8ccf879b81f156792719eb02617c56c891cd544 Mon Sep 17 00:00:00 2001 From: Kunal Bauskar Date: Thu, 28 May 2026 16:26:44 +0530 Subject: [PATCH] reduce pod eviction delay on node failure Set node-monitor-grace-period=40s, default-not-ready/unreachable-toleration-seconds=60 to cut pod eviction time from ~5min to ~60s after node failure. --- .../pf9/pf9-kube/conf/masterconfig/base/centos/master.yaml | 3 +++ .../pf9/pf9-kube/conf/masterconfig/base/ubuntu/master.yaml | 3 +++ 2 files changed, 6 insertions(+) diff --git a/nodelet/pkg/pf9kube/pf9/pf9-kube/conf/masterconfig/base/centos/master.yaml b/nodelet/pkg/pf9kube/pf9/pf9-kube/conf/masterconfig/base/centos/master.yaml index cf9cecec..30cdfaf8 100644 --- a/nodelet/pkg/pf9kube/pf9/pf9-kube/conf/masterconfig/base/centos/master.yaml +++ b/nodelet/pkg/pf9kube/pf9/pf9-kube/conf/masterconfig/base/centos/master.yaml @@ -20,6 +20,7 @@ spec: - "--v=__DEBUG_LEVEL__" - "--horizontal-pod-autoscaler-use-rest-clients=true" - "--use-service-account-credentials=true" + - "--node-monitor-grace-period=40s" volumeMounts: - name: "apiserver-certs" mountPath: "/srv/kubernetes/certs/apiserver" @@ -86,6 +87,8 @@ spec: - "--kubelet-client-key=/srv/kubernetes/certs/kubelet/apiserver/request.key" - "--v=__DEBUG_LEVEL__" - "--http2-max-streams-per-connection=1000" # See: PMK-2969 + - "--default-not-ready-toleration-seconds=60" + - "--default-unreachable-toleration-seconds=60" volumeMounts: - name: "authn" mountPath: "/srv/kubernetes/authn" diff --git a/nodelet/pkg/pf9kube/pf9/pf9-kube/conf/masterconfig/base/ubuntu/master.yaml b/nodelet/pkg/pf9kube/pf9/pf9-kube/conf/masterconfig/base/ubuntu/master.yaml index c513f9ef..68ba2504 100644 --- a/nodelet/pkg/pf9kube/pf9/pf9-kube/conf/masterconfig/base/ubuntu/master.yaml +++ b/nodelet/pkg/pf9kube/pf9/pf9-kube/conf/masterconfig/base/ubuntu/master.yaml @@ -20,6 +20,7 @@ spec: - "--v=__DEBUG_LEVEL__" - "--horizontal-pod-autoscaler-use-rest-clients=true" - "--use-service-account-credentials=true" + - "--node-monitor-grace-period=40s" volumeMounts: - name: "apiserver-certs" mountPath: "/srv/kubernetes/certs/apiserver" @@ -92,6 +93,8 @@ spec: - "--kubelet-client-key=/srv/kubernetes/certs/kubelet/apiserver/request.key" - "--v=__DEBUG_LEVEL__" - "--http2-max-streams-per-connection=1000" # See: PMK-2969 + - "--default-not-ready-toleration-seconds=60" + - "--default-unreachable-toleration-seconds=60" volumeMounts: - name: "authn" mountPath: "/srv/kubernetes/authn"