diff --git a/helm/ton-rust-node/CHANGELOG.md b/helm/ton-rust-node/CHANGELOG.md
index 8c8609c..87f30ab 100644
--- a/helm/ton-rust-node/CHANGELOG.md
+++ b/helm/ton-rust-node/CHANGELOG.md
@@ -5,6 +5,14 @@ All notable changes to the Helm chart will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/).
 Versions follow the Helm chart release tags (e.g. `helm/v0.3.0`).
 
+## [0.4.3] - 2026-04-01
+
+appVersion: `v0.3.0`
+
+### Added
+
+- `terminationGracePeriodSeconds` — configurable grace period before SIGKILL on pod termination. Defaults to 300s (5 minutes). The Kubernetes default of 30s is too short for a TON node — an unclean kill may corrupt the database and forces a cold boot
+
 ## [0.4.2] - 2026-03-18
 
 appVersion: `v0.3.0`
diff --git a/helm/ton-rust-node/Chart.yaml b/helm/ton-rust-node/Chart.yaml
index 12d65dc..e5d4546 100644
--- a/helm/ton-rust-node/Chart.yaml
+++ b/helm/ton-rust-node/Chart.yaml
@@ -2,7 +2,7 @@ apiVersion: v2
 name: node
 description: TON Rust Node deployment
 type: application
-version: 0.4.2
+version: 0.4.3
 appVersion: "v0.3.0"
 
 sources:
diff --git a/helm/ton-rust-node/templates/statefulset.yaml b/helm/ton-rust-node/templates/statefulset.yaml
index 34b35be..ffca504 100644
--- a/helm/ton-rust-node/templates/statefulset.yaml
+++ b/helm/ton-rust-node/templates/statefulset.yaml
@@ -46,6 +46,7 @@ spec:
       affinity:
         {{- toYaml . | nindent 8 }}
       {{- end }}
+      terminationGracePeriodSeconds: {{ .Values.terminationGracePeriodSeconds }}
       {{- if .Values.hostNetwork }}
       hostNetwork: true
       dnsPolicy: ClusterFirstWithHostDNS
diff --git a/helm/ton-rust-node/values.yaml b/helm/ton-rust-node/values.yaml
index 9d00a04..8d4d719 100644
--- a/helm/ton-rust-node/values.yaml
+++ b/helm/ton-rust-node/values.yaml
@@ -407,6 +407,10 @@ serviceAccount:
   name: ""
   annotations: {}
 
+## @param terminationGracePeriodSeconds Time (in seconds) given to the node process to shut down gracefully before SIGKILL. The default Kubernetes value (30s) is too short for a TON node — an unclean kill may corrupt the database and forces a cold boot. Set this to at least 300s.
+##
+terminationGracePeriodSeconds: 300
+
 ## @section Scheduling parameters
 
 ## @param nodeSelector [object] Node selector for pod scheduling