Avoid systemd ddos with kubelet if it fails to start
This patch introduces restart sleeps, e.g. systemd
will be wait 2s between stop and start kubelet.
This will help to avoid issues with systemd in case
if some salts going to reapplied due previous fails.
The problem appears if some of critial components
failed to install with first try. Usually, kubelet installed
successfully, but calico/criproxy/dockershim may
become a showstopper for kubelet. So, without this param
kubelet throttling systemd with restarts because can not
start correctly.
Change-Id: Ifaad5c60a1afb6c61bbd0b4549df7054d442979c
diff --git a/kubernetes/files/systemd/kubelet.service b/kubernetes/files/systemd/kubelet.service
index c1f2522..c353ece 100644
--- a/kubernetes/files/systemd/kubelet.service
+++ b/kubernetes/files/systemd/kubelet.service
@@ -25,7 +25,8 @@
$CADVISOR_PORT \
$DAEMON_ARGS
Restart=on-failure
+RestartSec=5
[Install]
WantedBy=multi-user.target
-Alias=cadvisor.service
\ No newline at end of file
+Alias=cadvisor.service