From 72d2d0567e1af16aa3464d66899d1bc9891cbdd7 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Micha=C5=82=20Dulko?= <mdulko@redhat.com>
Date: Tue, 16 Oct 2018 17:03:20 +0200
Subject: [PATCH] Enable `--restart on-failure` for K8s containers

To workaround gate issues with etcd performance that were causing
kubernetes-controller-manager and kubernetes-scheduler pods to die in
the gate, this commit adds `--restart on-failure` to Kubernetes
containers we start in DevStack plugin.

Minor issue here is that watching the logs will not get restarted
because of DevStack's inability to set restart policy for system.d
services.

Change-Id: I73715c14641236dead437357036e227e3e29f041
Closes-Bug: 1798120
---
 devstack/plugin.sh | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/devstack/plugin.sh b/devstack/plugin.sh
index c2333b5b7..2bb9e6bb4 100644
--- a/devstack/plugin.sh
+++ b/devstack/plugin.sh
@@ -572,6 +572,7 @@ function run_k8s_api {
 
     run_container kubernetes-api \
         --net host \
+        --restart on-failure \
         --volume="${KURYR_HYPERKUBE_DATA_DIR}:/srv/kubernetes:rw" \
         "${KURYR_HYPERKUBE_IMAGE}:${KURYR_HYPERKUBE_VERSION}" \
             /hyperkube apiserver \
@@ -598,6 +599,7 @@ function run_k8s_controller_manager {
     run_container kubernetes-controller-manager \
         --net host \
         --volume="${KURYR_HYPERKUBE_DATA_DIR}:/srv/kubernetes:rw" \
+        --restart on-failure \
         "${KURYR_HYPERKUBE_IMAGE}:${KURYR_HYPERKUBE_VERSION}" \
             /hyperkube controller-manager \
                 --master="$KURYR_K8S_API_URL" \
@@ -615,6 +617,7 @@ function run_k8s_scheduler {
     run_container kubernetes-scheduler \
         --net host \
         --volume="${KURYR_HYPERKUBE_DATA_DIR}:/srv/kubernetes:rw" \
+        --restart on-failure \
         "${KURYR_HYPERKUBE_IMAGE}:${KURYR_HYPERKUBE_VERSION}" \
             /hyperkube scheduler \
                 --master="$KURYR_K8S_API_URL" \