openstack
/
kuryr-kubernetes
Code Issues Proposed changes

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
This commit is contained in:
Michał Dulko 2018-10-16 17:03:20 +02:00
parent 94cb656c2e
commit 72d2d0567e
1 changed files with 3 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
devstack/plugin.sh
View File

@ -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" \