VOL-794: Assignment lists in etcd not updated correctly in presence of vcore failures
* Fixed coordinator method _sanitize_member_list
* Fixed keys-only GET operation for etcd KV store
* Fixed key DELETE operation for etcd KV store
* Modified vcore manifest to allow a failed pod to be re-spawned on a
different host but with the same pod name
* Modified zookeeper manifest to allow a failed pod to be re-spawned
on a different host. With the current manifest, when all 3 zookeeper
pods are scheduled on the same node and that node fails, none of the
pods get re-spawned.
* Added NodePort for the Envoy service
* Removed anti-affinity rule from OFAgent and NetConf manifests to allow
the deployment of multiple pods on the same node
Change-Id: I052d952d81a81cafb96acfc1d57a192596e2e9a1
diff --git a/k8s/zookeeper.yml b/k8s/zookeeper.yml
index 1c7f02d..93e4385 100644
--- a/k8s/zookeeper.yml
+++ b/k8s/zookeeper.yml
@@ -78,6 +78,7 @@
annotations:
cni: "calico"
spec:
+ terminationGracePeriodSeconds: 0
containers:
- name: zoo1
image: zookeeper:3.4.11
@@ -118,6 +119,7 @@
annotations:
cni: "calico"
spec:
+ terminationGracePeriodSeconds: 0
containers:
- name: zoo2
image: zookeeper:3.4.11
@@ -158,6 +160,7 @@
annotations:
cni: "calico"
spec:
+ terminationGracePeriodSeconds: 0
containers:
- name: zoo3
image: zookeeper:3.4.11