summaryrefslogtreecommitdiffstats
path: root/roles/openshift_node_upgrade
diff options
context:
space:
mode:
authorDan Williams <dcbw@redhat.com>2017-07-20 15:42:07 -0500
committerDan Williams <dcbw@redhat.com>2017-10-05 13:15:03 -0500
commit0164d95a2457dccc61163fde984abf7c80456624 (patch)
treee028ffe806ce023d9b62a72751e2e9e9eb761c91 /roles/openshift_node_upgrade
parent68358320f1f077d50f28c41bbe45700676302910 (diff)
downloadopenshift-0164d95a2457dccc61163fde984abf7c80456624.tar.gz
openshift-0164d95a2457dccc61163fde984abf7c80456624.tar.bz2
openshift-0164d95a2457dccc61163fde984abf7c80456624.tar.xz
openshift-0164d95a2457dccc61163fde984abf7c80456624.zip
node: make node service PartOf=openvswitch.service when openshift-sdn is used
Commit 7f805f9a0c41477365dd88b0ac73f0d221bd654a causes the behavior seen in https://bugzilla.redhat.com/show_bug.cgi?id=1453113 because openshift-node is no longer restarted when openvswitch is, due to the change from Requires to Wants. Turns out that making the openshift node service PartOf the OVS service can achieve the same result and ensure openshift-node gets restarted whenever OVS does, which ensures that networking doesn't break underneath the node. Suggested by Giuseppe Scrivano
Diffstat (limited to 'roles/openshift_node_upgrade')
-rw-r--r--roles/openshift_node_upgrade/templates/openshift.docker.node.service1
1 files changed, 1 insertions, 0 deletions
diff --git a/roles/openshift_node_upgrade/templates/openshift.docker.node.service b/roles/openshift_node_upgrade/templates/openshift.docker.node.service
index 864e4b5d6..07d1ebc3c 100644
--- a/roles/openshift_node_upgrade/templates/openshift.docker.node.service
+++ b/roles/openshift_node_upgrade/templates/openshift.docker.node.service
@@ -6,6 +6,7 @@ PartOf={{ openshift.docker.service_name }}.service
Requires={{ openshift.docker.service_name }}.service
{% if openshift_use_openshift_sdn %}
Wants=openvswitch.service
+PartOf=openvswitch.service
After=ovsdb-server.service
After=ovs-vswitchd.service
{% endif %}