Tolerate failures in the node upgrade playbook

4 files changed, 52 insertions, 2 deletions

diff --git a/inventory/byo/hosts.origin.example b/inventory/byo/hosts.origin.example
index 20f342023..310b8ab44 100644
--- a/inventory/byo/hosts.origin.example
+++ b/inventory/byo/hosts.origin.example
@@ -788,6 +788,31 @@ openshift_master_identity_providers=[{'name': 'htpasswd_auth', 'login': 'true',
 #
 #etcd_ca_default_days=1825
 
+# Upgrade Control
+#
+# By default nodes are upgraded in a serial manner one at a time and all failures
+# are fatal
+#openshift_upgrade_nodes_serial=1
+#openshift_upgrade_nodes_max_fail_percentage=0
+#
+# You can specify the number of nodes to upgrade at once. We do not currently
+# attempt to verify that you have capacity to drain this many nodes at once
+# so please be careful when specifying these values. You should also verify that
+# the expected number of nodes are all schedulable and ready before starting an
+# upgrade. If it's not possible to drain the requested nodes the upgrade will
+# stall indefinitely until the drain is successful.
+#
+# If you're upgrading more than one node at a time you can specify the maximum
+# percentage of failure within the batch before the upgrade is aborted. Any
+# nodes that do fail are ignored for the rest of the playbook run and you should
+# take care to investigate the failure and return the node to service so that
+# your cluster.
+#
+# The percentage must exceed the value, this would fail on two failures
+# openshift_upgrade_nodes_serial=4 openshift_upgrade_nodes_max_fail_percentage=49
+# where as this would not
+# openshift_upgrade_nodes_serial=4 openshift_upgrade_nodes_max_fail_percentage=50
+
 # host group for masters
 [masters]
 ose3-master[1:3]-ansible.test.example.com
diff --git a/inventory/byo/hosts.ose.example b/inventory/byo/hosts.ose.example
index f75a47bb8..e126bbcab 100644
--- a/inventory/byo/hosts.ose.example
+++ b/inventory/byo/hosts.ose.example
@@ -785,6 +785,31 @@ openshift_master_identity_providers=[{'name': 'htpasswd_auth', 'login': 'true',
 #
 #etcd_ca_default_days=1825
 
+# Upgrade Control
+#
+# By default nodes are upgraded in a serial manner one at a time and all failures
+# are fatal
+#openshift_upgrade_nodes_serial=1
+#openshift_upgrade_nodes_max_fail_percentage=0
+#
+# You can specify the number of nodes to upgrade at once. We do not currently
+# attempt to verify that you have capacity to drain this many nodes at once
+# so please be careful when specifying these values. You should also verify that
+# the expected number of nodes are all schedulable and ready before starting an
+# upgrade. If it's not possible to drain the requested nodes the upgrade will
+# stall indefinitely until the drain is successful.
+#
+# If you're upgrading more than one node at a time you can specify the maximum
+# percentage of failure within the batch before the upgrade is aborted. Any
+# nodes that do fail are ignored for the rest of the playbook run and you should
+# take care to investigate the failure and return the node to service so that
+# your cluster.
+#
+# The percentage must exceed the value, this would fail on two failures
+# openshift_upgrade_nodes_serial=4 openshift_upgrade_nodes_max_fail_percentage=49
+# where as this would not
+# openshift_upgrade_nodes_serial=4 openshift_upgrade_nodes_max_fail_percentage=50
+
 # host group for masters
 [masters]
 ose3-master[1:3]-ansible.test.example.com
diff --git a/playbooks/common/openshift-cluster/upgrades/upgrade_control_plane.yml b/playbooks/common/openshift-cluster/upgrades/upgrade_control_plane.yml
index 0ad934d2d..a335c1072 100644
--- a/playbooks/common/openshift-cluster/upgrades/upgrade_control_plane.yml
+++ b/playbooks/common/openshift-cluster/upgrades/upgrade_control_plane.yml
@@ -248,7 +248,7 @@
   # This var must be set with -e on invocation, as it is not a per-host inventory var
   # and is evaluated early. Values such as "20%" can also be used.
   serial: "{{ openshift_upgrade_nodes_serial | default(1) }}"
-  any_errors_fatal: true
+  max_fail_percentage: "{{ openshift_upgrade_nodes_max_fail_percentage | default(0) }}"
 
   pre_tasks:
   - name: Load lib_openshift modules
diff --git a/playbooks/common/openshift-cluster/upgrades/upgrade_nodes.yml b/playbooks/common/openshift-cluster/upgrades/upgrade_nodes.yml
index 4d455fe0a..91dbc2cd4 100644
--- a/playbooks/common/openshift-cluster/upgrades/upgrade_nodes.yml
+++ b/playbooks/common/openshift-cluster/upgrades/upgrade_nodes.yml
@@ -4,7 +4,7 @@
   # This var must be set with -e on invocation, as it is not a per-host inventory var
   # and is evaluated early. Values such as "20%" can also be used.
   serial: "{{ openshift_upgrade_nodes_serial | default(1) }}"
-  any_errors_fatal: true
+  max_fail_percentage: "{{ openshift_upgrade_nodes_max_fail_percentage | default(0) }}"
 
   pre_tasks:
   - name: Load lib_openshift modules