This section contains the description of the MCP 2019.2.7 known issue and its workaround. For other MCP known issues also applicable to MCP 2019.2.7, see Known issues and corresponding sections in the previous maintenance updates.
Fixed in 2019.2.8
The upgrade of OpenStack Pike to Queens fails when running the
Deploy - upgrade control VMs Jenkins pipeline job for the
mdb
role if dist-upgrade
is enabled. As a workaround, do not set the
OS_UPGRADE and OS_DIST_UPGRADE parameters to True
when running the Deploy - upgrade control VMs Jenkins pipeline job
on mdb
nodes.
Fixed in 2019.2.9
The novav21
client state removes hosts from aggregates if the hosts
are defined within the OpenStack compute node definitions and the aggregates
are ensured through a pillar structure on the OpenStack controller node.
Workaround:
Open your Git project repository with the Reclass model on the cluster level.
In openstack/control_init.yml
, specify the following pillar:
parameters:
nova:
client:
enabled: true
resources:
v21:
admin_identity:
endpoint_type: internalURL
aggregates:
MyAggregateName:
hosts:
- host1
- host2
...
Apply the changes:
salt -C 'I@nova:client' state.apply nova.client
Fixed in 2019.2.9
Restarting the OpenContrail schema-transformer
service may cause loss of
SNAT connectivity for the instances without a floating IP. The issue may also
occur during the update of OpenContrail 4.1 when applying maintenance updates
to 2019.2.7 and newer.
Workaround:
Identify the router ID and the external network ID for the instances that lost the SNAT connectivity:
Log in to an OpenStack controller node.
Clear the router gateway:
neutron router-gateway-clear <router-id>
Reset the router external gateway:
neutron router-gateway-set <router-id> <external-net-id>