Bug 1451023

Change to the defaults. 

https://github.com/openshift/openshift-ansible/commit/b50b4ea0b03feb9431abd7294fe4fb6b549ddfc0

A workaround is of course to set osm_cluster_network_cidr and osm_host_subnet_length to the old values before running the scaleup playbook. 

While we should fix this I'm lowering severity to medium based on the easy workaround.

Added an inventory check before upgrade which makes sure that these two variables are explicitly set. This is what it looks like when one or both are not set via inventory:


Tuesday 29 August 2017  11:35:50 -0400 (0:00:00.013)       0:00:03.567 ********                           
fatal: [192.168.124.234]: FAILED! => {                                                                    
    "assertion": "osm_cluster_network_cidr is defined",                                                   
    "changed": false,                                                                                     
    "evaluated_to": false,                                                                                
    "failed": true                                                                                                                                                                                                   
}                                                                                                         
                                                                                                                                                                                                                     
MSG:                                                                                                      
                                                                                                          
osm_cluster_network_cidr and openshift_portal_net are required inventory variables when upgrading. These variables should match what is currently used in the cluster. If you don't remember what these values are yo
u can find them in /etc/origin/master/master-config.yaml on a master with the names clusterNetworkCIDR(osm_cluster_network_cidr)  and hostSubnetLength (openshift_portal_net).



PR: https://github.com/openshift/openshift-ansible/pull/5256

Merged

Updated message: https://github.com/openshift/openshift-ansible/pull/5386

Added doc text.

From QE's perspective, the resolution described in comment 2 should be the best one, and the fix should be landed into scale up playbook, but not upgrade playbook.

> When performing a scaleup we need to read the CIDR values from an existing master then set that fact on the scaled up masters.

Even if can not implement that fix in short term, as a compromise, when performing a scaleup, once installer find CIDR values from an existing master is mismatched with the new CIDR values for new master, installer exit and prompt user to set osm_cluster_network_cidr and osm_host_subnet_length to the old values before running the scaleup playbook. 

From customer's perspective, when running upgrade, it is not reasonable to force user to set osm_cluster_network_cidr and osm_host_subnet_length into inventory host file. it is not relative to upgrade, but scaleup.

So assign this bug ack.

*** Bug 1493268 has been marked as a duplicate of this bug. ***

The final fix for the issue: https://github.com/openshift/openshift-ansible/pull/5473

Moving to MODIFIED as it's not built into rpm package.

Tested with openshift-ansible-3.7.0-0.134.0.git.0.6f43fc3.el7.noarch.rpm

1. Trigger master HA installation with original network parameters
# cat inventory_host
<--snip-->
osm_cluster_network_cidr=11.0.0.0/16
osm_host_subnet_length=8
openshift_master_portal_net=172.31.0.0/16
<--snip-->

2. Removed above network parameters from inventory file

3. Scale up one master against the env above

##Result:

Installation succeeded, but the master-controllers on the new master was indicating some errors:

" failed to start SDN plugin controller: cannot change the serviceNetworkCIDR of an already-deployed cluster"

Dig more found that the new master was still using the new default portal net:
# grep -nri "serviceNetworkCIDR:" /etc/origin/master/master-config.yaml
162:  serviceNetworkCIDR: 172.30.0.0/16

As `initialize_facts.yml` was executed prior to `set_network_facts.yml`, so installer would still take the default `portal_net` for the following tasks.

./playbooks/common/openshift-cluster/initialize_facts.yml:143:        portal_net: "{{ openshift_portal_net | default(openshift_master_portal_net) | default(None) }}"

https://github.com/openshift/openshift-ansible/pull/5621

Verified with openshift-ansible-3.7.0-0.147.0.git.0.2fb41ee.el7.noarch.rpm

Test steps:
1. Trigger master HA installation with original network parameters
# cat inventory_host
<--snip-->
osm_cluster_network_cidr=10.1.0.0/16
osm_host_subnet_length=8
openshift_portal_net=172.31.0.0/16
<--snip-->

2. Removed above network parameters from inventory file

3. Scale up one master against the env above

4. Check the new master parameters against the new master:

# grep -E "NetworkCIDR|Length" /etc/origin/master/master-config.yaml
  clusterNetworkCIDR: 10.1.0.0/16
  externalIPNetworkCIDRs:
  hostSubnetLength: 8
  serviceNetworkCIDR: 172.31.0.0/16

5. Trigger S2I build against the new master, works well

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHSA-2017:3188