Bug 1512581

Description of problem:

After upgrading cluster from 2.4 to 3.0 the old custom parameters used might be missed.

For eg: During 2.4 if customer uses different osd_jounral size and they edited some other param's for their environment, during upgrade ansible picks old released version of all.yml and upgrades the cluster. After this as per our doc upgrade suggestion replace the old all.yml with new all.yml.sample so there is possibility to overwrite those custom param's set by them and it could lead to failure.

It would be better to add or remove those custom param's before replacing that all.yml

During 2.4 to 3.0 upgrade got below problem, in which latest ceph-ansible expects new param in existing old all.yml file
=============
Ansible failure during upgrade:
TASK [ceph-common : make sure monitor_interface, monitor_address or monitor_address_block is defined] ***
task path: /usr/share/ceph-ansible/roles/ceph-common/tasks/checks/check_mandatory_vars.yml:67
skipping: [magna006] => {
    "changed": false,
    "skip_reason": "Conditional result was False",
    "skipped": true
}

TASK [ceph-common : make sure radosgw_interface, radosgw_address or radosgw_address_block is defined] ***
task path: /usr/share/ceph-ansible/roles/ceph-common/tasks/checks/check_mandatory_vars.yml:76
fatal: [magna006]: FAILED! => {
    "changed": false,
    "failed": true,
    "msg": "you must set radosgw_interface, radosgw_address or radosgw_address_block"
}
================

Not sure either this needs to be coded in ceph-ansible or manual way of matching the params but coding will be better to avoid problem.

Note: Filed this bug based on the info that ceph-ansible uses old version of all.yml to upgrade cluster to latest release build, if not please update the bug accordingly.

Version-Release number of selected component (if applicable):
NA

How reproducible:
2/2

Steps to Reproduce:
NA

Actual results:
NA

Expected results:
NA

Additional info:
NA