Bug 1847463

Script that executes get_role_hosts function from ovn_migration.sh 
The script should return list of controller nodes, expected output on a setup with 3 controllers: controller-0 controller-1 controller-2

==========================================================
#!/bin/bash

get_role_hosts() {
    inventory_file=$1
    role_name=$2
    roles=`jq -r  \.$role_name\.children\[\] $inventory_file`
    for role in $roles; do
        # During the rocky cycle the format changed to have .value.hosts
        hosts=`jq -r --arg role "$role" 'to_entries[] | select(.key == $role) | .value.hosts[]' $inventory_file`
        if [[ "x$hosts" == "x" ]]; then
            # But we keep backwards compatibility with nested childrens (Queens)
            hosts=`jq -r --arg role "$role" 'to_entries[] | select(.key == $role) | .value.children[]' $inventory_file`

            for host in $hosts; do
               HOSTS="$HOSTS `jq -r --arg host "$host" 'to_entries[] | select(.key == $host) | .value.hosts[0]' $inventory_file`"
            done
        else
            HOSTS="${hosts} ${HOSTS}"
        fi
    done
    echo $HOSTS
}

source ~/stackrc
/usr/bin/tripleo-ansible-inventory --list > /tmp/ansible-inventory.txt
get_role_hosts /tmp/ansible-inventory.txt neutron_api

osp16
(undercloud) [stack@undercloud-0 ~]$ jq -r  \.neutron_api\.children\[\] /tmp/ansible-inventory.txt
Controller

osp16.1
(overcloud) [stack@undercloud-0 ~]$ jq -r  \.neutron_api\.children\[\] /tmp/ansible-inventory.txt
overcloud_neutron_api


related snippet from tripleo-ansible-inventory on osp16
    "neutron_api": {
        "children": [
            "Controller"
        ],
        "vars": {
            "ansible_ssh_user": "heat-admin"
        }
    },



related snippet from tripleo-ansible-inventory on osp16.1

   "neutron_api": {
        "children": [
            "overcloud_neutron_api"
        ]
    },
    "overcloud_neutron_api": {
        "children": [
            "overcloud_Controller"
        ]
    },

Created attachment 1697608 [details]
tripleo ansible inventory osp16.1

Created attachment 1697609 [details]
tripleo ansible inventory osp16

hosts_for_migration file generated on osp16.1 environment with 3 controllers and 2 computes (missing 2 controllers and 1 compute)
====================================================================================================
[ovn-dbs]
controller-0 ansible_host=192.168.24.29 ovn_central=true ansible_ssh_user=heat-admin ansible_become=true

[ovn-controllers]
compute-0 ansible_host=192.168.24.33 ansible_ssh_user=heat-admin ansible_become=true ovn_controller=true
controller-0 ansible_host=192.168.24.29 ansible_ssh_user=heat-admin ansible_become=true ovn_controller=true


[overcloud-controllers:children]
ovn-dbs

[overcloud:children]
ovn-controllers
ovn-dbs


[overcloud:vars]
remote_user=heat-admin
public_network_name=nova
image_name=cirros
working_dir=/home/stack/ovn_migration
server_user_name=cirros
validate_migration=True
overcloud_ovn_deploy_script=/home/stack/overcloud-deploy-ovn.sh
overcloudrc=/home/stack/overcloudrc
ovn_migration_backups=/var/lib/ovn-migration-backup

[overcloud-controllers:vars]
remote_user=heat-admin
public_network_name=nova
image_name=cirros
working_dir=/home/stack/ovn_migration
server_user_name=cirros
validate_migration=True
overcloud_ovn_deploy_script=/home/stack/overcloud-deploy-ovn.sh
overcloudrc=/home/stack/overcloudrc
ovn_migration_backups=/var/lib/ovn-migration-backup

Note: for OSP16.1 we can run "get_role_hosts" function as follows :

get_role_hosts /tmp/ansible-inventory.txt overcloud_neutron_api   (L143 from tools/ovn_migration/tripleo_environment/ovn_migration.sh )

in this case output is:
controller-0 controller-1 controller-2

as expected.

Also L158 should be changed for OSP16.1, it should look like:

get_role_hosts  ansible-inventory_osp16.1_ovs  overcloud_neutron_ovs_agent

in this case output is correct (the nodes where we want to launch ovn-controller after the migration):

controller-0 controller-1 controller-2 compute-0 compute-1

Created attachment 1698297 [details]
tripleo ansible inventory osp16.1 ml2ovs

Possible solution is to replace L93 in  ./tools/ovn_migration/tripleo_environment/ovn_migration.sh
from
roles=`jq -r  \.$role_name\.children\[\] $inventory_file`
to 
roles=`roles=`jq -r  \.overcloud_$role_name\.children\[\] $inventory_file ||  jq -r  \.$role_name\.children\[\] $inventory_file`

In this case the function returns proper lists of nodes from OSP16 and OSP16.1  ansible-inventory file for ovn and ovs

(In reply to Roman Safronov from comment #12)
> Possible solution is to replace L93 in 
> ./tools/ovn_migration/tripleo_environment/ovn_migration.sh
> from
> roles=`jq -r  \.$role_name\.children\[\] $inventory_file`
> to 
> roles=`roles=`jq -r  \.overcloud_$role_name\.children\[\] $inventory_file ||
> jq -r  \.$role_name\.children\[\] $inventory_file`
> 
> In this case the function returns proper lists of nodes from OSP16 and
> OSP16.1  ansible-inventory file for ovn and ovs

Cannot hard-code the 'overcloud' stack name.

> get_role_hosts /tmp/ansible-inventory.txt neutron_api

I'm a bit confused by this usage. Are you trying to get the host list for a service (e.g neutron_api), or the host list for a role (e.g Controller)?

(In reply to Ollie Walsh from comment #13)
> 
> Cannot hard-code the 'overcloud' stack name.
> 
> > get_role_hosts /tmp/ansible-inventory.txt neutron_api
> 
> I'm a bit confused by this usage. Are you trying to get the host list for a
> service (e.g neutron_api), or the host list for a role (e.g Controller)?

We need to get host list for a service (e.g. neutron_api).

Comments from the code:

    # We want to run ovn_dbs where neutron_api is running
    OVN_DBS=$(get_role_hosts /tmp/ansible-inventory.txt neutron_api)

    # We want to run ovn-controller where OVS agent was running before the migration
    OVN_CONTROLLERS=$(get_role_hosts /tmp/ansible-inventory.txt neutron_ovs_agent)

Can use ansible-inventory with --graph to expand the groups.

E.g:

$ tripleo-ansible-inventory --stack overcloud --static-yaml-inventory static_inventory.yaml
$ ansible-inventory -i static_inventory.yaml --graph neutron_api 
@neutron_api:
  |--@overcloud_neutron_api:
  |  |--@overcloud_Controller:
  |  |  |--overcloud-controller-0
$ ansible-inventory -i static_inventory.yaml --graph neutron_api | sed -ne 's/^[ \t|]\+--\([a-z0-9\-]\+\)$/\1/p'
overcloud-controller-0

Note ansible-inventory fails with a non-zero exit code if the group does not exist in the inventory

Moving to z2, this was not approved for z1 which is in Blockers Only. If it meets Blocker criteria for 16.1.1 please follow blocker process.

Verified on puddle RHOS-16.1-RHEL-8-20200813.n.0 with python3-networking-ovn-migration-tool-7.2.1-0.20200611133439.15f2281.el8ost.noarch
Verified that ansible inventory file for migration (hosts_for_migration) contains all relevant overcloud nodes.

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 (openstack-neutron bug fix 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/RHBA-2020:3568