openQA: tweak ansible_ifcfg_whitelist handling for workers

As we have more and more tap workers, it's no longer practical
to do this for the whole group, as each one seems to have a
*different* set of interfaces a) present and b) 'active'. So
let's just take it out of the group vars and do it per-host
instead. Each tap worker only actually has *one* real hardware
interface that should be handled by network / NetworkManager,
so we just list that one per-host (it's eth0 for all of them
except the ppc64 host, which is eth2 for some damn reason).

Signed-off-by: Adam Williamson <awilliam@redhat.com>

inventory/group_vars/openqa-tap-workers

@@ -12,9 +12,3 @@ custom_rules: [
     '-A FORWARD -m state -i eth2 -o br0 --state RELATED,ESTABLISHED -j ACCEPT',
     '-A INPUT -i br0 -j ACCEPT'
 ]
-
-# we do stuff with ifcfg that base doesn't understand. terrible, terrible
-# stuff. seriously - it doesn't handle the openvswitch config well. so
-# let's tell it to just configure eth0 (and eth2, for ppc64) for us and
-# leave everything else alone.
-ansible_ifcfg_whitelist: ['eth0', 'eth2']

inventory/host_vars/openqa-aarch64-01.qa.fedoraproject.org

@@ -12,4 +12,11 @@ nrpe_procs_crit: 300
 
 sudoers: "{{ private }}/files/sudo/qavirt-sudoers"
 
+# openQA tap worker hosts (like this one) do stuff with ifcfg that base
+# doesn't understand. terrible, terrible stuff. seriously - it doesn't
+# handle the openvswitch config well. so let's tell it to just configure
+# the actual connected interface (which on this system is eth0) for us
+# and leave everything else alone.
+ansible_ifcfg_whitelist: ['eth0']
+
 openqa_workers: 4

inventory/host_vars/openqa-ppc64le-01.qa.fedoraproject.org

@@ -11,9 +11,12 @@ nrpe_procs_crit: 300
 
 sudoers: "{{ private }}/files/sudo/qavirt-sudoers"
 
-# Given that the only interface here is eth2, we can't really make ansible do its magic
-ansible_ifcfg_whitelist:
-- nonexistent
+# openQA tap worker hosts (like this one) do stuff with ifcfg that base
+# doesn't understand. terrible, terrible stuff. seriously - it doesn't
+# handle the openvswitch config well. so let's tell it to just configure
+# the actual connected interface (which on this system is eth2) for us
+# and leave everything else alone.
+ansible_ifcfg_whitelist: ['eth2']
 
 # this is a powerful machine, can handle more openQA workers
 openqa_workers: 8

inventory/host_vars/qa09.qa.fedoraproject.org

@@ -6,5 +6,13 @@ nrpe_procs_warn: 1400
 nrpe_procs_crit: 1600
 eth0_ip: 10.5.124.239
 gw: 10.5.124.254
+
+# openQA tap worker hosts (like this one) do stuff with ifcfg that base
+# doesn't understand. terrible, terrible stuff. seriously - it doesn't
+# handle the openvswitch config well. so let's tell it to just configure
+# the actual connected interface (which on this system is eth0) for us
+# and leave everything else alone.
+ansible_ifcfg_whitelist: ['eth0']
+
 # this is a powerful machine, can handle more openQA workers
 openqa_workers: 10

inventory/host_vars/qa14.qa.fedoraproject.org

@@ -6,5 +6,13 @@ nrpe_procs_warn: 900
 nrpe_procs_crit: 1000
 eth0_ip: 10.5.131.22
 gw: 10.5.131.254
+
+# openQA tap worker hosts (like this one) do stuff with ifcfg that base
+# doesn't understand. terrible, terrible stuff. seriously - it doesn't
+# handle the openvswitch config well. so let's tell it to just configure
+# the actual connected interface (which on this system is eth0) for us
+# and leave everything else alone.
+ansible_ifcfg_whitelist: ['eth0']
+
 # this is a powerful machine, can handle more openQA workers
 openqa_workers: 10