- name: disable resolv.conf control from NM
  ini_file: dest=/etc/NetworkManager/NetworkManager.conf section=main option=dns value=none
  notify:
  - restart NetworkManager
  when: ansible_distribution_major_version|int >=7 and nmclitest|success and ( not ansible_ifcfg_blacklist)
  tags:
   - config
   - resolvconf
   - base
   - ifcfg

$ git grep ansible_ifcfg_blacklist | grep -E '(taskotron|resultsdb)'
inventory/host_vars/resultsdb-stg01.qa.fedoraproject.org:ansible_ifcfg_blacklist: true
inventory/host_vars/resultsdb01.qa.fedoraproject.org:ansible_ifcfg_blacklist: true
inventory/host_vars/taskotron-stg01.qa.fedoraproject.org:ansible_ifcfg_blacklist: true
inventory/host_vars/taskotron01.qa.fedoraproject.org:ansible_ifcfg_blacklist: true

<puiterwijk> Okay, taskotron down it seems. Looking
<kparal> it's up, just ssl seems not working
<puiterwijk> kparal: that means it's not up.
<puiterwijk> Since http is also handled by proxies
<kparal> interesting, just a while ago http returned with 302 to https
<puiterwijk> yes
<puiterwijk> That's done by the proxies
<puiterwijk> There's no communication to taskotron01
<puiterwijk> Its VPN link is offline
<puiterwijk> Ah. The resolv.conf is broken
<puiterwijk> kparal: running base role
<kparal> is that something we caused somehow?
<puiterwijk> Well, somehow the NetworkManager config got lost
<puiterwijk> We configure NM to not mess up the resolv.conf
<kparal> I'm not aware of any action that would cause that
<kparal> there was a dnf upgrade yesterday, but NM was not involved
<puiterwijk> Okay, the network config file is incorrect
<puiterwijk> Aug 10 09:51:31 taskotron01.qa.fedoraproject.org openvpn[2115]: Initialization Sequence Completed
<puiterwijk> and it's working again
<puiterwijk> https://taskotron.fedoraproject.org/
<kparal> puiterwijk: so all we needed to do was to run the full playbook (which would include the base role) and it would get fixed, right?
<kparal> just making sure I know what to do next time
<puiterwijk> kparal: well, not really, since normally NM should not do dns in our instances, but for some reason taskotron is setup to do so
<puiterwijk> I can look at that in a few, first looking at something else
<kparal> jskladan: ^^ taskotron landing page and buildmaster is up now. resultsdb is not, though. is that something you're going to fix on your side?
<puiterwijk> kparal: likely resultsdb server has the same issue
<puiterwijk> yep
<puiterwijk> kparal: coming back
<kparal> thanks a lot puiterwijk 
<kparal> puiterwijk++
<puiterwijk> same story: please check your NetworkManager.conf why it's not excluding dns
<kparal> my cookies are depleted
<puiterwijk> :)
<kparal> puiterwijk: the reason seems because we have ansible_ifcfg_blacklist: true for those machines
<kparal> I can ask tflink about that, I have no idea why it is setup that way
<puiterwijk> ack
<kparal> puiterwijk: what exactly did you do to keep NM out of dns? I don't see the config file changed
<puiterwijk> kparal: for now I've set the /etc/sysconfig/network-scripts/ifcfg-eth0 to include the search domain and both DNS servers
<puiterwijk> So that NM generates a correct resolv.conf
<kparal> ok