Comment 29 for bug 1656150

I'm having a similar problem with fuel-community-10.0.

Some nodes do not start mcollective. The workaround described above (running /etc/rc.local manually) unmasked the mcollective service, but it is not able to start it.

systemctl status mcollective prints:

# systemctl status mcollective
● mcollective.service - The Marionette Collective
   Loaded: loaded (/lib/systemd/system/mcollective.service; enabled; vendor preset: enabled)
   Active: inactive (dead)

Mar 08 09:12:33 bootstrap systemd[1]: mcollective.service: Failed to send unit change signal for mcollective.service: Transport endpoint is not connected
Mar 08 09:31:26 bootstrap systemd[1]: mcollective.service: Failed to send unit change signal for mcollective.service: Transport endpoint is not connected
Mar 08 09:37:20 bootstrap systemd[1]: mcollective.service: Failed to send unit change signal for mcollective.service: Transport endpoint is not connected

The "Transport endpoint is not connected" message is printed for each every time /etc/rc.local was executed.

Manually starting mcollective via systemctl start works, but the hosts are printed as 'bootstrap':
[root@fuel etc]# mco ping
bootstrap time=34.90 ms
17 time=37.08 ms
master time=39.54 ms
21 time=43.94 ms
25 time=45.40 ms
bootstrap time=46.36 ms
22 time=47.25 ms
23 time=63.95 ms

---- ping statistics ----
8 replies max: 63.95 min: 34.90 avg: 44.80

They are also not recognized during network connectivity check stating that nodes 11 and 20 are not available via mcollective.