devops.error.DevopsCalledProcessError: Command '/etc/init.d/ntp stop' returned exit code 1

Actually the command should not work:

'/etc/init.d/ntp stop' returned exit code 1 while expected [0]

The NTPD was rebuild to fix the LP #1585751 and now it should start/stop using Upstart.
See the commit and description with explanations at the link
https://review.fuel-infra.org/#/c/25388/

Please consider to update the test and using Upstart for managing NTP.

Albert, if you have introduced the new functionality, you have to support it in tests!
Why QA team have to do such alignment w/o any significant notification?

It's quite hard to reproduce the bug.
It's a race conditions.
I just noticed from the logs, that there is attempt to stop ntp service, which actually has been stopped.
And at this point we got the "ntp: stop: Unknown instance" message.

The root cause of the issue is race conditions.
Configuring network interfaces (if-up script) follows to ntpd restart.

Here the details:
When one interface is restarting ntpd (and the service is actually in the state stop/waiting, but not starting yet), a second interface is go to restarting to we got the case when the second interface are trying stopping (and starting) the ntpd interface at the same time.
As the result the "ntp: stop: Unknown instance" message is appeared.

Nothing wrong with that, the service will successfully start and run later.

Nevertheless attempt to sync time with NTP (the following commands):
--------
service stop ntp
ntpdate ...
service start ntp
-------
when interfaces are restarting follow to the issue:

1. Interface one restart ntpd (the ntpd service has been stopped bun not started yet).
2. The command `service stop ntp` will attempt to stop ntpd, but got the error (the service has been stopped).
3. Interface one (start the ntpd service). Ntp daemon is in running state.
4. ntpdate command attempt to sync time (- will failed because nptd is running now)
5. service start ntp ( - will failed due to the service has been running).

I am curious, is this scenario possible in fields?

Since the issue has not been discovered in the field (customers have not reported yet), the case when interfaces are restarted and ntp restarted and caught the race conditions are very rare.
Change the bug severity to high.
We agreed, that @vkhlyunev will try to reproduce the bug (or scenario) on HW lab.

Move the bug to Vladimir @vkhlyunev waiting for test results.

Not reproduced on baremetal - its ok to move to 9.2

Moving the bug to the 9.2 will enable us update scripts and avoid racing in management ntp.

Reproduced on today swarm 9.1 snapshot #323,
When I try to revert snapshot for failed swarm test on:
http://paste.openstack.org/show/583477/

failed test:
https://product-ci.infra.mirantis.net/job/9.x.system_test.ubuntu.command_line/77/consoleFull

https://product-ci.infra.mirantis.net/view/9.x_acceptance/job/9.x.acceptance.ubuntu.load/17/console