Move manager.instance to the part of code which doesn't get preempted 

/reviewed-on nexedi/slapos.core!21

The logic implemented on client is (for started instances):

Notify instances are available to master
run promises if fail report error
Notify instances are started to master

The change on this commit prevent us to consider the notification of availability as "no error", as promise can still fail, so we just ignore and let the decision to suspend for the next run.

This makes the ease to get the message, on certain special cases.

See ERP5 commit 8995b64508f6d7c0fee1ef475de3dc3145ddc84d.

  This version include initial plugin system and cgroups integration.

  We also included a significant change for provide the partitions network informations, which now
  don't require to come from slapos master anymore, so it is possible to make local extensions.

This report list the usage of resilience instances per user, and give the % of the adoption.

This report lists the Computers that host at least one instance of a cluster which is returning error (see Related ticket columns). This list also the utilisation ratio of the partitions, and as well the percentage of the partitions inside which are reporting errors.

This report aims to help on Support Decision Process to eliminate the biggest amount of errors.

-  Move all Computer attribute initialization into Computer constructor
-  Rename 'construct' to 'format' because we are in slap.format
-  Introduce Manager base class
-  Implement CGroupManager to take care of cgroup's cpuset

-  Add TUN class
-  Check Interface is a bridge before performing bridgey operations
-  Replace outdated 'tunctl' binary with 'ip tunctl'

  This raise my lead to a significant change on the API, which
  may require to refactor several tests, to this change keep s a
  backward compatible API.

This reverts commit 889711ed

This approach to solve checking of commands by logging is wrong. It can lead to malicious commands being executed so this test is no longer reliable.

We just should not test issued commands - test should use an isolated environment and then inspect the impact.

/reviewed-on nexedi/slapos.core!15

Fix parameters order to use request()

/cc @rafael @alain.takoudjou 

/reviewed-on !13

When creating a TAP interface we try to add it to a bridge no mater what kind of the external interface is. Check whether it is a bridge first.

/reviewed-on nexedi/slapos.core!12

slapos/cli/entry.py, slapos/cli/format.py, slapos/format.py: Do not double the log output when "slapos node boot" failed.

Pre-check if the instance is allocated before check if it was allocated on the right place. This problem seems unreproducible on a running instance.