Shorten deployment time with:
- Remove redundand roles if duplicated by a dependency and vice versa
- When a member of k8s-cluster, always install docker as a dependency
  of the etcd role and drop the docker role from cluster.yaml.
- Drop etcd and node role dependencies from master role as they are
  covered by the node role in k8s-cluster group as well. Copy defaults
  for master from node role.
- Decouple master, node, secrets roles handlers and vars to be used w/o
  cross references.

Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>

Fix handler triggering for kubelet restart

Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>

Check only for AWS, wrote some docs on actually using AWS

Fix resolv.conf search/nameserver

* Ensure additional nameserver/search, if defined as vars.
* Don't backup changed dhclient hooks as they are going to be
  executed by dhclient as well, which is not what we want.
* For debian OS family only:
- Rename nodnsupdate hook the resolvconf hook to be sourced always
before it.
- Ensure dhclient restarted via network restart to apply the
nodnsupdate hook.
* For rhel OS family, the fix TBD, it doesn't work the same way.

Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>

Revert regressions

This reverts commit f35e5e86.

This reverts commit 977f82c3.

Adjust collect-info playbook

pass cloud provider flag in all cases, not just openstack

Cleanup collected artifacts,
drop unrelated files/commands.
Always install gitinfos script to binaries for external
use.

Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>

remove host ca-certs, as they aren't necessary

Fix gen-gitinfos.sh

Fix the error gen-gitinfos.sh: 57: [: foo: unexpected operator

Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>

Fix resolv.conf search/nameserver

Rename nodnsupdate hook the resolvconf hook to be sourced always
before it.
Ensure dhclient restarted via network restart to apply the
nodnsupdate hook.
Ensure additional nameserver/search, if defined as vars.

Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>

Restart kubelet if launcher changed

Fixes #409

Fix etcd play to only run on non-k8s-cluster nodes

This decreases the time required to deploy a cluster with
3 nodes, but none are standalone etcd roles.

Copy hyperkube cni plugins optionally for calico deployment

upstreaming to k8s

Fix init scripts for etcd. Fixes #383

Fixes Ubuntu 14.04 deployment of etcd.

force fact check of all nodes first thing

OpenStack minor fix

Fix resolvconf executable discovery

If resolvconf was installed and then removed, the file
/etc/resolvconf/resolv.conf.d/head remains in the filesystem

- change discovery of 'resolvconf' executable to check if it
  can be located with 'which resolvconf' command or not.

Hyperkube from CoreOS now ships with all binaries required for
calico and flannel (but not weave). It simplifies deployment for
some network plugin scenarios to not download CNI images.

TODO: Optionally disable downloading calico to /opt/cni/bin

Collect ansible logs as artifacts

Extends https://github.com/kubespray/kargo/pull/406



Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>

Fix calico-node service unit