* return a proper result set for getmaster/getslave
when not on a master/slave.
This allows for a cleaner error handling.
* A more uniform return of result keys for getmaster/slave
got connected. It is possible for an intiator to successfully connect to a
target, whilst getting no LUN's back. If no devicenodes get detected, it makes
more sense to return an empty list than plainly None.
This potentially avoids further tasks to have to check if devicenodes is
iterable.
This fixes a corner case where ini files live in a subdir
of the main inventory directory.
Reproducing the original error:
mkdir -p inventory/ini
cat > inventory/ini/hosts << EOF
[www]
www1
EOF
$ ansible -i inventory/ all -m ping
ERROR! 'all'
(or without the [www] group, it would complain about 'ungrouped')
This addresses a specific case with multiple vars files
in a group_vars/${groupname}/ directory where one of those files
is empty, which returns None instead of an empty dict.
inventory vars: make loading from a directory obey the same rules as
when checking the base paths, looking at the file name extensions
as defined in CONSTANTS.YAML_FILENAME_EXTENSIONS
Fixes Github issue #11017
and without hosts and vars
Without this patch, the simplified syntax is triggered when a group
is defined like this:
"platforms": {
"children": [
"cloudstack"
]
}
Which results in a group 'platforms' with 1 host 'platforms'.
modified: lib/ansible/inventory/script.py
Only print a blank line between plays when also doing --list-hosts and/or
--list-tasks, otherwise this output just a long list of blank lines, one for
each play.
This avoids a stale situation where name/path contains some impossible path,
but gets configured (faultly) in fstab, and the module only fails after that,
when creating that path.
Avoid resolving a pattern that is a plain host. When matching a hostname in the
hosts_cache, just use the host object from there.
When running a task on say 750 hosts, this yields a huge improvement.
When no vg_options are passed to the module, 'vg_options' still exists
in the module.params dict with a value of None, so the default empty string in
the get method is never used. None cannot be "splitted", which backtraced.
In some cases, where a host is mentioned in multiple groups, and those
groups are referenced in multiple ini files, a group could still contain
multiple instances of a group in its host,groups list, where only one of them
is the right group, that exists in the inventory.
rewrite deepcopy in util.merge_hash and just iterate
on an inventory with 500 groups and 800 hosts this brings back the
inventory initialisation from 13s to 3s (with hash_behaviour=merge)
Split out parsing of vars files to per host and per group
parsing, instead of reparsing all groups for each host. This enhances
performance.
Extend vars_plugins' API with two new methods:
* get host variables: only parses host_vars
* get group variables: only parses group_vars for specific group
The initial run method is still used for backward compatibility.
Parse all vars_plugins at inventory initialisation, instead of
per host when touched first by runner. Here we can also loop through
all groups once easily, then parse them.
This also centralizes all parsing in the inventory constructor.
modified: bin/ansible
modified: bin/ansible-playbook
modified: lib/ansible/inventory/__init__.py
modified: lib/ansible/inventory/vars_plugins/group_vars.py
Make sure all hosts and groups are unique objects
and that those are referenced uniquely everywhere.
Also fixes test_dir_inventory unit tests which were broken after previous
patches.
modified: lib/ansible/inventory/dir.py
Fixes a bug where vault_password parameter was not passed through in
_load_vars_from_folder()
modified: lib/ansible/inventory/vars_plugins/group_vars.py
- extra properties were not set at creation, only when updating
which can be overlooked when running the module from more than 1 node...
- fix bas var as time_until_up didn't get used
- checks if address already assigned to other node name
- add description for node
- check for node addres changes
- add missing code "node exists, potentially modify attributes"
by ensuring all basedirs, plugin paths and extra
paths are handled as absolute paths and are checked
to not add any doubles.
This fixes the corner case where e.g. the user has
an additional plugin path configured to a dir
relative to his playbooks or inventory location,
which also matches the _plugin subdir relative to
one of the basedirs in the play.
For most plugins this doesn't show as an obvious issue
except for callback_plugins, which might fire more
than once. Other plugins (inventory and template
plugins) might unnecessarily be ran twice.
e.g. ansible.cfg has
callback_plugins = ./plays/callback_plugins
and plays/ contains a playbook file:
.
├── ansible.cfg
├── inventory
└── plays
├── callback_plugins
│ └── timestamp.py
└── site.yml
modified: lib/ansible/utils/plugins.py
to ensure consistent behavior, hosts should look like this:
hosts: webservers:&boston:!rack42
So when applying the host selectors, run those without the "&" first,
then the &s, then the !s.
Closes#3500
Use case: e.g. dual homed hosts on production en management network
The inventory_hostname is the regular host name and matches the
dns name on the production network; ansible connects to the host
through a management network; the dns name on the management network
is standardized and equals ${inventory_hostname}-mgt.mynetwork.com
Now this can be configured as the default in group_vars/all:
ansible_ssh_host: {{ inventory_hostname + '-mgt.mynetwork.com' }}
