* ensure exit_json does not fail from stderr=False
- do a little bit of safety-checking in exit_json to not try to .splitlines() on a boolean
- remove the stderr=boolean from uri.py, this is the only spot that uses it (at least so obviously)
- add unit tests that call exit_json. These are useless because the stderr parsing is in _execute_module and is difficult to mock; deleting these tests after the commit.
* remove added unit tests per prev commit
exit_json doesn't do the param parsing, that is buried deep inside _execute_module.
Fixes#22575 - issue under new exec wrapper where unconstrained handle inheritance (for stdin) caused WinRM to block on breakaway processes. Uses explicit handle inheritance to ensure that only stdin read handle gets inherited. Adds test to ensure that async is actually async.
* removes cli functions from shared lib
* adds cli functions to junos_netconf module
* statically pins junos_netconf to cli transport
* all other modules use netconf transport
* adds command rpc function to junos shared
* Fixes for junos_config errors
* Check transport settings for core Junos
* Don't pop from the same list you iterate over
* use of persistent connections are now explicitly enabled in junos
* modules must now explicitly enable persistent connections
* adds rpc support to junos_command
fixes#22166
The provider arguments where not correctly being mapped in the action
plugin for eapi transport. This will now correctly map existing
arguments as well as set the correct defaults for values.
fixes#22442
* Move retry logic into _ssh_retry decorator, and apply to exec_command, put_file and fetch_file
* Update tests to reflect change
* Move _ssh_retry to _run, and update tests to reflect
* piped should use exec_command instead of removed _exec_command
* Rework tests to support selectors instead of select.select
* WIP: wait_for_connection: Wait for system to be reachable
This action plugin allows to check when a system is back online and
usable by Ansible.
As an example, when doing a SysPrep and running Enable-WinRM.ps1, it
takes between 10 to 20 seconds between the WinRM TCP port to open, and
it actually being able to server Ansible requests. This time is variable
and depends on the boot process.
Current implementation is specific for Windows (WinRM) only, this will
be fixed shortly.
This fixes#19998
* Support other transport types
* Various improvements
- Fix reported typo
- Add transport_test support in accelerate plugin
- Ensure port is an integer
* Improve examples
* Small fixes
- Use correct ConfigureRemotingForAnsible.ps1 script name
- Only use win_ping when remote shell is known to be Powershell
- Add integration tests to CI framework
* clarify facts assignment for several corner cases
run_once/delegate_facts:
now delegate_facts > run_once, previously run_once always published facts to all hosts in play
include_vars/delegate_to:
now include_vars allows to delegate to a specific host
also fix task_vars exception in delegate_facts/loop as var was removed
fixes#15365
* removed unused loop_var
* Fix vyos signatures to match new versions
* Fix test cases referring to 'updates' instead of 'commands'
* I think this is an artifact of `connection: network_cli`?
* win_copy: Add force parameter and check-mode support
The rationale behind this is that if you're working with +3GB files,
creating the checksum takes a lot of time, which we can avoid by simply
testing if the file exists.
I also took the liberty to put the various parameters together. It
probably takes a (neglible) performance hit but makes the code a bit
easier to inspect/work with, as its closer to all other windows modules.
On a normal run, the action plugin does a local checksum of the source
and a remote checksum of the destination. And afterwards, the module
will do another remote checksum of the copied source, a remote checksum
of the original destination, and another remote checksum of the copied
destination.
On a very huge file (think 4GB) that means 5x reading the complete file
(if you have a large cache you may get away with it, otherwise you're
doomed !).
This patch will ensure with `force: no` that not checksums are being
performed.
* Moving presence check before remote checksum
* Adapted to wishes
* Even more performance improvements
Fix 'task name is not templated in retry callback'
Add a task_name property to TaskResult that knows to
check in TaskResult._task_fields.
Add integration test for v2_retry_runner callback
Fixes#18236
The requests python module is needed, however it is not a dependency of
the python-winrm package. The python-winrm package does require
python-requests_ntlm, which does not seem to pull python-requests.
So for the time being (until Red Hat fixes their package) give a more
informative error message.
trying to import from the wrong shared lib, this patch changes
network_common to connection. Also fixes a bug found in the nxos
action plugin to detect transport
fixes#21829
The nxos action handler did not map the nxapi value to the provider
argument properly. There as an additional fix in the nxos shared lib
to properly detect nxapi being set
added new base class for file based cache plugins as 99% of code was common
now also catches unexpected decoding exceptions
allows per module file modes and encoding
moved jsonfile code to base
This allows getting the Vault token from the `VAULT_TOKEN` env var or
from the file `$HOME/.vault-token`, as both of these are understood by
the Vault CLI and are a common place to put Vault tokens. This allows
avoiding hard-coding a Vault token into playbooks or having to include
lookups.
`HOME/.vault-token` is nice because a user can authenticate with the CLI
using `vault auth` and then the token will be stored in
`$HOME/.vault-token`. If we read this file, then we allow someone to do
`vault auth` "out of band" to set up Vault access.
* fixes error where eos would close the cli shell
* fixes network_cli connection plugin to check before calling open_shell()
* fixes json commands being sent over eapi
This patch adds some checks on the path that is accessed as a container,
making sure it looks like one. It implements the connection method and
add adaptations to the modern way of writing connections for Ansible.
It also rewords docs and vars to use the nspawn terminology instead of
chroot.
This commit adds a connection driver built on top of systemd-nspawn.
This is similar to the existing `chroot` driver, except that nspawn
offers a variety of additional services. For example, it takes care of
automatically mounting `/proc` and `/sys` inside the chroot environment,
which will make a variety of tools work correctly that would otherwise
fail.
You can take advantage of other system-nspawn features to perform more
complicated tasks. For example, on my x86_64 system I have a Raspberry
Pi disk image mounted on `/rpi`. I can't use `chroot` with this because
the binaries contained in the image are for the wrong architecture.
However, I can use the systemd-nspawn `--bind` option to automatically
insert the appropriate qemu-arm binary into the container using an
inventory file like this:
pi ansible_host=/rpi ansible_nspawn_extra_args='--bind /usr/bin/qemu-arm --bind /lib64'
See http://www.freedesktop.org/software/systemd/man/systemd-nspawn.html
for more information about systemd-nspawn itself.
* removes unneeded supports_multiplexing var
* refactors terminal_prompts_re to terminal_stdout_re
* refactors terminal_errors_re to terminal_stderr_re
* updates network_cli unit test cases
The network_cli plugin would return immediately if an error was
detected. This patch will force the connection plugin to still try to
detect the current prompt even if an error is found.
Moving it to after the blocks where per-item results are calculated,
as it's not used there and causes quite a performance hit being there.
Fixes#21340
* refactors supports_sessions to a property
* exposes supports_sessions as a toplevel function
* adds open_shell() to network_cli
* implements open_shell() in eos action plugin
allows user to force persistent connection to close, needed for when
you want to benefit from changes applied to the current play but persistent connections
prevent them from being realized.
This change will now check the cli context after a module runs and if
the cli is still in config mode it will exit config mode. Also fixes a
minor issue with converting list of commands to a dict
fixes#21481
I have from time to time a need of random password without
wanting to write them down (one example is mailman list creation,
that requires a password to be given to be sent to the list owner).
But using /dev/null do not return null, but the empty string, which
doesn't generate a password at all and so do not achieve my use case.
hashlib hashes operate on byte strings. When given a text string on
Python3, hashlib backtraces. When given a text string on Python2,
hashlib will backtrace if the string contains non-ascii characters.
Encode the text string to utf-8 prior to hashing to avoid this problem.
Fixes#21452
* removes unused code
* removes module_utils/local.py
* removes plugins/action/network.py
* removes action_handler from connection plugins
* removes code to use action_handler in task_executor
* updates action plugins to subclass from normal
This completes the refactor of the nxos modules to use the persistent
connection. It also updates all of the nxos modules to use the
new connection module and preserves use of nxapi as well.
* updates all iosxr modules to support persistent socket
* adds iosxr action plugin to connect to device
* adds exec_command() to iosxr shared module
* fixes iosxr_config and iosxr_template local action
* update all unit test cases
* adds base test module for iosxr module testing
* fixes issue with load_provider() not checking for an existing key
* adds updates to eos_config results key
* lots of minor syntax fixes in eos shared module
* adds eos_argument_spec to eos_eapi
fixes#21402
When loading the params in the local action, the provider key was
inadvertently being loaded as well. This created the circular
reference. The load_provider() method will now check for the provider
key and skip it when encountered.
fixes#21399
* updates eos modules to use persistent connection socket
* removes split eos shared module and combines into one
* adds singular eos doc frag (eos_local to be removed after module updates)
* updates unit test cases
* adds conversion to command dict into ios shared lib
* fixes hardware commands in ios_facts
* adds network_os value to play_context in ios action plugin
fixes#21190
* updates all vyos modules to use socket connection
* adds vyos local action handler
* adds exec_command() to vyos
* updates vyos_config local action
* update unit test cases
* add base class for testing vyos modules
* updates all ios modules to support persistent socket
* adds ios action plugin to connect to device
* adds exec_command() to ios shared module
* fixes ios_config and ios_template local action
* update all unit test cases
* adds base test module for ios module testing
* Add a surrogate_then_replace error strategy to keep to_bytes from tracebacking by default
* Port all code that explicitly used surrogate_or_replace to surrogate_then_replace
* replaces persistent connection digest with _create_control_path()
* adds _ansible_socket to _legal_inputs in basic.py
* adds connection_user to play_context
* maps remote_user to connection_user when connection is local
* maps ansible_socket in task_vars to module_args _ansible_socket if exists
- centralized skipping
- also fixed module name broken by previous refactor
- let action modules handle async processing
- moved async into base action class's module exec
- action plugins can now run final action as async
- actually skip copy if base skips
- fixed normal for new paths
- ensure internal stat is never async
- default poll to 10 as per docs
- added hint for callback fix on poll
- restructured late tmp, now a pipeline query
- moving action handler to connection as networking does
- fixed network assumption invocation is always passed
- centralized key cleanup, normalized internal var
- _supress_tmpdir_delete now in _ansible_xxx and gets removed from results
- delay internal key removal till after we use em
- nicer tmp removing, using existing methods
- moved cleanup tmp flag to mking tmp func
At the moment, this change will use EPoll on Linux, KQueue on *BSDs,
etc, so it should alleviate problems with too many open file
descriptors.
* Bundle a copy of selectors2 so that we have the selectors API everywhere.
* Add licensing information to selectors2 file so it's clear what the
licensing terms and conditions are.
* Exclude the bundled copy of selectors2 from our boilerplate code-smell test
* Rewrite ssh_run tests to attempt to work around problem with mocking
select on shippable
Fixes#14143
* Make the module_utils path configurable
* Add a config value to define the path site module_utils files
* Handle module_utils that do not have source as an error
* Make an integration test for module_utils envvar working
* Add documentation for the ANSIBLE_MODULE_UTILS config option/envvar
* Add it to the sample ansible.cfg
* Add it to intro_configuration.
* Also modify intro_configuration to place envvars on equal footing with
the config options (will need to document the envvar names in the
future)
* Also add the ANSIBLE_LIBRARY use case from
https://github.com/ansible/ansible/issues/15432 so we can close out
that bug.
* Check for the prompt as a substring of the output
sudo sometimes spits out warnings to stdout before getting to the
password prompt. Account for that when trying to match a password
prompt.
Fixes#20858
We need to use ssh_executable instead of hardcoding ssh in the command
we run but we need to use "ssh" when we lookup the value of the
{command}_extra_args variable. Do this by leaving binary as "ssh" and
only expanding when we place it into b_command.
Fixes#20862
* A method to validate and alter the ssh control path automatically.
* First tries %C to use the shortened hash
* On further failure, it removes section by section from the original path
* Fix hostname
* Implement bcoca's suggested changes
* Remove unused option
* Remove unused class var
* Use to_string to avoid unicode error
* Switch from to_text to to_bytes
* Update the example config for the new controlpath feature
This version just gets the relevant paths from PluginLoader and then
uses the existing imp.find_plugin() calls in the AnsiballZ code to load
the proper module_utils.
Modify PluginLoader to optionally omit subdirectories (module_utils
needs to operate on top level dirs, not on subdirs because it has
a hierarchical namespace whereas all other plugins use a flat
namespace).
Rename snippet* variables to module_utils*
Add a small number of unittests for recursive_finder
Add a larger number of integration tests to demonstrate that
module_utils is working.
Whitelist module-style shebang in test target library dirs
Prefix module_data variable with b_ to be clear that it holds bytes data
added better way of adding warnings to return data
backwards compatible if warnings key already exists
added deprecations made iface more generic
changed to enforce type per item
added logging of warnings/deprecations
also display deprecations by default
* update eos_config to use eapi exclusively and remove cli transport
* add unit test cases for eos_config
* updates action plugin to handle both eapi and network_cli connections
* Add jinja2 groupby filter override to cast namedtuple to tuple. Fixes#20098
* Address some of the requested changes
* Quoting
* Print the python path and version
* Be less explicitly verbose, rely on implicit verbosity
* checks if signal hander is set and sets it if not (will be set if coming
from ansible-connection)
* will now timeout long running commands based on DEFAULT_TIMEOUT setting
* removed bad iteration from execute meta
most of the tasks should not be iterated over, others needed to include unreachable hosts
fixes#19673
* corrected host var
Because we add the names of all filters to the callable whitelist used
by safe_eval, adding a filter named type makes it so code calling "type()"
gets eval'd. We can't think of a way to exploit this but it's
sufficiently sketchy that we're renaming it in case someone smarter than
us can think of a problem.
When you become: with synchronize and docker it sets the rsync-path to
"sudo rsync" to launch rsync on the server as root. Unfortunately due to
docker exec doing stricter argument parsing than ssh this fails to
launch rsync on the server and the sync fails.
For docker though we don't need to launch rsync with sudo we can simply
docker exec -u <user> and rsync as normal to get around the problem.
Closes#20117
The behavior now matches GNU diff.
Fixes#14094.
Example of output before this change:
TASK [healthchecks.io : hourly healthchecks.io ping] ***************************
changed: [ranka]
--- before: /etc/cron.hourly/mg-healthchecks-dot-io
+++ after: /tmp/tmpOTvXTw
@@ -1,2 +1,2 @@
#!/bin/sh
-curl -sS https://hchk.io/xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx > /dev/null+curl -sS https://hchk.io/xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx > /dev/null
after this change:
TASK [healthchecks.io : hourly healthchecks.io ping] ***************************
changed: [ranka]
--- before: /etc/cron.hourly/mg-healthchecks-dot-io
+++ after: /tmp/tmpOTvXTw
@@ -1,2 +1,2 @@
#!/bin/sh
-curl -sS https://hchk.io/xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx > /dev/null
\ No newline at end of file
+curl -sS https://hchk.io/xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx > /dev/null
The added unit tests contain more examples.
This commit also takes care to avoid "no newline at EOF" warnings when
no_log is in effect, and also when modules return dicts rather than
strings. (It also removes trailing whitespace from using json
serialization when diffing dicts, because I hate trailing whitespace in
Python source files, even if they're test files.)
* updates the ios_config module to use the network_cli plugin
* updates the local action plugin to derive from network
* add unit test cases for ios_config
* updates the deprecated ios_template module to use network_cli
* adds unit test cases for ios_template
* adds check for provider argument and displays warning message
can be per run or per host, also aggregate or not
set_stats action plugin as reference implementation
added doc stub
display stats in calblack
made custom stats showing configurable
* net_config now subclasses action plugin network
* net_template now subclasses action plugin network
This will break existing modules until those modules have been refactored.
In some cases it is desirable to have a send only function that doesn't
wait for the response from the CLI (such as reloading a device). This
adds a new key to the command json string sendonly that will
achieve this behavior.
* new lookup module: mongodb lookup
* fix versionadded for MongoDB Lookup
* tests should run again
* removed use of basestring
* we don't use iteritems anymore
* run tests again
* run tests again2
* run tests again3
* run tests again4
Added iocage connector that extends the jail connector. Uses iocage to translate iocage tags or UUIDs/partial UUIDs to the actual jail name and then uses the jail connector for actual functionality.
This plugin can be used with the lpass cli interface for lastpass.
[lastpass-cli](https://github.com/lastpass/lastpass-cli)
Example:
Add a lookup to your playbooks/variables somewhere:
```
some_variable: "{{ lookup('lastpass','Some Lastpass entry name or ID', field='username') }}"
```
Usage:
* start a lpass session prior to using ansible
* run ansible
* logout when finished
```
lpass login user@domain.com
ansible-playbook foo.yml
lpass logout
```
Since we no longer use a post-validated task in _process_pending_results, we
need to be sure to template fields used in original_task as they are raw and
may contain variables.
This patch also moves the handler tracking to be per-uuid, not per-object.
Doing it per-object had implications for the above due to the fact that the
copy of the original task is now being used, so the only sure way is to track
based on the uuid instead.
Fixes#18289
This is a redesign in how plugins call _remote_checksum().
- _remote_stat() has been modified to report the real error as
AnsiblError
- Action plugin **unarchive** calls _remote_stat() directly instead of
_remote_checksum()
- Action plugin **unarchive** also handles the exceptions directly
- Ensure get_exception() returns native text
Two other action plugins, **template** and **fetch**, also do a remote checksum.
In **template** we already call _remote_stat(), just like we now do for
unarchive, in **fetch** we do call _remote_checksum() and we make the
exact same mistake as the unarchive plugin. So that one could use a
redesign as well.
This fixes#19494
Before:
```
[dag@moria ansible.testing]$ ansible-playbook -v test137.yml
Using /home/dag/home-made/ansible.testing/ansible.cfg as config file
PLAY [localhost]
******************************************************************************************************
TASK [unarchive]
******************************************************************************************************
fatal: [localhost]: FAILED! => {"changed": false, "failed": true, "msg":
"python isn't present on the system. Unable to compute checksum"}
PLAY RECAP
******************************************************************************************************
localhost : ok=0 changed=0 unreachable=0
failed=1
```
After:
```
[dag@moria ansible.testing]$ ansible-playbook -v test137.yml
Using /home/dag/home-made/ansible.testing/ansible.cfg as config file
PLAY [localhost]
*************************************************************************************************************
TASK [unarchive]
*************************************************************************************************************
fatal: [localhost]: FAILED! => {"changed": false, "failed": true, "msg":
"Failed to get information on remote file (/tmp/): sudo: unknown user:
foobar\nsudo: unable to initialize policy plugin\n"}
PLAY RECAP
*******************************************************************************************************************
localhost : ok=0 changed=0 unreachable=0
failed=1
```
For devices that do not support mutliplexing, we cannot automatically
determine the network os. This removes the os guess static method
from the terminal plugin. For this devices, the network_os
value must be configured
This fix ensures that if there are specific module errors (in our case
the python interpreter was not found) then command and shell returns a
proper error.
It also fixes a few other imperfections that we noticed during
troubleshooting:
- Return the real RC if it were available
- Improve a dictionary evaluation using .get()
- Return an RC of -1 if it is unknown (instead of returning 0)
This fixes#18846
Brian Coca
Matt Martz
Ted Timmons
Peter Sprygada
Matt Davis
jralbert
Brian Coca
Ganesh Nalawade
Ricardo Carrillo Cruz
Matt Clay
Dag Wieers
Nathaniel Case
Paul Arthur
Adrian Likins
jctanner
Toshio Kuratomi
James Cammarata
Marc Abramowitz
René Moser
Thomas Szymanski
Lars Kellogg-Stedman
James
Michael Scherer
Patrick Deelman
Ssawa
Pilou
Samuel Boucher
Doug Bridgens
hloeffler
Peter Sprygada
John R Barker
jonathanbouvier
David Barroso
Marcos Diez
Andrew Gaffney
Matt Davis
Jamie Lennox
Ryan Brown
Marius Gedminas
Alexander Gubin
Tom Booth
Timur Vasyunin
slohse
azenk
Sorin Sbarnea
Chris Houseknecht