There was a traceback when setting permissions on a directory tree when
there were broken symlinks inside of the tree and follow=true. chmod -R
ignores broken symlinks inside of the tree so we've fixed the file
module to do the same.
Fixes#39456
(cherry picked from commit 6b159fdb03)
* Fix for file module with symlinks to nonexistent target
When creating a symlink to a nonexistent target, creating the symlink
would work but subsequent runs of the task would fail because it was
trying to operate on the target instead of the symlink.
Fixes#39558
(cherry picked from commit 4f664f8ff6)
* vyos_interface require multiple network nodes to run
We don't have the ability to run these currently, so disable them.
The original logic was also incorrect, the tests don't pass on lab, DCI
nor single instance nodepool, so disable
https://github.com/ansible/ansible/issues/39667 tracks getting these
enabled again
* eth0 -> Gi0/0
* Correctly detect if we should run lldp
* Correctly detect if we should run lldp
(cherry picked from commit 923a81e9e5)
* removed additional check for name parameter
(cherry picked from commit 91357d07f4)
* Added empty default
Fix for issue https://github.com/ansible/ansible/issues/38482
(cherry picked from commit 60a3277103)
* [ec2_ami] Ensure name or image_id is provided for state=present (#38972)
Add integration tests for backward compatibility and ensuring name or image_id is provided
(cherry picked from commit e2aa1155ba)
Fixes#27262
(cherry picked from commit 81b2529159)
Add tests for template with non-ascii filenames
This is a test in response to #27262 but I could not provoke the error
so it only shows that the current code is working with non-ascii
filenames in this case. It doesn't show whether there's some other bug
somewhere.
(cherry picked from commit f91d961cb4)
* [cloud] ec2_vpc_route_table: ignore routes without DestinationCidrBlock
Add module warnings rather than silently skipping
* Permit warnings for routes tables containing vpc endpoints to be turned off
* Add tests to ensure a VPC endpoint associated with a route table does not result in a traceback
(cherry picked from commit da3f7a8db1)
* win_file: Handle [] in paths, fix touch in check mode (#37901)
* win_file: Handle [] in paths, fix touch in check mode
* Fixed typo for p/invoke command
(cherry picked from commit 4b57fa91d0)
* Added changelog fragment for win_file special char fix
(cherry picked from commit 61bcf4740f)
* Ansible.ModuleUtils.FileUtil - Add ability to test non file system provider paths (#39200)
(cherry picked from commit b6afe6946d)
* Added changelog fragment
* Skip self._parent on dynamic, defer to grandparent for attr lookup (#38827)
* Skip self._parent on dynamic, defer to grandparent for attr lookup
* Revert _inheritable
* Add tests for include inheritance from static blocks
Fixes#38037#36194
(cherry picked from commit 354aa8d602)
* Add changelog for #38827
* Consider parent also when comparing IncludedFile (#37083)
* Consider parent also when comparing IncludedFile
* Add new tests for IncludedFile and convert to pytest
(cherry picked from commit cdb79b0e3a)
* Add changelog for #37083
* module_common: set required parameter templar
Fix the following error (related to b455901):
$ ./hacking/test-module -m ./lib/ansible/modules/system/ping.py -I ansible_python_interpreter=/usr/bin/python
Traceback (most recent call last):
File "./hacking/test-module", line 268, in <module>
main()
File "./hacking/test-module", line 249, in main
(modfile, modname, module_style) = boilerplate_module(options.module_path, options.module_args, interpreters, options.check, options.filename)
File "./hacking/test-module", line 152, in boilerplate_module
task_vars=task_vars
File "ansible/lib/ansible/executor/module_common.py", line 910, in modify_module
environment=environment)
File "ansible/lib/ansible/executor/module_common.py", line 736, in _find_module_utils
shebang, interpreter = _get_shebang(u'/usr/bin/python', task_vars, templar)
File "ansible/lib/ansible/executor/module_common.py", line 452, in _get_shebang
interpreter = templar.template(task_vars[interpreter_config].strip())
AttributeError: 'NoneType' object has no attribute 'template'
* module_common.modify_module: templar is required
(cherry picked from commit 7908f78fa6)
* fix inventory plugin source caching
- avoid caching invetnory sources in loader in base
- same fix for yaml plugin
- idem for 'auto' plugin
fixes#37162
* fix mock dataloader func sig
(cherry picked from commit 886c4edfb9)
This offers an optimization that allows loading larger
inventories of various structure by improving the
scaling laws involved for adding hosts and groups.
The primary speed benefit is the elimination of duplicate
recusion from traversing converging paths.
(cherry picked from commit 153c9bd539)
If user does not specify any network type then set network type
to dhcp. There are additional checks around 'ip', 'netmask' and
'type' in network spec.
Fixes: #38466
Signed-off-by: Abhijeet Kasurde <akasurde@redhat.com>
(cherry picked from commit 9b139baacb)
* fix for when status codes are provided as a comma separated list (#38080)
* fix for when status codes are provided as an array of strings
* convert status codes to int, additional tests
(cherry picked from commit c38617a736)
* Added changelog fragment
* Fix 'New Vault password' on vault 'edit'
ffe0ddea96 introduce a
change on 'ansible-vault edit' that tried to check
for --encrypt-vault-id in that mode. But '--encrypt-vault-id'
is not intended for 'edit' since the 'edit' should always
reuse the vault secret that was used to decrypt the text.
Change cli to not check for --encrypt-vault-id on 'edit'.
VaultLib.decrypt_and_get_vault_id() was change to return
the vault secret used to decrypt (in addition to vault_id
and the plaintext).
VaultEditor.edit_file() will now use 'vault_secret_used'
as returned from decrypt_and_get_vault_id() so that
an edited file always gets reencrypted with the same
secret, regardless of any vault id configuration or
cli options.
Fixes#35834
(cherry picked from commit 6e737c8cb6)
* openssl_certificate: Handle dump() in check_mode (#38386)
Currently, when ones run the module in check_mode it tries to retrieve
values from the actual certificate generated in the generate() function.
Since in check_mode we call dump() without calling generate(), self.cert
is None, leading to self.cert.get_notBefore(), self.cert.get_notAfter()
and self.cert.get_serial_number() raising an error.
> NoneType' object has no attribute 'get_notBefore'
The solution is to have two way to handle dump() method, whether its run
in check_mode=True or check_mode=False leading to different way the
information is retrieved.
(cherry picked from commit 99497ce54c)
* added eos_smoke tests
* removed left over file
* added note to uncomment broken eapi test when #36919 is fixed
* uncommented fixed test, added unbecome test
* skip become tests when connection=local
(cherry picked from commit 712d30f46c)
* Eos :do not push config to device if check_mode is enabled (#37287)
* eos can not check config without config session support
* add testcase for check_mode without config session
* fix eos eapi to read use_session env var
(cherry picked from commit a1026dbce5)
* Backport/2.5/37287 - Eos do not push config is check_mode is enabled
* Close & remove paramiko connection where appropriate (#37528)
* Update unit test
(cherry picked from commit 594840c1d6)
* Put back $PATH checking in ansible-connection call (#37933)
(cherry picked from commit 169209c32a)
* issue:38167 add support for onyx version 3.6.6000 for onyx_linkagg (#38191)
Signed-off-by: Samer Deeb <samerd@mellanox.com>
(cherry picked from commit 72d42bd065)
* fragment
* issue:37307 Add support for changes in pfc output in onyx 3.6.6000 (#37651)
* issue:37307 Add support for changes in pfc output in onyx 3.6.6000
Signed-off-by: Samer Deeb <samerd@mellanox.com>
(cherry picked from commit 9dfb665e43)
* fragment
* issue:37306 Fix issue with vlan support for onyx version 3.6.6000
Signed-off-by: Samer Deeb <samerd@mellanox.com>
(cherry picked from commit a89bafce2e)
* nso_verify handle leaf-list in 4.5 and identityref (#37393)
NSO verify did not handle leaf-list value verification in 4.5 and
later due to changes made for configuration writing made.
map prefix for identityref types in verification.
(cherry picked from commit 6308047dc9)
* Remove timeout
* win_service: fix when dealing with paths with special chars and change WMI to CIM cmdlets (#37897)
* win_service: fix when dealing with paths with special chars and change WMI to CIM cmdlets
* compare username in lowercase for test
(cherry picked from commit f37a44430f)
* Added changelog fragment
* azure_rm_networkinterface: fixed issue when public ip address should not be created (#36824)
* fixed issue when public ip address should not be created
* adding test for public ip address
* fixed samples
* another fix to sample formatting
* fixed test
* fix test
* fixed test
* another attempt to fix test
* maybe it works now
* still wrong
* improved check per customer request
* removed stupid semicolon
* updated test to match main scenario
* changed ip configurations to list
* another attempt
(cherry picked from commit 89401f13f7)
* Added changelog fragment
* Fix redundant yaml error blurbs on ModArgs parse errors
Some of the AnsibleParserErrors from parsing.mod_args
are created with the obj=some_yaml_ds options but
some are not.
If they were, we don't want to add another yaml_ds to
it, because that will result in double yaml error blurbs.
And since we dont need to add info, we can just re raise it.
But if there is no ._obj, add it here so we get the extra
detail in the error message (see issue #14790) and raise
a new AnsibleParserError instance.
Fixes#36848
* cleanup existing test_tasks pep8/sanity issues
(cherry picked from commit e166946a0a)
This fix adds environment variables for connection in vmware_*
modules.
Signed-off-by: Abhijeet Kasurde <akasurde@redhat.com>
(cherry picked from commit 0ae7a0e88c)
The generated file was completely unusable by the system
therefore the fix which ensures that diffing the file
prior to changes and after only shows diffs
Furthermore the code did not work for Python 3.6
> f.writelines(to_bytes(lines, errors='surrogate_or_strict'))
E TypeError: a bytes-like object is required, not 'int'
The other modifications (lambda variable renaming) is to
comply with default flake8 rules
(cherry picked from commit 612d0d6634)
Fixes#37208
If check_mode is enabled instead of committing th config need to
discard all the chnages to cnadidate db
In case of cli to discard changes issue `rollback 0` command
and for netconf execute `discard-changes` rpc call
(cherry picked from commit 8eaa9cc938)
Fixes#36979
If `abort` is not issued in the top level session prompt
the existing session goes to pending state.
The fix is to come out of config mode by issuing `end` command
and again to same config session and execute `abort` which
`abort` is issued at the top level session prompt.
(cherry picked from commit 017ea018d0)
SGs created when a VPC ID was not specified would not necessarily
get the default egress rule, even when no explicit egress rules
were set.
Add some checks for egress rules in results from existing tests
(cherry picked from commit 98b29f8ad6)
Fixes#35993 - Changes to update_size in commit eb4cc31 made it so
the group dict passed into update_size was not modified. As a result,
the 'replace' call does not see an updated min_size like it previously
did and doesn't pause to wait for any new instances to spin up. Instead,
it moves straight into terminating old instances. Fix is to add batch_size
to min_size when calling wait_for_new_inst.
Fixes#28087 - Make replace_all_instances and replace_instances behave
exactly the same by setting replace_instances = current list of instances
when replace_all_instances used. Root cause of issue was that without lc_check
terminate_batch will terminate all instances passed to it and after updating
the asg size we were querying the asg again for the list of instances - so terminate batch
saw the list including new ones just spun up.
When creating new asg with replace_all_instances: yes and lc_check: false
the instances that are initially created are then subsequently replaced.
This change makes it so replace only occurs if the asg already existed.
Add integration tests for #28087 and #35993.
(cherry picked from commit a2b3120e85)
* Fix name parameter templating in include_role module (#36372)
An IncludedFile() object built using the original_task will have
its _task bound to the original_task. The iterative reassignment of
original_task._role_name during with_item loops leaves all returned
included_files with the same ._task._role_name (the final name from
the with_items list). This commit builds IncludedFile() objects
from an original_task.copy() to avoid the problematic binding.
(cherry picked from commit 54e70fc783)
* Test include role with items in name #36372 (#37001)
* Tests for #36372
* Tests for #36372
* Tests for #36372
(cherry picked from commit 8c4f349743)
* Add changelog for #36372
* Connection error messages are unsafe: wrap them (#37329)
* Check that connection error msg are not unsafe
* Connection error messages are unsafe: wrap them
For example, in case of error, docker connection plugin returns exception
message containing Go template. These messages weren't tagged as unsafe
and were consequently rendered:
The conditional check 'result is failed' failed. The error was:
{
'msg': u'Docker version check ([\'/usr/bin/docker\', \'version\', \'--format\', "\'{{.Server.Version}}\'"]) failed: Got permission denied while trying to connect to the Docker daemon socket at unix:///var/run/docker.sock: Get http://%2Fvar%2Frun%2Fdocker.sock/v1.35/version: dial unix /var/run/docker.sock: connect: permission denied\n',
'failed': True
}:
template error while templating string: unexpected '.'.
String: Docker version check (['/usr/bin/docker', 'version', '--format', "'{{.Server.Version}}'"]) failed: Got permission denied while trying to connect to the Docker daemon socket at unix:///var/run/docker.sock: Get http://%2Fvar%2Frun%2Fdocker.sock/v1.35/version: dial unix /var/run/docker.sock: connect: permission denied
(cherry picked from commit 4378542ac7)
* Add a changelog for the no-template error message fix
* win: handle non string as an environment value (#37215)
* win: handle non string as an environment value
* Changed powershell environment handler to use .net function instead for special chars
(cherry picked from commit 708869edd6)
* Added changelog fragement for powershell environment handler fix
* typo in changelog
* win_uri: return response info on non 200 responses, convert DateTime values to ISO 8601 (#37104)
(cherry picked from commit c1f5e11cdf)
* Added changelog fragment for win_uri return value fix
* Win lineinfile fix (#35100)
* win_lineinfile: fix#33858. Removed conversion from \r\n
* win_lineinfile: added test for #33858
* win_lineinfile: added documentation and more tests for change
* win_lineinfile: fixed wrong hash in testing
(cherry picked from commit e15a903bdf)
* Added changelog for win_lineinfile fix
* fix typo in changelog
* Remove default admin_distance and fix the idempotence thereof
Fixes#33290
* Fix tests and use yaml anchors to shorten tests
* Add test for undefined admin_distance
* Read config from `show run` if `show ip static route` fails
* Restore flags to ios.get_config & use get_config where appropriate
(cherry picked from commit 7016b3b9ca)
False assumption that values can not have cyclic dependencies. Fix by
removing dependency on self and look for cycles, if found remove
dependency to get a partial sort done.
(cherry picked from commit 042c111563)
Fix issues in ValueBuilder used in nso_config and nso_verify so that it
can handle leaf-list in NSO 4.5 and detect identityref types from
unions.
Fail gracefully if a type is not found.
(cherry picked from commit 2789cc5c09)
* allows ib_spec attrs to be filtered in update
This change will allow the ib_spec entries to be be filtered on a change
object by setting the update keyword to false. The default value for
update is true. When the update keyword is set to false, the keyed
entry will be removed from the update object before it is sent to the
api endpoint.
fixes#36563
* fix up pep8 issues
(cherry picked from commit 93b795baf0)
* Decouple config and state check in {network_os }_vlan and { network_os }_vrf modules
Fixes#35567Fixes#34754
`interfaces` option is used for configuration as well as operational state
check. If interface is configured to given vlan or vrf but if
operational state of interface is disabled it results in module failure.
Fix is to decouple same option usage for config and state.
With this fix `interfaces` is used as config option and a new
option named `associated_interfaces` will be used for intent check
for assigned interfaces.
* Fix CI failures
* Fix review comment
* Fixed integration test failure
(cherry picked from commit 5a6b893240)
* win_feature: better error handling to make it easier to debug issues (#36491)
* win_feature: better error handling to make it easier to debug issues
* removed ignroed pslint rules that are no longer needed
(cherry picked from commit ef4f8851dc)
* Added changelog for win_feature error handling fix
* fetch: set fail_on_missing: True as default as per docs
* Updated docs for fetch to say behaviour was changed in 2.5 and updated tests
(cherry picked from commit df8a5d7a4f)
* Bug in del(list) logic. Say you have a list of 4 elements a[0-3]
and you have to remove index 1 and 3, if you remove index 1 first
then list is cut short a[0-2] and does not have index 3
Fix: Remove indexes in reverse sorted order e.g. 3 and 1 in above
example so that order of indexes remain preserved even after deleting
fix is to remove indexes in reverse sorted order
* Add test cases for failed case
(cherry picked from commit 0bbea9a579)
* Fix nxos_system tests (#36201)
* Fix nxos_system tests
* Add debug connection plugin info
* Move sanity test under common
(cherry picked from commit 5b5d24631a)
* Fixes for the N3500 platform that uses the A8 image (#36261)
* fix nxos_l3_interface tests as n35 doesn't support ipv6
* add terminal dont-ask to nxos_feature and nxos_lldp
* put interfaces in L2 mode for N35
* fix nxos_feature unit-tests
(cherry picked from commit e24c547a3a)
This PR includes:
- Rename 'protocol' parameter to 'url_protocol'
- Add limited integration tests (as we have no public firmware access)
- Add missing examples
(cherry picked from commit e1b5bc049c)
When ACI modules are being used for querying MOs, we should not return
the previous state, as there is no previous state, there's only the
current state.
This impacts a lot of tests that were used to testing the current state
as 'previous'.
(cherry picked from commit dba561efa7)
This fixes, cloning operation where template or existing VM
does not have network or DVPG. Also, adds some strict type checking in
network parameters.
Signed-off-by: Abhijeet Kasurde <akasurde@redhat.com>
(cherry picked from commit a377302d6b)
This PR includes:
- A fix to the module where domain_type without domain caused havoc
- A fix to the integration tests when querying all binding objects
- Improvements to integration tests
(cherry picked from commit 7897558b95)
This PR includes:
- Fixes related to the recent merge of #31637 and #34537
- A generic fix for a reference for assignment issue
- Fixes to aci.boolean() in order to catch exception
(cherry picked from commit 01ba3a4efc)
The modules in this patch include waits that need to happen to ensure
something is correctly configured on a BIG-IP. These waits were
raised as an issue in a recent ansible-testing meeting.
This patch eliminates the waits by mocking time.sleep
(cherry picked from commit 77fa41795e)
* Normalize usage of temp and tmp on tmp
* Rename system_tmps system_tmpdirs
* Add ANSIBLE_REMOTE_TMP spelling of environment variables
(cherry picked from commit 06f73ad578)
Start using this to construct shade OpenStack Cloud objects in a
consistent manner. This will let us centralize things like dealing with
password arguments and whatnot. It also allows us to introduce the
ability to pass a fully formed config dict directly to the module.
Migrate all OpenStack modules to use openstack_cloud_from_module.
Have it return the shade library since it's responsible for
importing shade and shade is needed for the exceptions.
Only pull specific OpenStack arguments for the constructor
Rather than passing **module.params to the shade constructor, pull out
only the values that make sense. This should prevent the issues with
module parameters stepping on shade parameters.
Replace module.params.pop with module.params.get
We don't need to pop these anymore since the shade constructor is now
using opt-in values.
Using real urls is ungood. Use example.com domains. Also, get rid of the
antiquated port numbers.
(cherry picked from commit 0f893027c4)
Toshio Kuratomi