Release notes
0.6.1
- In addition to login/password, Ranger authentication can now be performed through Kerberos.
0.6.0
- /var/run/supervisor_xxxx folder was sometime deleted (i.e. after a hard restart). Now automatically recreated
- Yarn service failed when existing job list was empty. Fixed
- Improving documentation on Yarn services integration with Kerberos.
- Added handling of secured access to elasticsearch cluster (Authentication and SSL).
INCOMPATIBILITY
- kafka_relay has a new mandatory parameter:
kafka_version
. - Due to change in the Kerberos handling of the Kafka plugin,
principal
andxxx_keytab_path
parameters has been removed from the kafka_relay definition.
0.5.7
- Fixed this [DEPRECATION WARNING]: Using tests as filters is deprecated. Instead of using
result|succeeded
instead useresult is succeeded
--action dumpvars
now display the file where the variable is defined.
0.5.6
- Added a Yarn plugin to handle Yarn services (long running jobs) lifecycle.
- Storm topologies ACTIVE status is now displayed in upper case, for better visibility.
- HDFS plugin:
hadoop_conf_dir
andwebhdfs_endpoint
was not propagated on all HDFS commands. Fixed. - Users plugin: When a user is not managed and an authorized_keys is to be set, HADeploy ensure home folder is accessible by performing an
su
command.
0.5.5
- Added Elasticsearch indices and Templates management.
- Scope for Storm's notification handler was erroneous. Fixed
- Now allow missing both
ssh_private_key_file
andssh_password
in host definition. (As already stated in the doc) - Notification system generated some errors when scope was limited to
files
. Fixed by suppressing notification in such case. - Modification of grooming order for
Storm
,Systemd
andSupervisor
plugins. Now grooming occurs beforefiles
, to handle notifications correctly. - ansible_inventories is now compatible with
ansible_user/host/...
variables (Formerlyansible_ssh_user/host/...
). - Refactoring of supervisors plugin to have one configuration file per program and group (Instead of a single file for all).
- Added a
scope
attribute to supervisor_programs items. - Action
remove
on Storm plugin now kill all running topologies.
0.5.4
- Added an action
dumpvars
, to dump all variables. - With Ranger 0.7, there was unjustified 'changed' on policies settings. Fixed
- Added a Storm plugin to handle Storm topologies lifecycle.
- Added Storm Ranger policies management.
- Action
status
as been implemented in Storm, Systemd and Supervisor plugins.
INCOMPATIBILITY
- Notification syntax has been modified for
files
definition.
0.5.3
- In some cases, using include directive disrupted some relative file relocation. Fixed
- An alternate variable notation (
<<...>>
) has been introduced to allow non-string variable in flow style notation. - Conditional deployment was implemented in all plugins.
- A switch
no_log
has been added toranger_relay
to ease debugging.
0.5.2
- A change in the API of Ansible 2.4 broke
ansible_inventory
plugin. Fixed.
0.5.1
- Added almost all existing Ansible configuration variables for
hosts
andhost_override
in inventory. - Added maven artifact download, using 'maven_repositories` definition.
- Changed all references to ansible variable
ansible_ssh_user
toansible_user
(Following Ansible evolution). - Some (small) documentation improvements.
- Added
priority
attribute onhost_overrides
- Added
systemd
plugin withsystemd_units
service management. - Added
supervisor
plugin withsupervisor
process controler management.
0.5.0
- Plugins architecture refactoring
- New plugin:
ansible
, to insert raw Ansible playbook or role in the deployment. - scope value are now checked.
- Added an
encrypted_vars
block to have a more generic encryption capability.
INCOMPATIBILITY
- The
ranger_relay.admin_password
encrypted with 0.4.0 method must be modified to comply to new, generic syntax.
0.4.1
- Added scope limitation mechanism on performed operation
- Default HDFS relay cache is now /tmp/hadeploy_{{ansible_ssh_user}}/files
- Added
remote_tmp = /tmp/.ansible-${USER}/tmp
in generated ansible.cfg
0.4.0
- Some quotes was missing in plugin yaml files. This generated errors when a HADeploy variable was substituted with a string begining with {{some_ansible_variable}}. Fixed
hdfs_relay.cache_folder
default value is modified, to be in the user home folder.hbase_relay.tools_folder
default value is modified to be in/tmp/hadeploy_{{ansible_ssh_user}}
hive_relay.tools_folder
default value is modified to be in/tmp/hadeploy_{{ansible_ssh_user}}
kafka_relay.tools_folder
default value is modified to be in/tmp/hadeploy_{{ansible_ssh_user}}
- Local file definition now works with ~/xxx
hdfs_relay.user
default value is nowhdfs
only ifssh_user
is root. Otherwise, isssh_user
- Added global
exit_on_fail
flag. - Added some retry on user/groups creation/removal
- Added
become_user/become_method
onhbase_relay
- Added
become_user/become_method
onkafka_relay
- Added
hbase_relay.local_keytab_path
- Added
kafka_relay.local_keytab_path
- Added
hdfs_relay.local_keytab_path
- Added
source_host_credential.local_keytab_path
- Added
groups
inhost_groups
- Previous version was unable to fetch an existing Ansible inventory when it contains some encrypted file(s).
The Description of Ansible inventory has been modified (now
ansible_inventories
) to include a password file or user password request. - Added a method to encrypt password in
ranger_relay.admin_password
.
DEPRECATION
ansible_inventory_files
has been marked deprecated. Replaced by ansible_inventories
.
INCOMPATIBILITY
There is some incompatible change with previous version. You may need to modify your source files:
hive_relay.user
is renamedhive_relay.become_user
.hbase_relay.keytab_path
is renamedhbase_relay.relay_keytab_path
kafka_relay.keytab_path
is renamedkafka_relay.relay_keytab_path
hdfs_relay.keytab_path
is renamedhdfs_relay.relay_keytab_path
source_host_credential.keytab_path
is renamedsource_host_credential.relay_keytab_path
- For
files
andtrees
,<node>:///
is replaced bynode://<node>/...
notation