Skip to content

Latest commit

 

History

History
702 lines (543 loc) · 30.1 KB

ReleaseNotes.md

File metadata and controls

702 lines (543 loc) · 30.1 KB

The ReleaseNotes file

3.5-Beta

Contains a list of the released versions with a summary of the main changes in each version.

v3.5-Beta

  • Nov 2018

MAJOR RELEASE

  • The major changes were:

    1. LCM (LocalCatalogManager) v.0.2.4-Beta - Manages, Interprets, and Distributes MetaData - was add as automated deployment of Kave on AWS (see releases at https://github.com/KaveIO/LocalCatalogManager/releases )
    2. New version of ESKAPADE v0.7 - see releases at https://github.com/KaveIO/Eskapade/releases
    3. New version of MongoDB v.3.6
    4. Removed Kaveganglia from AmbariKave - deprecated, replaced by Ambari Metrics
    5. Removed JBOSS from AmbariKave - deprecated, replaced by WildFly
    6. Cloudbreak 2.7 support
    7. HDP stack version switched from 2.6.3.4 to 2.6.3.5, So the stack now looks like 2.6.3.5.KAVE (HDP-version.KAVE-version.KAVE)
  • Minor changes were:

  • Cloudbreak deployment improvements and bug fixes:

      - improved error handling
      - a lot of previously hard-coded data is now configurable - CB credential, volume size/count, node count, security group
      - added validations for hostgroups/ blueprints
      - separate mandatory and custom recipes
      - Cloudbreak credentials protection
      - Mechanism to stop/delete clusters via Cloudbreak
      - Fix broken formatting in documentation
      - Fix log message in Cloudbreak deployment script
    
  • Improvements in KaveToolbox service:

      - Ability to define custom env. script for all users
      - Default settings for Spark env. compatibility with Anaconda python (3.6)
      - KaveToolboxNode defined as slave instead of client service
    
  • Self-signed SSL certificates for Jenkins and Gitlab created at install time.

v3.4-Beta

  • November 2017

MAJOR RELEASE

  • The major changes were:

    1. Cloudbreak based automated deployement of Kave on Azure: 3.4-Beta
    2. New version of ESKAPADE 0.6 - A light-weight, Python-based data-analysis framework, meant for all sorts of data-analysis problems (see releases at https://github.com/KaveIO/Eskapade/releases, or http://eskapade.kave.io/)
    3. New version of KaveToolbox 3.6-Beta, see details here https://github.com/KaveIO/KaveToolbox/releases
  • Minor changes were:

    • Kave on Azure migration leveraging Cloudbreak
    • For cloudbreak freeipa was removed,
    • Do not stop httpd service when stopping Airflow service
    • Added guide how to enable SSL on Oozie
    • Use recommended ciphers and MACs for SSH
    • Do not expose Jetty version in HTTP requests
    • Disable version disclosure for Apache
    • Add security configurations for Apache, installed by FreeIPA

v3.3-Beta

  • August 2017

MAJOR RELEASE

  • The major changes were:

    1. New Service added - ESKAPADE 0.6 - A light-weight, Python-based data-analysis framework, meant for all sorts of data-analysis problems (see http://eskapade.kave.io/)
    2. HDP stack version switched from 2.6.0.3 to 2.6.1.0
    3. Moving to KaveToolbox-3.4.-Beta which sets up the kaveEvn for all non-system users.
  • The minor changes were:

    1. Some small fixes, enhancements in different services
      a. Airflow starts/stops unexpectedly (Removed airflow-scheduler status check from status function) b. ssh client time-out increased to 600s in clusters.py

KAVE stack versioning change

  • Users were confused by the version numbering of KAVE
  • So the stack now looks like 2.6.3.4.KAVE (HDP-version.KAVE-version.KAVE)
  • KAVE is a as usual a small extension of HDP and this should improve the clarity
  • Read more here about KAVE versioning: https://github.com/KaveIO/AmbariKave/wiki/kave-versioning

New Services:

  • ESKAPADE: A light-weight, python-based data analysis framework, meant for all sorts of data analysis problems.

Minor improvements in existing services:

  • A PEN Test was conducted

Other topics:

  • Eskapade service and integration tests
  • Modified role_command_order and created eskapade blueprints
  • Switched to Spark2 in ESKAPADE blueprint
  • Fixed continuing loop on ESKAPADE service stop
  • Changed Spark to Spark2 in examplehadoop blueprint
  • As a cluster administrator I want my installation to be reboot-proof: Tested
  • Incorrect restart hook in GitLab
  • As a cluster admin I want to be sure that Spark on YARN and Kafka are fully functional: Tested

Bugfixes in installer:

  • Added known forwarders to /etc/resolv.conf
  • Changed placeholder name and edit all places where /etc/resolv.conf templete is defined
  • Changed ktb env. path
  • Removed "Beta" from eskapade version
  • Changed KTB version to 3.4-Beta in order to have Python 3 installed
  • Eskapade install prodecure fixes
  • Remove Gate component from Eskapade service
  • Configure metainfo.xml for Eskapade service
  • Do not raise exception for Eskapade service status
  • Move Eskapade and KTB package name into separate variable
  • Package name depends on release version

v3.2-Beta

MAJOR RELEASE

Minor changes:

  • The minor changes were
    1. Some small fixes, enhancements in different services (Airflow, package dependencies, KTBox head test, Increased ssh client timeout to 60s in kavedeploy)
      a. kaveToolBox node removed from nameNodes
      b. Fixed CustomInstallscript.py to be created before install
      c. Disabled KTB version of spark
      d. common services removed form package builder
      e. Add .gitlab-ci.yml for gitlab-ci, put on hold for now. Will be part of one of the upcoming releases.
      f. Fixed Issues: ABK-508,
      g. Password protected notebooks
      h. Show restart warnings for services
      i. KTB on Python 3 breaks video /unity desktop
      j. Path odities for system users
      k. VNC
      l. Working Kave, KaveToolbox

v3.1-Beta

MAJOR RELEASE

MAJOR CHANGES: The major changes were:

	1.	Removal of CentOS6 and Ubuntu support. However we do support certain versions of Ubuntu for KaveToolBox.
	2.	Addition of Dovecot
	3.	Gitlab tunneling from the gateway node
	4.	PDSH issue resolution in integration test
	5.	Addition of Fish shell
	6.	MongoDB instance type resolution
	7.	MongoDB db_path resolution
	8.	HDFS-Log4j size modification
	9.	Port configuration for MongoDB
	10.	KaveLanding:
			a. Drop down menu fix.
	11.	Port Configuration for Nagios
	12. Lightning-Viz to KaveToolBox: Addition of new tool for visualization
	    a framework for interactive data visualization, including a server, visualizations, and client libraries
	13. KaveToolBox: small fixtures
	        a. CairoSVG version
	        b. Avoid loading default kaveenv for few users
	        c. Improved linux version detection code
	        d. Addition of remote desktop (via xrdp) support for the gateway node
	14. FreeIPA: Custom port configuration no longer supported. FREEIPA needs to be installed on a separate node as per the official documentation.
	15. Airflow: platform to programmatically author, schedule and monitor workflows
	16. Continuous Integration:
			a. Jenkins is running now nightly builds for KaveToolBox
			b. For Ambari Kave: The following tests are running with success
			    i. Unit Tests
			    ii. Service Tests
			    iii. Integration Tests
	This was a long overdue, that we have managed in this release.
	18. KaveGanglia integration with kave: tasks finalized.

Important information:

v3.0-Beta

MAJOR RELEASE not guaranteed backwards compatible

MAJOR CHANGES

Important information:

  • KaveToolbox 3.0: ROOT 6.04, Pentaho Kettle not installed by default, anaconda 4.1

  • HDP 2.5 · New Services:

       o   SmartSense: SmartSense, which is a collection of tools and services that help Hortonworks
       Data Platform’s operators quickly resolve issues, and also act on proactive recommendations
       that help avoid future issues. This service has been added
       as part of HDP 2.5 migration.
    
       o   HDFS Log viewer: Log collection for HDFS client and server. This service has been added
       as part of HDP 2.5 migration.
    
       o   Zeppelin: Apache Zeppelin interpreter concept allows any language/data-processing-backend
       to be plugged into Zeppelin. Currently Apache Zeppelin supports many interpreters such as
       Apache Spark, Python, JDBC, Markdown and Shell. This service has been added as part
       of HDP 2.5 migration.
    
        §  Prerequisites:
    
          ·         HDP 2.4.2 is installed
    
          ·         The cluster contains Spark 1.6 or 1.5
    
          ·         Git is installed on the node running Ambari Server
    
          ·         You can install git as ‘ sudo yum install git’
    
          ·         Java 8 is installed on the node where Zeppelin is installed
    
    
    
      Minor improvements in existing services:
    
       o   Ambari Metrics
    
       §  role-command order fixes for more stable installation
    
    
    
          ·         KAVE stack versioning change
    
          ·         In 3.0 we are now including the KAVE version as well as the HDP version
                    in the stack name
    
          ·         So the stack now looks like 2.5.3.1.KAVE (HDP-version.KAVE-version.KAVE)
    
          ·         KAVE is a as usual a small extension of HDP and this should improve the clarity
    
          ·         Read more here about KAVE versioning:
                    https://github.com/KaveIO/AmbariKave/wiki/kave-versioning
    
  • Storm 1.0.2 Migration from storm 0.10 to latest storm version 1.0.2

       -Latest configuration file in the form of yaml file does not have certain configurations which were
        present in the older version of storm such as
        nimbus.impersonation.authorizer: "backtype.storm.security.auth.authorizer.ImpersonationAuthorizer"
        has been removed
    
    
       -For some of the configuration values which earlier began with "backtype.storm" they have now been
        replaced with "org.apache"
    
       -For example storm.thrift.transport: "backtype.storm.security.auth.SimpleTransportPlugin" is now
       storm.thrift.transport: "org.apache.storm.security.auth.SimpleTransportPlugin"
    
       -As a part of the configuration we have kept bare minimum
       configurations which are absolutely required to run a storm cluster and those configurations which we
       have overridden in our code.
    
       -Pacemaker as a separate service has been incorporated in storm 1.0.2, however the implementation
       would be part of the next release of KAVE.
    
  • Centos6 deprecated - We no longer run tests on Centos6. It does not mean that Centos6 will not work, but it's not tested or supported by us any longer.

  • HUE deprecated - HUE used to be the one place to go to get all databases interfaces within KAVE. However, it does not work in Centos7 and so, has been deprecated.- We notice that most of our users are switching to TEZ&Spark as default processors- Hortonworks HUE does not support Tez (now the default) or Centos7/Redhat7- So not only is the use case reduced but the compatibility is removed- We advise users familiar with Hue to switch to ipython notebooks and pyspark

  • Added validators for the params

  • Minor changes in the test scripts

  • Added the new example blueprints have new HDP 2.5 services

v2.2-Beta

This may be one of the last versions with centos6 support. More than 100 independent improvements made since 2.1

Important information:

New OS support

  • We now support the following OSes for AmbariKave
  • Centos6 (well tested and stable, to be removed in a future release)
  • Centos7 (for all services except HUE, well tested and stable)
  • Redhat7 (for all services except HUE, not yet fully tested)
  • Since we now support multiple OSes, we have removed the centos6-flag from the installer
  • Note that KaveToolbox has been similarly modified to support C6/C7/U14/U16/RH7

Deprecated HUE on centos7/redhat7

  • In the latest HDP, the Ambari web itself contains interfaces to HIVE
  • We notice that most of our users are switching to TEZ&Spark as default processors
  • Hortonworks HUE does not support Tez (now the default) or Centos7/Redhat7
  • So not only is the use case reduced but the compatibility is removed
  • We advise users familiar with Hue to switch to ipython notebooks and pyspark

FreeIPA independence

  • In previous versions of AmbariKave FreeIPA was forced to be installed on the ambari node
  • In 2.2-Beta FreeIPA can now be installed on a different node within the cluster
  • Installing on a separate node is now the preferred method due to separation of concerns
  • FreeIPA is the central security system for KAVE, so much better if it is isolated
  • The FreeIPA node root must also be able to ssh into all other machines as with ambari

JBOSS -> WildFly

  • JBOSS 7.1 is deprecated in favor of the sucessor project, WildFly 10.1.0
  • JBOSS does not work on Centos7, but is kept for now for the sake of backwards-compatibility
  • JBOSS will be removed in a future release

KAVE stack versioning change

  • Users were confused by the version numbering of KAVE
  • In 2.2-Beta we are now including the KAVE version as well as the HDP version in the stack name
  • So the stack now looks like 2.4.2.2.KAVE (HDP-version.KAVE-version.KAVE)
  • KAVE is a as usual a small extension of HDP and this should improve the clarity
  • Read more here about KAVE versioning: https://github.com/KaveIO/AmbariKave/wiki/kave-versioning

New Services:

  • Nagios: A common monitoring and alerting framework with a large userbase. In HDP 2.4 Nagios was deprecated, however no sufficiently stable alternative was added. In this release we have restored a KAVENAGIOS service with basic Nagios functionality for users to use in the intervening time before migration to Ambari Metrics

  • Ganglia: A common monitoring and alerting framework with a large userbase. In HDP 2.4 Ganglia was deprecated, however no sufficiently stable alternative was added. In this release we have restored a KAVEGANGLIA service with basic Ganglia functionality for users to use in the intervening time before migration to Ambari Metrics

  • WildFly: Successor of JBOSS, java application server.

Minor improvements in existing services:

  • FreeIPA

    • port numbers: certain ports are now configurable via a script Changing ports in FreeIPA is not supported by the IPA developers, but is possible with complex editing of the FreeIPA python. We have tested and wrapped this editing to enable the user more control over the ports of FreeIPA however this is not necessary unless FreeIPA is installed next to different services which is no longer required or recommended, best to isolate FreeIPA
    • Minor fixes in user generation order for code quality
    • the freeipa service administrator (admin) is now by default not a command-line (shell) user
    • role-command order fixes for more stable installation
    • --force-join in client install to fix partial installation failures
  • MongoDB

    • Add MONGODB_ARBITER component an Arbiter is a mongodb master component which does not store data but does contribute to voting on the primary. Useful to break ties
  • Archiva

    • Small modifications for RH/Centos7
    • Allow configuration of Archiva admin user via Ambari
  • JBOSS

    • Fix java version incompatibility
    • Deprecate in favor of WildFly

Httpoxy:

  • The simple fix for the recently announced CVE known as httpoxy is now applied by default to all apache-based installations. For previous releases the admin can patch manually with a small configuration change in Apache services

Other topics:

  • resource_wizard.py is now independent from pandas and updates with latest specs
  • restart_all_services.sh improved to perform better on larger clusters
  • clean.sh overhauled to solve database dropping issues on centos7 and to enable cleaning within a cluster using pdsh
  • test scripts and deployment libraries have been slightly improved we now also auto-deploy in frankfurt and mumbai regions for testing
  • additional unit tests for pep8 compliance
  • We have begun docker experimentations in both AmbariKave and Centos7

v2.1-Beta

  • June 2016

Minor release with significant bugfixes, new services and explorations of centos7

New services:

  • We have added a new implementation of Ganglia to our stack, separate from the old HDP implementation. This is optional and can be used in the interim period while Ambari Metrics grows in maturity.

Minor improvements:

  • Update to Ambari 2.2.2, bringing Grafana and other minor fixes
  • Further code style fixes and additional unit tests
  • Improvements in deployment libraries for stability

Minor improvements in services:

  • Move forward to Java 1.8 consistently across all services
  • HUE allow configuration of service username
  • Move up to KaveToolbox v 2.1-Beta (described elsewhere)

Bugfixes in services:

  • epel repo seems not very stable recently, yum clean all seems to solve most issues
  • FreeIPA:
    • says install was successful in the client even when it failed actually
    • FreeIPA now supports domain names up to 61 characters (experimental patch)
    • resolv.conf overwriting now only occurs on successful client install
  • KaveLanding:
    • KaveLanding now supports modified ambari password correctly
  • Archiva:
    • Re-installation is now possible with limited cleanup, small changes in properties

Bugfixes in installer:

  • No longer breaks with local directory in /etc/kave/mirror file
  • New methods to check pdsh version compatibility
  • Checkout/wget into temporary directory for cleanliness

Centos7:

  • We have begun to test KAVE on the Centos7 operating system. In version 2.1 Centos6 is still the only supported os, but several of the basic services already deploy on Centos7. In the near future we intend to supply a version which also supports Centos7.
  • We have already fixed several issues in the installer, Apache, deployment library, Sonar, FreeIPA
  • At the moment there are major issues in FreeIPA, SonarQube and HUE specific to Centos7.

v2.0-Beta

  • May 2016
  • MAJOR release to migrate to the latest Ambari version 2.2.1
  • KAVE stack now on top of HDP 2.4 (2.4.KAVE)
  • HOTFIX: gitlabs CVE-2016-4340 necessitated a quick hotfix on this version, updating gitlabs from 8.6.3 -> 8.7.1 (3rd May)

Major modifications in services:

  • MongoDB, version increase to MongoDB 3. MongoDB 3 is not 100% backwards compatible
  • FreeIPA Major modifications to how we handle keytabs for kerberos. During the kerberization of your cluster with Ambari, you can download a kerberos.csv containing the list of all keytabs to be created With this list, on the ambari node, as the root user:
    • kinit admin
    • createketabs.py kerberos.csv Then continue with the wizard through the web interface See the wiki and readme for details In the new Ambari version we need to distribute the Java Cryptographic Extensions into the correct directories before enabling kerberos on the cluster. This is now take into account by the FreeIPA client install
  • KaveToolbox Version update to 2.0-Beta, many changes described separately in the KTB readme Most major change: creation of versioned directories
  • StormSD Added the Storm Logviewer service, an optional service for each node, providing a web interface for the storm logs

General improvements:

  • Example blueprints and diagrams modified to respect new services
  • Tests modified to reduce fragility
  • Modifications to installer to improve pdsh and Ambari install with new Ambari version
  • Workarounds only necessary for old Ambari versions reviewed and removed
  • Check pdsh version and group names during restart_all_services.sh script
  • addition of resource_wizard.py script, which naively guesses cluster size requirements

Minor modifications in services:

  • MongoDB Conf file and repo file now configurable with an InlineTeplate through the web interface
  • JBOSS In previous versions JBOSS would overwrite the standalone.xml with defaults on restart Now the standalone.xml is an InlineTemplate and configurable
  • StormSD Updated version of Supervisord to v3.X. This new version of supervisord respects import statements which makes it a lot easier to configure and does not invoke duplication of code
  • SonarQube Updated to version 5.4 of the server, only a minor update
  • Jenkins: updated to version 1.642
  • Gitlabs: updated to version 8.7.1

Bugfixes:

  • StormSD re-installation of partially completed install fixed. installation errors now with better error messages installation error when on top of FreeIPA fixed
  • JBOSS fix removing of users during install fix re-installation of partially completed install
  • FreeIPA: Installation without DNS fixed

Notes about the new ambari/HDP version:

  • Ganglia and Nagios are replaced by AmbariMetrics The new HDP assumes the user will install Ganglia/Nagios themselves if needed
  • The latest stack requires more RAM on the ambari (+500MB) node and more disk space to run HDP components (+2GB)
  • Ranger, Atlas and Falcon, although they have been added are not installable in a simple way the installer must install and create specific databases before following the install wizard

v1.4-Beta

  • Jan 2016
  • Bugfix release in preparation to migration to Ambari 2.X
  • Over 50 independent fixes

General improvements:

  • Use Ambari's protected strings feature to hide sensitive strings such as passwords from logs
  • Modify test/deployment framework to avoid ever using the ambari password as plain text on the command line
  • Review initial passwords across services
  • Finalize integration of remaining services with FreeIPA
  • Copy a lot of templates into the configuration files, so that the templates are also configurable

New features/improvements in services:

  • StormSD: Moved to Storm 10.0
  • FreeIPA: kadm.acl is now configurable through ambari to control the kerberos admin users resolv.conf now configurable
  • Gitlab: set initial gitlab root admin password based on configurable
  • Sonar: Integrate with Host-based access control to centralize user management with FreeIPA
  • MongoDB: Replication of datasets is now supported
  • Twiki: Integrate with Host-based access control and also LDAP to centralize user management with FreeIPA
  • Jenkins: set initial jenkins root admin password based on configurable

Bugfixes:

  • FreeIPA: initial_passwords does not need to be of password type. prepare for moving to version 2.X by coping with reading all hosts from all_hosts
  • Apache: start/stop of Apache services now more reliable
  • KaveLanding: no longer show false-negative or false-positive running status
  • Twiki: Fix installation error where PhP is missing
  • Packaging script: Fix packaging/installation script so that it copes with missing repositories in the mirrors file

Modifications to tests and deployment:

  • Improved documentation
  • Check for PDSH at start of scripts
  • Respect git branches
  • Modify example blueprints for latest changes
  • New unit tests verify python completeness of files
  • More prior checking of blueprints within tests
  • Fix creating Centos7 clusters to avoid needing a tty terminal

v1.3-Beta

  • Sept 2015
  • Bugfix release in preparation for migration to Ambari 2.X
  • Over 50 independent fixes

New features in services:

  • KaveToolbox: move to 1.3 - Beta(lots of minor improvements), e.g. the ProtectNotebooks.sh script
  • KaveToolbox: Possibility to customize CustomInstall.py added as configuration parameter
  • Hue: now integrated with HBAC(host - based access control) and thus also with FreeIPA
  • Jenkins: Migration to latest stable version
  • Jenkins: Installer and list of default plugins now cached on our repo server
  • FreeIPA: you can now specify also a list of sudoers in the initial configuration
  • FreeIPA: Full usernames(first and last) can now be added through the configuration
  • KaveLanding: Links are now sorted
  • KaveLanding: Custom links can be provided
  • TWiki: HBAC and LDAP access supported, thus integrated with FreeIPA
  • All services reviewed to ensure they can be restarted effectively without downloads after installation
  • Ambari / clusterHostInfo / can be used to determine most server locations in the cluster

New features in test and deployment framework:

  • restart_all_services.sh script: recovers from most aborted blueprints see the in-built help
  • Add check for datetime(ntp) in deploy_from_blueprint.py to avoid unknown timing issues for ambari agents
  • Default blueprints updated, several spurious parameters no longer need to be specified in current ambari versions
  • Slight modification to host authentication in the gitwrap script used for testing
  • FDisk part of mount command is now guessed per region, simplifying the test scripts
  • In tests, threads now unlock if an exception is thrown
  • Small bug in packaging script fixed

Bugfixes:

  • TWiki: LocalSiteCfg was broken through misapplication of a template
  • TWiki: different default parameters being required through webUI vs blueprints fixed with defaults
  • StormSD: Parameters renamed so that the install also works through the web interface
  • StormSD: Partial fix for multiply writing the same configuration files into supervisord.conf, now less common
  • StormSD: Start / Stop / Restart methods heavily modified to spawn background processes, otherwise they hung forever
  • KaveLanding: bootstrap / bower not actually needed in most cases
  • KaveLanding: removal of duplicate code
  • KaveLanding: False positive status fixed(see status methods below)
  • KaveToolbox: --ignore - missing - groups accidentally always applied, new logic added
  • KaveToolbox: missing git install
  • FreeIPA: did not declare it's forwarders configuration parameter
  • Gitlab: Unicorn port(8080) was not configurable leading to port conflicts, fixed
  • SonarQube: Unused parameters removed
  • MongoDB: in the case a user - specified and IP address for a mongo server, it was not actually being applied
  • Start and Install methods need to both call configure in most cases so that web interface reconfiguring is possible
  • Status method needs to throw an exception when service not running
  • Start / Stop / Restart methods reviewed especially for backgrounded - services

v1.2-Beta

  • June 2015
  • Bugfix release of AmbariKave with over 40 independent fixes
  • Thanks to the beta testing program for providing so much feedback

New features in services:

  • storm added to the path for all users when storm is installed
  • StormSD various parameters such as ports, zookeeper - servers, drpc servers, now allow for comma - separated - lists
  • FreeIPA service now has a parameter to set groups, users, and initial user passwords
  • FreeIPA now sets default user shell(bash as default, but it is configurable)
  • new MONGODB_CLIENT adds mongo libraries for gateway machines
  • MONGODB_CLIENT implements mongok script for simpler connectivity(mongok - -help)
  • KaveLanding, small changes in website layout and better parsing of list of services

New features in test and deployment framework:

  • current branch is now the default in tests
  • deploy_from_blueprint.py now also creates pdsh groups for simpler local cluster management
  • new restart_all_services.sh script to recover from aborted blueprint deploys
  • New more fully tested blueprints which match better the examples on the twiki
  • Cleaning script now also cleans content of the ambari - agent directory

Bugfixes:

  • HUE integration with beeswax / hive minor fix in port number resolution
  • add_ebs_volume script major fixes
  • FreeIPA previously made a bind user with no password
  • FreeIPA and SonarQube both had missing defaults in their configuration xml
  • FreeIPA throwing a certain exception caused a different exception to be raised, confusingly
  • FreeIPA now can be recovered from partial failures with a re - install
  • tWiki installation was failing on jinja template resolution

v1.1-Beta

  • May 2015
  • First version of the AmbariKave installer for public beta release(open source release)
  • Stack 2.2.KAVE derived from HDP 2.2

Implemented 14 new services:

  • FreeIPA as user - management component
  • MongoDB and StormSD as key components in the complete lambda architecture
  • Gitlabs, Jenkins, Twiki, SonarQube, Archiva, for the development line
  • KaveToolbox installation in one of two modes for the analytical tools
  • HUE on top of HDP for simplified hive access
  • Apache and JBOSS webservers
  • KaveLanding, a simple website of internal links within a KAVE
  • Mail, postfix and dovecot for mail servers

Integrated services together:

  • Integration of HUE into HDP
  • Integration of FreeIPA with Twiki, Jenkins, Gitlabs, HDP(Kerberos, LDAP, DNS and unix users)

Implemented common libraries:

  • src / shared / kavecommon is automatically distributed during patching

Automated deployment and testing:

  • As part of our development cycle automated deployment and testing of deployed instances is done through aws or vagrant
  • We extended the python unittests package for our testing purposes
  • Implemented several developer scripts to speed up the development process
  • We wrapped aws cli in a python suite to simplify these tests, but that's not really needed for most purposes
  • We wrapped blueprint deployment into a simple script which is quite useful for easily deploying / redeploying clusters