- Including PR #111 - "Handle steps status that hasn't run yet". Thank you @robert2d!
- Including PR #112 - "Parse Dates when step/cluster still starting". Thank you @robert2d!
- Including PR #113 - "Add job flow steps remove extra steps key". Thank you @robert2d!
- Including PR #114 - "Allow tags to be set on runJobFlow". Thank you @robert2d!
Amazon is in the process of transitioning from the notion of "Job Flows" to "Clusters" and is updating their APIs as such. You've already seen this in the EMR web UI as all mentions of "job flows" are gone and now you create "Clusters".
On the API side, all of the newer commands take cluster_id
rather than job_flow_id
. On the API submission side, they are transitioning from a 'flat' structure to a nested JSON structure requiring no transformation. Finally, XML is all gone and commands return JSON (i.e. no more Nokogiri as a dependency).
They've also begun deprecating APIs, starting with DescribeJobFlows
. Given the sweeping set of changes, a major release was deemed appropriate.
- #88 - Removed support for deprecated
DescribeJobFlows
. - #89 - Add support for
AddTags
. - #90 - Add support for
RemoveTags
. - #91 - Add support for
SetVisibleToAllUsers
. - #92 - Add support for
ListSteps
. - #93 - Add support for
ListInstances
. - #94 - Add support for
ListInstanceGroups
. - #95 - Add support for
ListClusters
. - #96 - Add support for
ListBootstrapActions
. - #97 - Add support for
DescribeCluster
. - #98 - Add support for
DescribeStep
. - #101 - Fix plurality of
TerminateJobFlows
; now requires an array of IDs to terminate. - #102 - Simplify interface to
AddJobFlowSteps
; no longer require extraneous:steps => []
. - #104 - Expose return value from
AddJobFlowSteps
. - #105 -
JobFlow#status
has been removed in favour ofJobFlow#cluster_status
andJobFlow#cluster_step_status
. - #107 - Add support for temporary credentials via
Elasticity.configure
. - #109 - Credential specification relocated to
Elasticity.configure
.
- Fix for issue #86.
- Fix for issue #83,
elasticity
has now transitioned to the AWS Signature Version 4 Signing Process. - Removed the ability to create insecure (HTTP) connections to the EMR endpoints.
- Bear with me here :) Backmerged into 4.0.4 to add IAM Service Role support per @alexanderdean. As part of the forward merge, bumping the version to trigger an update.
- Major version bump as there are language support implications.
- Now supporting only the latest version of the 2.x minors (e.g. 2.0.x, 2.1.x, etc.) versus specific minor-minor releases. This was an oversight on my part in how I both configured Travis and communicated support for the 2.x line.
- Removed support for Ruby 1.9.3 as it has been unsupported since 2/23/2015.
- Removed "support" for JRuby. It was in poor form that I originally claimed to support JRuby as I do not have the time to dedicate to such an endeavor. Compatibility is merely coincidental and it would not be responsible for me to continue communicating ongoing support when that is not my intention. Apologies to those this may inconvenience.
- Thanks to @alexanderdean, pull request #82 adds support for profile roles.
- Thanks to @jshafton, pull request #79 adds support for job flow roles.
- Thanks to @ilyakava, pull request #78 fixes a bug in the "wait for completion" feature.
- Thanks to @jshafton, now with support for Ganglia!
- Now tracking the Master Instance ID and Created At timestamp for job flows and steps, via @AuraBorea (Thanks!) #73.
- Fix for issue #69. The AWS region was previously being derived from the placement. With the advent of VPC/subnet IDs being set, placement is not always relevant (as the VPC subnet ID implicitly defines a placement). Since region cannot be derived it is now available directly on
JobFlow
. - Fix for issue #79. Removing the ability to set the now deprecated Hadoop version. It is now set via specifying the AMI version. See the EMR docs for more details.
- Building against JRuby 1.8. While I don't necessarily have the resources to optimize for JRuby compatibility (given that I know next to nothing about it :) I'll use this as a canary to tackle any low-hanging issues that may arise.
- Second go at a fix for Issue #60, "Specifying a VPC subnet does not work" as placement and subnet cannot both be specified.
- Fix for Issue #60, "Specifying a VPC subnet does not work".
- Add support for Ruby 2.1.0 and 2.1.1.
- Major/minor bump because of breaking API changes to
Elasticity::BootstrapAction
due to Issue #55. After spending some time deprecating, I realized I don't have the code bandwidth to do it in a way that I would be happy with. Move fast and break things ;) - Dev update: Added
Elasticity::EMR#describe_jobflow_xml
to assist in job flow debugging.
Elasticity::S3DistCp
andElasticity::ScriptStep
added to provide easy access to remote copying and arbitrary script execution.Elasticity::EMR
can now autodetect AWS credentials.- Baseline support now is now 1.9.3-p484. Dropped support for Ruby 1.9.2 as it was EOL on 2013/06/01.
- Dev update: No longer using
guard-rspec
. - Dev update: Added
unf
as an explicit dependency asfog
no longer includes it since it's AWS-only.
- Added debugging support via
JobFlow#enable_debugging
. Note that this requiresJobFlow#log_uri
to be set. - Added job flow completion polling via
JobFlow#wait_for_completion
. - Added testing to support Ruby 2.0.
- Removed support for REE and 1.8.7 as these are now unsupported versions of Ruby. Common strategy in the Ruby community has been to only perform a minor version bump in this case; Elasticity is following suit.
- Now specifying minor versions in the gemspec. With the release of Ruby 2.0, I'm anticipating breaking changes coming to many gems, and hoping that this mitigates those effects.
- Dev update: guard-rspec added.
- Dev update: All development dependencies now require Ruby >= 1.9.2.
- Dev update: Latest version of Ruby 1.9.2 (p320).
- Dev update: Migrated from .rvmrc => .ruby-version and .ruby-gemset.
- Pull request from Aaron Olson, removing requirement that a
JobFlow
has steps before running. - Updating development to ruby-1.9.3-p385.
- Pull request from Aaron Olson, adding
StreamingStep#arguments
.
- Pull request from Aaron Olson, adding
JobFlowStatus#normalized_instance_hours
.
- Added
#visible_to_all_users
toJobFlow
. Thanks to dstumm for the contribution! - Added
#ended_at
toJobFlowStatus
. - Added
#duration
calculated field toJobFlowStatus
.
- Configuration of Hive installations via
hive_site
is now supported. See the README.md for details.
- When
JobFlow#placement
is specified, instances are created in that availability zone. Previously, this setting was only used to derive the EMR API endpoint to connect to (i.e. the region). - Updated development dependencies.
SyncToS3
supports S3 region specification.SyncToS3#sync
supports being called with both files and directories.
SyncToS3
added to enable one-way asset synchronization.- Generic bootstrap actions are now supported via
BootstrapAction
. - If you have several Hadoop bootstrap actions (15 is the current EMR limit), store all of your Hadoop configuration options in a file, ship it up with
SyncToS3
and use the newHadoopFileBootstrapAction
to point at that file. - If no parameters are passed to
JobFlow.new
, it will use the standard AWS environment variables to lookup the access and secret keys -AWS_ACCESS_KEY_ID
andAWS_SECRET_ACCESS_KEY
. - New dependencies: fog (S3 access), fakefs (filesystem stubbing - development only), timecop (freezing and manipulating time - development only).
- Birthday release! ;)
- Bumped the default version of Hadoop to 1.0.3.
- Amazon now requires the
--hive-versions
argument when installing Hive (thanks to Johannes Wuerbach). JobFlowStatus#master_public_dns_name
is now available (thanks to Johannes Wuerbach).
JobFlow
now supports specifying availbility zone instance specification viaJobFlow#placement=
.JobFlow::from_jobflow_id
now supports region specification so that jobs created in regions other than us-east-1 can be recreated.
- Hadoop streaming jobs are now supported via
Elasticity::StreamingStep
.
JobFlow::from_jobflow_id
factory method added so that you can operate on running job flows (add steps, shutdown, status, etc.) that you didn't start in the same Ruby instance.- Updated to rspec-2.11.
- TASK instance group support added.
- SPOT instance support added for all instance group types.
- Removed name of jar from default name of
CustomJarStep
since the AWS UI already calls it out in a separate column when looking at job flow steps.
2.0 is a rewrite of the simplified API after a year's worth of daily use at Sharethrough. We're investing heavily in our data processing infrastucture and many Elasticity feature ideas have come from those efforts.
In order to move more quickly and support interesting features like a command-line interface, configuration-file-based launching, keep-alive clusters and more - a remodeling of the simplified API was done. This is going to result in breaking changes to the API, hence the bump to 2.0. I hope that most of you were using gem 'elasticity', '~> 1.5'
in your Gemfile :)
- The
SimpleJob
-based API has been removed in favour of a more modular step-based approach using the "job flow" and "step" vernacular, in line with Amazon's own language. If you're familiar with the AWS UI, using Elasticity will be a bit more straightforward. - The functionality provided by
JobFlow
andJobFlowStep
has been transitioned toJobFlowStatus
andJobFlowStatusStep
respectively, clearing the path for use ofJobFlow
andJobFlowStep
in job submission.
- When submitting jobs via
JobFlow
API, it is now possible to specify the version of the AMI, whether or not the cluster is keep-alive, and the subnet ID (for launching into a VPC). Keep in mind that AWS will error out if you specify an unsupported combination of AMI and Hadoop version. - The default version of Hadoop in
JobFlow
is now 0.20.205. The previous default was 0.20 in case you'd like to set it yourself. - It is now possible to name Hadoop bootstrap actions, making it easier to understand the actions when looking in the AWS UI after a job is submitted.
- AWS requests are now POSTs (thanks to Menno van der Sman) in order to avoid server-imposed GET request size limits. Rather than maintain two separate code paths for GET and POST, we decided to only support POST as there is no reason to support both.
- Drastic simplification of the testing around EMR submission, reducing LoC (however important that metric is you :) and complexity by ~50%.
- Development dependency updates: updated to ruby-1.9.3-p194 and rspec-2.10. Removed dependency on VCR and WebMock (no longer using either of these).
- Added support for Hadoop bootstrap actions to all job types (Pig, Hive and Custom Jar).
- Added support for REE 1.8.7-2011.12, Ruby 1.9.2 and 1.9.3.
- Updated to the latest versions of all development dependencies (notably VCR 2).
- Added
Elasticity::EMR#describe_jobflow("jobflow_id")
for describing a specific job. If you happen to run hundreds of EMR jobs, this makes retrieving jobflow status much faster than usingElasticity::EMR#describe_jobflowS
which pulls down and parses XML status for hundreds of jobs.
- Added
Elasticity::CustomJarJob
for launching "Custom Jar" jobs.
- Explicitly requiring 'time' (only a problem if you aren't running from within a Rails environment).
Elasticity::JobFlow
now exposeslast_state_change_reason
.
This release primarily contains contributions from Wouter Broekhof
- The default mode of communication is now via HTTPS.
Elasticity::AwsRequest
new option:secure => true|false
(whether to use HTTPS).Elasticity::AwsRequest
new option:region => eu-west-1|...
(which region to run the EMR job).Elasticity::EMR#describe_jobflows
now accepts additional params for filtering the jobflow query (see docs).
HiveJob
andPigJob
now support configuring Hadoop options via#add_hadoop_bootstrap_action()
.
- Shipping up E_PARALLELS Pig variable with each invocation; reasonable default value for PARALLEL based on the number and type of instances configured.
- Added
PigJob
!
HiveJob
critical bug fixed, now it works :)- Added
log_uri
andaction_on_failure
as options toHiveJob
. - Added integration tests to
HiveJob
.
- Added
HiveJob
, a simplified way to launch basic Hive job flows. - Added HISTORY.
- Added LICENSE.
- Released!