Bug Fixes
Bug Fixes
Internal Changes
Features
Bug Fixes
Breaking Changes
- Remove deprecated SDK API code (#671) (fc50536), closes #671
- Remove custom scripts (#673) (9e32b4b), closes #673
- Replaced states with Enums for ClusterState, JobState, ApplicationState (#677) (e486536), closes #677
Features
- Spark retry docker pull (#672) (18b74e4), closes #672
- Spark scheduling target (#661) (4408c4f), closes #661
- Spark submit scheduling internal (#674) (8c2bf0c), closes #674
Bug Fixes
- Fix: pin all node dependencies not in Pipfile (#667) (0606598), closes #667
- Fix: vsts integration tests block (#657) (4a60c8a), closes #657
- Fix: vsts mutliline secrets (#668) (cb62207), closes #668
Breaking Changes
- spark roll back scheduling disable (#653) (93615d9), closes #653
- remove custom scripts (#650) (442228a), closes #650
- 0.9.0 deprecated code removal (#645) (eef36dc), closes #645
- SDK refactor (#622) (b18eb69), closes #622
Features
- Add ability to specify docker run options in toolkit config (#613) (9d554c3), closes #613 #3
- add brief flag to debug tool (#634) (b7bdd8c), closes #634
- first run docs update (#644) (9098533), closes #644
- SDK refactor (#622) (b18eb69), closes #622
Bug Fixes
- diagnostics function write error result bug (#649) (293f297), closes #649
- expose get cluster configuration API (#648) (7c14648), closes #648
- remove bad node scripts import (#652) (0a9ce94), closes #652
- typo in vsts build (#654) (7c37b06), closes #654
- update incompatible dependencies in setup.py (#639) (f98d037), closes #639
Internal Changes
- fix pylint warnings (#651) (828162e), closes #651
- update vsts build (#635) (6eda21e), closes #635
- verify code formatting in build (#633) (7730c46), closes #633
Bug Fixes
- docs links version (#614) (a8f8e92), closes #614
- set defaults for SparkConfiguration, add tests (#606) (5306a2a), closes #606
- spark debug tool filter out .venv, make debug tool testable (#612) (4e0b1ec), closes #612
- Suppress msrest warnings (#611) (883980d), closes #611
Deprecated Features
- ClusterConfiguration fields vm_count and vm_count_low_pri have been renamed to size and size_low_priority
- command line flag
--size-low-pri
foraztk spark cluster create
has been replaced with--size-low-priority
default
secrets.yaml block has been deprecated, place all child parameters directly at the root- Spark version 1.6 has been deprecated
Added Features
- add cluster list quiet flag, ability to compose with delete (#581) (88d0419), closes #581
- add node run command (#572) (af449dc), closes #572
- Add VSTS CI (#561) (66037fd), closes #561
- Disable scheduling on group of nodes (#540) (8fea9ce), closes #540
- New Models design with auto validation, default and merging (#543) (02f336b), closes #543
- nvBLAS and OpenBLAS plugin (#539) (603a413), closes #539
- pure python ssh (#577) (f16aac0), closes #577
- Support passing of remote executables via aztk spark cluster submit (#549) (f6735cc), closes #549
- TensorflowOnSpark python plugin (#525) (1527929), closes #525
- Conda, Apt-Get and Pip Install Plugins (#594) (fbf1bab), closes #594
- Warnings show stacktrace on verbose (#587) (b9a863b), closes #587
Bug Fixes
- add toolkit to sdk docs and example (d688c9c)
- --size-low-pri being ignored (#593) (fa3ac0e), closes #593
- fix typos (#595) (7d7a814), closes #595
- getting started script reuse aad application (#569) (3d16cf3), closes #569
- models v2 deserialization (#584) (1eeff23), closes #584
- optimize start task (#582) (e5e529a), closes #582
- remove deprecated vm_count call (#586) (dbde8bc), closes #586
- Remove old spark-defaults.conf jars (#567) (8b8cd62), closes #567
- set logger to stdout (#588) (3f0c8f9), closes #588
- switch create user to pool wide (#574) (49a890a), closes #574
- switch from pycryptodome to pycryptodomex (#564) (19dde42), closes #564
- allow cluster config to be printed when no username has been set (#597) (1cc71c7), closes #597
Bug Fixes
- Fix: create virtual environment even if container exists (2db7b00)
- Fix: gitattributes for jar files (#548) (a18660b), closes #548
- Fix: pass docker repo command back to the cluster config (#538) (a99bbe1), closes #538
AZTK is now published on pip! Documentation has migrated to readthedocs
This release includes a number of breaking changes. Please follow the migration for upgrading from 0.6.0..
Breaking Changes
- Moved
docker_repo
under a newtoolkit
key.docker_repo
is now only used for custom Docker images. Use toolkit for supported images. - Docker images have been refactored and moved to a different Dockerhub repository. The new supported images are not backwards compatible. See the documentation on configuration files.
Deprecated Features
- Custom scripts have been removed in favor of Plugins, which are more robust. See, the documentation on Plugins.
Added Features
- add internal flag to node commands (#482) (1eaa1b6), closes #482
- Added custom scripts functionality for plugins with the cli(Deprecate custom scripts) (#517 (c98df7d), closes #517
- disable msrestazure keyring log (#509) (3cc43c3), closes #509
- enable mixed mode for jobs (#442) (8d00a2c), closes #442
- getting started script (#475) (7ef721f), closes #475
- JupyterLab plugin (#459) (da61337), closes #459
- managed storage for clusters and jobs (#443) (8aa1843), closes #443
- match cluster submit exit code in cli (#478) (8889059), closes #478
- Plugin V2: Running plugin on host (#461) (de78983), closes #461
- Plugins (#387) (c724d94), closes #387
- Pypi auto deployment (#428) (c237501), closes #428
- Readthedocs support (#497) (e361c3b), closes #497
- refactor docker images (#510) (779bffb), closes #510
- Spark add output logs flag (#468) (32de752), closes #468
- spark debug tool (#455) (44a0765), closes #455
- spark ui proxy plugin (#467) (2e995b4), closes #467
- Spark vnet custom dns hostname fix (#490) (61e7c59), closes #490
- New Toolkit configuration (#507) (7a7e63c), closes #507
Bug Fixes
- add gitattributes file (#470) (82ad029), closes #470
- add plugins to cluster_install_cmd call (#423) (216f63d), closes #423
- add spark.history.fs.logDirectory to required keys (#456) (4ef3dd0), closes #456
- add support for jars, pyfiles, files in Jobs (#408) (2dd7891), closes #408
- add timeout handling to cluster_run and copy (#524) (47000a5), closes #524
- azure file share not being shared with container (#521) (07ac9b7), closes #521
- Dependency issue with keyring not having good dependencies (#504) (5e79a2c), closes #504
- filter job submission clusters out of cluster list (#409) (1c31335), closes #409
- fix aztk cluster submit paths, imports (#464) (c1f43c7), closes #464
- fix broken spark init command (#486) (a33bdbc), closes #486
- fix job submission cluster data issues (#533) (9ccc1c6), closes #533
- fix spark job submit path (#474) (ee1e61b), closes #474
- make node scripts upload in memory (#519) (0015e22), closes #519
- pypi long description (#450) (db7a2ef), closes #450
- remove unnecessary example (#417) (f1e3f7a), closes #417
- Remove unused ssh plugin flags (#488) (be8cd2a), closes #488
- set explicit file open encoding (#448) (5761a36), closes #448
- Spark shuffle service worker registration fail (#492) (013f6e4), closes #492
- throw error if submitting before master elected (#479) (a59fe8b), closes #479
- hdfs using wrong conditions (#515) (a00dbb7), closes #515
- AZTK_IS_MASTER not set on worker and failing (#506) (b8a3fcc), closes #506
- VNet required error now showing if using mixed mode without it (#440) (9253aac), closes #440
- Worker on master flag ignored and standardize boolean environment (#514) (5579d95), closes #514
- Fix job configuration option for
aztk spark job submit
command (#435) (4be5ac2), closes #435 - Fix keyring (#505) (12450fb), closes #505
- Fix the endpoint (#437) (bcefca3), closes #437
- Fix typo in command_builder 'expecity' -> 'explicitly' (#447) (27822f4), closes #447
- Fix typo load_aztk_screts -> load_aztk_secrets (#421) (6827181), closes #421
- Update file to point at master branch (#501) (4ba3c9d), closes #501
- Update storage sdk from 0.33.0 to 1.1.0 (#439) (f2eb1a4), closes #439
Internal Changes
- Internal: Cluster data helpers and upload_node_script into cluster_data module (#401) (2bed496), closes #401
- Internal: Move node scripts under aztk and upload all aztk to cluster (#433) (dfbfead), closes #433
Features:
aztk spark init
customization flagsaztk spark cluster run
command addedaztk spark cluster copy
command added- enable Spark dynamic allocation by default
- add SDK support for file-like objects
- add Spark integration tests
- add flag
worker_on_master
option for cluster and job submission mode - Spark driver runs on master node for single application job submission mode
Bug Fixes:
- load jars in
.aztk/jars/
in job submission mode - replace outdated error in cluster_create
- fix type error crash if no jars are specified in job submission
- stop using mutable default parameters
- print job application code if exit_code is 0
- job submission crash if executor or driver cores specified
- wrong error thrown if user added before master node picked
Major Features:
- Job Submission
- AAD Support, see Getting Started
- VNET Support
Breaking changes:
SecretsConfiguration
inputs changed. Check in SDK for the new format