- 10 3月, 2015 7 次提交
-
-
由 Fabian Hueske 提交于
This closes #458
-
由 Stephan Ewen 提交于
-
由 Stephan Ewen 提交于
-
由 Stephan Ewen 提交于
-
由 Stephan Ewen 提交于
-
由 Henry Saputra 提交于
Add parentheses on Exception#getMessage calls from pattern of: "Initializing the input processing failed" + e.getMessage() == null ? "." : ": " + e.getMessage() to: "Initializing the input processing failed" + (e.getMessage() == null ? "." : ": " + e.getMessage()) Extra parentheses needed to make sure ternary operator check on e.getMessage scope call. Author: Henry Saputra <henry.saputra@gmail.com> Closes #461 from hsaputra/fix_parentheses_exception_getmessage and squashes the following commits: 1345cbf [Henry Saputra] Fix checking null for Exception#getMessage call from pattern of:
-
由 Henry Saputra 提交于
JobManagerStartupTest should not assume user could not create /does-not-exist-no-sir dir. Sometimes, user that run tests could have write access to root directory so creating /does-not-exist-no-sir is possible, hence no exception thrown. Need to construct a Flink test directory under directory specified under "java.io.tmpdir" and change the permission to not allow create new directory. Author: Henry Saputra <henry.saputra@gmail.com> Closes #460 from hsaputra/FLINK-1651_fix_jobmanager_fail_test and squashes the following commits: 5d46716 [Henry Saputra] [FLINK-1651] Fix test case at JobManagerStartupTest should not assume user could not create /does-not-exist-no-sir dir.
-
- 09 3月, 2015 4 次提交
-
-
由 Robert Metzger 提交于
This closes #450
-
由 mbalassi 提交于
This closes #463
-
由 Akshay Dixit 提交于
-
由 Gábor Hermann 提交于
-
- 08 3月, 2015 1 次提交
-
-
由 andralungu 提交于
This closes #440
-
- 07 3月, 2015 1 次提交
-
-
由 Henry Saputra 提交于
-
- 06 3月, 2015 2 次提交
-
-
由 Henry Saputra 提交于
This PR consists of: 1. Remove unnecessary brackets in ExecutionVertex#getPreferredLocations 2. Throw illegal argument with proper message for consistencies of not nullable argument. 3. Add final modifier to RecordWriter#serializers since the content being used as lock. 4. Wrap too long of lines in some of the Java code for readibility. 5. Add missing JavaDoc parameter. 6. Remove final modifier in the OutputEmitter's private methods because it is redundant. Author: Henry Saputra <henry.saputra@gmail.com> Closes #457 from hsaputra/cleanup_javadoc_and_longlines_1 and squashes the following commits: 8815302 [Henry Saputra] Some cleanups and doc updates while looking at runtime package: 1. Remove unnecessary brackets in ExecutionVertex#getPreferredLocations 2. Throw illegal argument with proper message for consistencies of not nullable argument. 3. Add final modifier to RecordWriter#serializers since the content being used as lock. 4. Wrap too long of lines in some of the Java code for readibility. 5. Add missing JavaDoc parameter. 6. Remove final modifier in the OutputEmitter's private methods because it is redundant.
-
由 Fabian Hueske 提交于
This closes #453
-
- 05 3月, 2015 17 次提交
-
-
由 Stephan Ewen 提交于
This closes #455
-
由 Stephan Ewen 提交于
This closes #456
-
由 Maximilian Michels 提交于
-
由 mbalassi 提交于
This closes #449
-
由 Gyula Fora 提交于
[FLINK-1625] [streaming] [api-breaking] Added proper cancellation to StreamInvokables + Sink- and SourceFunction interfaces extended with cancel method
-
由 Gyula Fora 提交于
[FLINK-1625] [streaming] Refactored StreamVertex and subclasses to clean up after invoke and properly log and propagate exceptions
-
由 vasia 提交于
-
由 balidani 提交于
This closes #441
-
由 balidani 提交于
-
由 andralungu 提交于
This closes #429
-
由 andralungu 提交于
-
由 Stephan Ewen 提交于
-
由 Stephan Ewen 提交于
-
由 Ufuk Celebi 提交于
Add a hint about how to configure the number of network buffers.
-
由 Stephan Ewen 提交于
-
由 Stephan Ewen 提交于
- Fix bugs with non-serializable messages - Separate parser and action logic - Clean up tests - Vastly improve logging in CLI client - Additional tests for parsing / config setup in the command line client
-
由 Ufuk Celebi 提交于
The fixed race condition could lead to an infinite loop, in which the network I/O thread is continuously trying to request a buffer from an already destroyed buffer provider. The problem was noticed in certain builds during recovery test cases, where the blocked network I/O thread prevented restarted tasks from connecting to remote task managers.
-
- 04 3月, 2015 6 次提交
-
-
由 Till Rohrmann 提交于
-
由 Robert Metzger 提交于
-
由 Stephan Ewen 提交于
-
由 Stephan Ewen 提交于
-
由 Robert Metzger 提交于
-
由 Robert Metzger 提交于
This closes #415
-
- 03 3月, 2015 2 次提交
-
-
由 Ufuk Celebi 提交于
Task thread Exceptions were always logged, even if the task was cancelled or failed externally. In certain scenarios this lead to misleading error messages and made it hard to figure out the root exception. When an exception is thrown w/o the task being canceled or failed, the error is still handled via Task#markFailed(Throwable).
-
由 Robert Metzger 提交于
-