- 25 5月, 2017 2 次提交
- 24 5月, 2017 12 次提交
-
-
由 lingjinjiang 提交于
-
由 zentol 提交于
This closes #3948.
-
由 Nico Kruber 提交于
This closes #3966.
-
由 Nico Kruber 提交于
-> use a JUnit '@Rule' that does the cleanup This closes #3962.
-
由 Nico Kruber 提交于
-> use a JUnit '@Rule' that does the cleanup
-
由 zentol 提交于
This closes #3970.
-
由 zentol 提交于
This closes #3969.
-
由 zentol 提交于
This closes #3968.
-
由 sunjincheng121 提交于
This closes #3958.
-
由 Stefan Richter 提交于
-
由 Stefan Richter 提交于
[FLINK-6685] Prevent that SafetyNetCloseableRegistry is closed prematurely in Task::triggerCheckpointBarrier
-
由 Till Rohrmann 提交于
When recovering state from a ZooKeeperStateHandleStore it can happen that the deserialization fails, because one tries to recover state from an old Flink version which is not compatible. In this case we should output a better error message such that the user can easily spot the problem. This closes #3972.
-
- 23 5月, 2017 14 次提交
-
-
由 David Anderson 提交于
This closes #3964.
-
由 huafengw 提交于
This closes #3898.
-
由 Till Rohrmann 提交于
Let first the snapshotting thread completely finish before checking the test condition. This prevents race conditions between the checkpointing thread and the verification.
-
由 kkloudas 提交于
This is to leverage the fact that RocksDB already returns the keys sorted. So now elements, instead of being stores in a PQ and all of them being deserialized and serialized at each incoming element, the are stored in a MapState with the key being the timestamp and the value, a List of elements that refer to the same timestamp.
-
由 Till Rohrmann 提交于
The lifecycle of savepoints is not managed by the CheckpointCoordinator and fully in the hand of the user. Therefore, the CheckpointCoordinator cannot rely on them when trying to recover from failures. E.g. a user moving a savepoint shortly before a failure could completely break Flink's recovery mechanism because Flink cannot skip failed checkpoints when recovering. Therefore, until Flink is able to skip failed checkpoints when recovering, we should not add savepoints to the CompletedCheckpointStore which is used to retrieve checkpoint for recovery. The distinction of a savepoint is done on the basis of the CheckpointProperties (CheckpointProperties.STANDARD_SAVEPOINT). This closes #3965.
-
由 Greg Hogan 提交于
Updates / reverts the import order by sections: - org.apache.flink.* - all other imports - javax.* - java.* - static imports Adds EmptyLineSeparator to enforce an extra newline (not enforced between field or local variable declarations). This closes #3941
-
由 zentol 提交于
-
由 zentol 提交于
-
由 zentol 提交于
-
由 zentol 提交于
-
由 zentol 提交于
-
由 zentol 提交于
This closes #3947.
-
由 zjureel 提交于
This closes #3935.
-
由 sunjincheng121 提交于
This closes #3944.
-
- 22 5月, 2017 5 次提交
-
-
由 Tzu-Li (Gordon) Tai 提交于
This commit adds also the nested serializers themselves to the configuration snapshots of composite serializers. This opens up the oppurtunity to use the previous nested serializer as the convert deserializer in the case that a nested serializer in the new serializer determines that state migration is required. This commit also consolidate all TypeSerializer-related serialization proxies into a single utility class. This closes #3937.
-
由 twalthr 提交于
-
由 Till Rohrmann 提交于
The ClusterClient starts its ActorSystem lazily. In order to find out the address to which to bind, the ClusterClient tries to connect to the JobManager. In order to find out the JobManager's address it is important to use the HighAvailabilityServices instead of retrieving the address information from the configuration, because otherwise it conflicts with HA mode. This closes #3949.
-
由 Till Rohrmann 提交于
The ClientConnectionTest passed even though it was failing the test because we were expecting an exception and checking a special word to contained in the exception's message. Unfortunately, we generated an AssertionError with the same word if the actual logic we wanted to test failed. That cause the test to pass.
-
由 Stefan Richter 提交于
-
- 20 5月, 2017 7 次提交
-
-
由 zentol 提交于
This closes #3955.
-
由 zentol 提交于
This closes #3954.
-
由 zentol 提交于
This closes #3953.
-
由 Stefan Richter 提交于
-
由 Till Rohrmann 提交于
The ClusterClient should not call HighAvailabilityServices#closeAndCleanupAllData when being shut down. The reason is that this call will delete all HA data needed for a future recovery. Only the JobManager should be allowed to decide when to discard HA data or not.
-
由 zjureel 提交于
This closes #3904.
-
由 zentol 提交于
This closes #3907.
-