- 14 10月, 2017 3 次提交
-
-
由 Stephan Ewen 提交于
This adds two implementations of a file system that write to S3. Both are not actual re-implementations but wrap other implementations and shade dependencies. (1) A wrapper around Hadoop's s3a file system. By pulling a smaller dependency tree and shading all dependencies away, this keeps the appearance of Flink being Hadoop-free, from a dependency perspective. (2) The second S3 file system is from the Presto Project. Initial simple tests seem to indicate that it responds slightly faster and in a bit more lightweight manner to write/read/list requests, compared to the Hadoop s3a FS, but it has some semantical differences. This closes #4818
-
由 Chris Ward 提交于
This closes #4816
-
由 Stephan Ewen 提交于
-
- 13 10月, 2017 5 次提交
-
-
由 Aljoscha Krettek 提交于
This also adds a test that verifies the deep copy.
-
由 Aljoscha Krettek 提交于
Pinning it to the latest version means the order in which we deploy to maven doesn't matter anymore.
-
由 kkloudas 提交于
-
由 Fabian Hueske 提交于
-
由 Till Rohrmann 提交于
This commit implements the ClusterOverview generation on the Dispatcher. In order to do this, the Dispatcher requests the ResourceOverview from the ResourceManager and the job status from all JobMasters. After receiving all information, it is compiled into the ClusterOverview. Note: StatusOverview has been renamed to ClusterOverview This closes #4793.
-
- 12 10月, 2017 11 次提交
-
-
由 Till Rohrmann 提交于
This closes #4811.
-
由 Bowen Li 提交于
-
由 Bowen Li 提交于
-
由 Aljoscha Krettek 提交于
-
由 gyao 提交于
This closes #4791.
-
由 gyao 提交于
-
由 Aljoscha Krettek 提交于
This way of passing parameters does not work with the streaming API and is actually confusing for users. This closes #4797.
-
由 zentol 提交于
This closes #7807.
-
由 zentol 提交于
This closes #4800.
-
由 Zhijiang 提交于
This closes #4698.
-
由 Till Rohrmann 提交于
The test did not properly wait for the registration of the TaskExecutor at the ResourceManager. Therefore, it could come to a race condition between sending an initial SlotReport and a separate message for the newly added slots.
-
- 11 10月, 2017 16 次提交
-
-
由 kkloudas 提交于
Previously the QueryableStateClient could connect to the JM and the TMs directly to fetch the required state. Now, there is a proxy running on each TM and the remote client connects to one of these proxies in order to get its state. The proxy receives the request from the client, performs all necessary message exchanges within the Flink cluster, receives the state and forwards it back to the client. This architecture allows for more security features to be integrated in the future, as the proxy is running in the cluster, it exposes less information about the cluster to the outside world, and is more suitable for containerized environments.
-
由 kkloudas 提交于
Creates a separate for the queryable state and moves the client code outside the runtime. The Task Manager is now instantiating the KvStateServer using reflection.
-
由 Till Rohrmann 提交于
This closes #4772.
-
由 Till Rohrmann 提交于
-
由 zentol 提交于
This closes #4789.
-
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 #4792.
-
由 军长 提交于
This closes #4624.
-
由 Shuyi Chen 提交于
This closes #4585.
-
由 Xpray 提交于
This closes #4785.
-
由 Fabian Hueske 提交于
-
由 Xingcan Cui 提交于
This closes #4625.
-
由 yew1eb 提交于
This closes #4768.
-
- 10 10月, 2017 5 次提交
-
-
由 Till Rohrmann 提交于
Disable failing when not all creator properties are known Move CheckpointStatsCache out of legacy package; Remove unused CheckpointingStatistics#generateCheckpointStatistics method Remove JsonInclude.Include.NON_NULL from CheckpointStatistics; Pull null check out of CheckpointStatistics#generateCheckpointStatistics; Make CheckpointStatistics#checkpointStatisticcsPerTask non nullable; Add fail on missing creator property This closes #4763.
-
由 zentol 提交于
This close s#4788.
-
由 Zhijiang 提交于
This closes #4735.
-
由 Zhijiang 提交于
This closes #4499.
-
由 Zhijiang 提交于
This closes #4485.
-