- 24 6月, 2016 2 次提交
-
-
由 Maximilian Michels 提交于
This removes repeated printing of messages retrieved from the Yarn cluster. Only new messages are printed. - reduce waiting time between subsequent cluster queries This closes #2147
-
由 Maximilian Michels 提交于
- remove ActorRunner thread, print status in finalizeCluster instead - prevent premature shutdown of actor system in shutdown method - prevent timeout exceptions due to poisoning the ApplicationClient
-
- 23 6月, 2016 9 次提交
-
-
由 Till Rohrmann 提交于
The SUSPENDED job status is a new ExecutionGraph state which can be reached from all non-terminal states when calling suspend on the ExecutionGraph. Unlike the FAILED, FINISHED and CANCELED state, the SUSPENDED state does not trigger the deletion of the job from the HA storage. Therefore, this state can be used to handle the loss of leadership or the shutdown of a JobManager so that the ExecutionGraph is stopped but can still be recovered. SUSPENDED is also a terminal state but it can be differentiated as a locally terminal state from FAILED, CANCELED and FINISHED which are globally terminal states. Add test case for suspend signal Add test case for suspending restarting job Add test case for HA job recovery when losing leadership Add online documentation for the job status Add ASF license header to job_status.svg Not throw exception when calling ExecutionGraph.restart and job is in state SUSPENDED This closes #2096.
-
由 Till Rohrmann 提交于
A job can get stuck in FAILING if fail is called on a restarting job which has not yet reset its ExecutionJobVertices, because these vertices would not call jobVertexInFinalState. This method, however, must be called in order to transition from FAILING to FAILED. Accept state FAILED when calling ExecutionGraph.restart This closes #2095.
-
由 twalthr 提交于
This closes #2088.
-
由 Ufuk Celebi 提交于
- Adds two new config keys for the start up scripts: * env.java.opts.jobmanager * env.java.opts.taskmanager - These are used in addition to the regular env.java.opts for the JobManager and TaskManager, respectively. - Current behaviour is not changed. This does not address the JMX port setting, because that is handled differently. This closes #2143.
-
由 Ufuk Celebi 提交于
This closes #2069.
-
由 Ufuk Celebi 提交于
- Adds a deserialization schema from byte[] to Row to be used in conjunction with the Table API.
-
由 chobeat 提交于
This closes #1996.
-
由 Greg Hogan 提交于
Update ValueTypeInfo to use specialized serializers and comparators, many of which were already present. This closes #1983
-
由 kl0u 提交于
This closes #2142
-
- 22 6月, 2016 6 次提交
-
-
由 Andrea Sella 提交于
[FLINK-4097] Fix NullPointerException on CassandraSinkBase and CassandraTupleWriteAheadSink's close() This closes #2144
-
由 MaBiao 提交于
This closes #2111
-
由 Shannon Quinn 提交于
This closes #2136 Create a task_id message in PythonStreamer that is passed through to the underlying process and included in the RuntimeContext, where it is accessible to the user and to functions.
-
由 Shahin 提交于
This closes #2130
-
由 zentol 提交于
This closes #2133
-
由 Aljoscha Krettek 提交于
By default this is set to disabled because there are known issues when users specify a custom TypeInformation.
-
- 21 6月, 2016 9 次提交
-
-
由 Ufuk Celebi 提交于
-
由 Ufuk Celebi 提交于
-
由 Ufuk Celebi 提交于
-
由 Ufuk Celebi 提交于
-
由 Aljoscha Krettek 提交于
I added this once with the though of varying the start time for window tests but it just makes stuff harder to parse for now.
-
由 Aljoscha Krettek 提交于
These will make it easier to read the tests for our future selves and others.
-
由 Aljoscha Krettek 提交于
-
由 Aljoscha Krettek 提交于
In onElement() we registered a timer for the case where the watermark is already past the end of the window and we're firing anyways. Now, only add a timer if the watermark is not already past the end of the window.
-
由 kl0u 提交于
Handle late elements and take care of cleaning the window state.
-
- 20 6月, 2016 4 次提交
-
-
由 MaBiao 提交于
This closes #2132
-
由 Maximilian Michels 提交于
-
由 Stefan Richter 提交于
This closes #2116.
-
由 Stefan Richter 提交于
This closes #2117.
-
- 19 6月, 2016 3 次提交
-
-
由 Fabian Hueske 提交于
This closes #2122
-
由 zentol 提交于
This closes #2119
-
由 Fabian Hueske 提交于
- setOutputType calls are forwarded to wrapped functions. - test added for InternalWindowFucntions. This closes #2118
-
- 17 6月, 2016 7 次提交
-
-
由 Trevor Grant 提交于
This closes #1898.
-
由 Till Rohrmann 提交于
Hardens ZooKeeperElectionTest by allowing the testing listener to return out-dated leader information. This can happen if the ZooKeeper connection was suspended and the new leader information has not been sent to the testing listener. In this case, the testing listener will be queried again to return the actual leader information. Add debug statements to ZooKeeperLeaderElectionTest.testZooKeeperReelection This closes #2103.
-
由 Maximilian Michels 提交于
-
由 Ufuk Celebi 提交于
Trying to prevent failures like [1] from happening again. I could not explain who deleted the savepoint file concurrently with the exists check. The savepoint is triggered and retrieved successfully. Shutting down the cluster does not remove any savepoints. [1] https://s3.amazonaws.com/archive.travis-ci.org/jobs/136396433/log.txt
-
由 kl0u 提交于
-
由 Maximilian Michels 提交于
-
由 Maximilian Michels 提交于
- integrates with and extends the refactoring of FLINK-3667 - enables to resume from Yarn properties or Yarn application id - introduces additional StandaloneClusterDescriptor - introduces DefaultCLI to get rid of standalone mode switches in CliFrontend - various fixes and improvements - remove legacy code from CliFrontend - change activation code of CustomCommandLine interface - use checked exceptions to signal supported operations - remove all checked exceptions of type Exception - fix logging and reduce verbosity of per-job clusters - print 'id' argument in YarnSessionCli - minor renaming of methods names - improve documentation - deprecate streaming option - extend CliFrontendYarnAddressConfigurationTest - move loading of custom CLIs to CliFrontend This closes #2085
-