- 27 5月, 2015 13 次提交
-
-
由 Robert Metzger 提交于
This closes #717
-
由 Robert Metzger 提交于
-
由 Robert Metzger 提交于
-
由 Pietro Pinoli 提交于
This closes #721
-
由 Fabian Hueske 提交于
This closes #712
-
由 vasia 提交于
added SSSP example for vertex-centric; added a comparison section for vertex-centric and GSA This closes #722
-
由 andralungu 提交于
This closes #678
-
由 andralungu 提交于
This squashes the following commits: [gelly] Removed trailing comment [gelly] Made remove methods use a coGroup fun
-
由 Stephan Ewen 提交于
This closes #728
-
由 Ufuk Celebi 提交于
-
由 Theodore Vasiloudis 提交于
Added tables for parameter values Adds links to optimization parameter values in the description This closes #715.
-
由 hsaputra 提交于
The FlinkMiniCluster contain member variable called HOSTNAME which all caps. The naming of al caps usually reserved for constants and static variable. The PR is changing the name of the variable and the usages. Author: hsaputra <hsaputra@apache.org> Closes #711 from hsaputra/update_hostname_to_match_conventions and squashes the following commits: 2a2f3bd [hsaputra] Change FlinkMiniCluster#HOSTNAME to FlinkMiniCluster#hostname to match naming convention.
-
由 Ufuk Celebi 提交于
-
- 26 5月, 2015 17 次提交
-
-
由 Maximilian Michels 提交于
-
由 Fabian Hueske 提交于
-
由 Stephan Ewen 提交于
-
由 Stephan Ewen 提交于
This closes #718
-
由 Stephan Ewen 提交于
- Streaming mode sets the memory manager to lazy memory allocation to ensure heap is not blocked by batch memory manager
-
由 Stephan Ewen 提交于
[FLINK-2085] [runtime] Add an option to let the MemoryManager allocate and release memory as needed. This is an alternative mode to the current mode that pre-allocates all memory. The default remains to pre-allocate all memory.
-
由 Ufuk Celebi 提交于
-
由 Robert Metzger 提交于
This closes #689
-
由 Ufuk Celebi 提交于
-
由 Till Rohrmann 提交于
[FLINK-2053] [ml] Adds automatic type registration of flink-ml types. Adds de-duplication of registered types at ExecutionConfig. Fixes bug in Breeze SparseVector to Flink SparseVector conversion. This closes #723.
-
由 Ufuk Celebi 提交于
-
由 Theodore Vasiloudis 提交于
This closes #714.
-
由 Ufuk Celebi 提交于
-
由 Aljoscha Krettek 提交于
-
由 Aljoscha Krettek 提交于
Before this, a sliding time window would keep emitting the last result because the number of elements per pre-aggregation result was not correctly reset on eviction.
-
由 Gyula Fora 提交于
-
由 Ufuk Celebi 提交于
Problem: Failures in the network stack were not properly handled and correctly attributed. Solution: Failures are always attributeed to the client (consumer). This change introduces TransportException, which indicates whether the problem ocurred locally or remotely. This makes it easy to reason about the source of a problem. This closes #713.
-
- 25 5月, 2015 2 次提交
-
-
由 Gyula Fora 提交于
-
由 Gyula Fora 提交于
-
- 24 5月, 2015 1 次提交
-
-
由 mbalassi 提交于
Closes #697
-
- 22 5月, 2015 7 次提交
-
-
由 Aljoscha Krettek 提交于
This was using != before, now uses !(a.equals(b))
-
由 Gyula Fora 提交于
-
由 Ufuk Celebi 提交于
Problem: cancelling of tasks sometimes leads to misleading error messages about "not found partitions". This is an artifact of task cancelling. If a task (consumer) consumes data from another remote task (producer), its sends a partition request over the network. If the producer fails concurrently with this request, the request returns with a PartitioNotFoundException to the consumer. If this error message is received *before* the consumer is cancelled (as a result of the failing producer), you see the misleading error being attributed to the consumer. This makes it hard to trace the root cause of the problem (the failing producer). Solution: when a consumer receives a remote PartitionNotFoundException, it asks the central job manager whether the producer is still running or has failed. If the producer is still running, the partition request is send again (using an exponential back off). If the following requests fail again, the consumer fails with a PartitionNotFoundException. If the producer has failed, the consumer is cancelled. If the producer is not running and has not failed, there is a bug either in the consumer task setup (e.g. requesting a non-existing result) or in the network stack (e.g. unsafe publication of produced results), in which case the error is attributed to the consumer. --- The new Akka messages introduced with this change are only exchanged in error cases and don't affect normal operation. Normal operation (not affected by this change): - TM1=>TM2: request result - TM2=>TM1: result Error case: - TM1=>TM2: request result - TM2=>TM1: PartitionNotFoundException - TM1=>JM: check partition state - JM=>TM1: retrigger request -OR- cancel consumer This closes #705.
-
由 Till Rohrmann 提交于
Adds pipeline comments Adds pipeline IT case
-
由 Till Rohrmann 提交于
[FLINK-2050] [ml] Introduces new pipelining mechanism using implicit classes to wrap the algorithm logic This closes #704.
-
由 Theodore Vasiloudis 提交于
Also added attribution for some of the Latex in optimization framework. This closes #688.
-
由 Theodore Vasiloudis 提交于
Added test case for convergence This closes #692.
-