BoWang created FLINK-11309:
------------------------------
Summary: Batch Job Failover Using RestartPipelinedRegionStrategy
Fails in Some Scenes
Key: FLINK-11309
URL: https://issues.apache.org/jira/browse/FLINK-11309
Project: Flink
Issue Type: Bug
Components: Core
Affects Versions: 1.7.1, 1.7.0, 1.6.2
Reporter: BoWang
Assignee: BoWang
Hi all,
When running the batch WordCount example, I configured the job execution mode
as *BATCH_FORCED*, and failover-strategy as *region*, I manual injected some
errors to let the execution fail in different phases. In some cases, the job
could recovery from failover and became succeed, but in some cases, the job
retried several times and failed.
Example:
# If the failure occurred before task read data, e.g., failed before
*invokable.invoke()* in Task.java, failover could succeed.
# If the failure occurred after task having read data, failover did not work.
Problem diagnose:
Running the example described before, each ExecutionVertex is defined as a
restart region, and the ResultPartitionType between executions is *BLOCKING.*
Thus, *SpillableSubpartition* and *SpillableSubpartitionView* are used to
write/read shuffle data, and data block is described as *BufferConsumer* stored
in a list called *buffers,* when task requires input data from
*SpillableSubpartitionView,* *BufferConsumer* is REMOVED from buffers. Thus,
when failures occurred after having read data, some *BufferConsumers* have
already released, although tasks retried, the input data is incomplete.
Fix Proposal:
# *BufferConsumer* should not be removed from buffers until *ExecutionVertex*
terminates.
# *SpillableSubpartition* should not be released until *ExecutionVertex*
terminates.
# Each *SpillableSubpartition* contains multi *SpillableSubpartitionView*s,
each of which is corresponding to a *Execution*.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
