[
https://issues.apache.org/jira/browse/FLINK-2462?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14699218#comment-14699218
]
ASF GitHub Bot commented on FLINK-2462:
---------------------------------------
Github user aljoscha commented on a diff in the pull request:
https://github.com/apache/flink/pull/1017#discussion_r37168797
--- Diff:
flink-staging/flink-streaming/flink-streaming-core/src/main/java/org/apache/flink/streaming/runtime/tasks/SourceStreamTask.java
---
@@ -39,61 +37,40 @@
*/
public class SourceStreamTask<OUT> extends StreamTask<OUT,
StreamSource<OUT>> {
- private static final Logger LOG =
LoggerFactory.getLogger(SourceStreamTask.class);
-
@Override
- public void invoke() throws Exception {
- final SourceOutput<StreamRecord<OUT>> output = new
SourceOutput<StreamRecord<OUT>>(outputHandler.getOutput(), checkpointLock);
-
- boolean operatorOpen = false;
-
- if (LOG.isDebugEnabled()) {
- LOG.debug("Task {} invoked", getName());
- }
-
- try {
- openOperator();
- operatorOpen = true;
-
- streamOperator.run(checkpointLock, output);
-
- closeOperator();
- operatorOpen = false;
-
- if (LOG.isDebugEnabled()) {
- LOG.debug("Task {} invocation finished",
getName());
- }
-
- }
- catch (Exception e) {
- LOG.error(getEnvironment().getTaskNameWithSubtasks() +
" failed", e);
-
- if (operatorOpen) {
- try {
- closeOperator();
- }
- catch (Throwable t) {
- LOG.warn("Exception while closing
operator.", t);
- }
- }
- throw e;
- }
- finally {
- this.isRunning = false;
- // Cleanup
- outputHandler.flushOutputs();
- clearBuffers();
- }
+ protected void init() {
+ // does not hold any resources, so no initialization needed
+ }
+ @Override
+ protected void cleanup() {
+ // does not hold any resources, so no cleanup needed
}
+
@Override
- public void cancel() {
- super.cancel();
+ protected void run() throws Exception {
+ final Object checkpointLock = getCheckpointLock();
+
+ final SourceOutput<StreamRecord<OUT>> output =
+ new
SourceOutput<StreamRecord<OUT>>(outputHandler.getOutput(), checkpointLock);
+
+ streamOperator.run(checkpointLock, output);
+ }
+
+ @Override
+ protected void cancelTask() throws Exception {
streamOperator.cancel();
}
+ //
------------------------------------------------------------------------
+
+ // TODO:
+ // does this help with anything? The losk should be already held by the
source function that
+ // emits. If that one does not hold the lock, then this does not help
either.
+
--- End diff --
This is for the case where the source does not actually acquire the lock
(because it is not participating in checkpointing). This might be a degenerate
case, however, and not really necessary.
> Wrong exception reporting in streaming jobs
> -------------------------------------------
>
> Key: FLINK-2462
> URL: https://issues.apache.org/jira/browse/FLINK-2462
> Project: Flink
> Issue Type: Bug
> Components: Streaming
> Affects Versions: 0.10
> Reporter: Stephan Ewen
> Assignee: Stephan Ewen
> Priority: Blocker
> Fix For: 0.10
>
>
> When streaming tasks are fail and are canceled, they report a plethora of
> followup exceptions.
> The batch operators have a clear model that makes sure that root causes are
> reported, and followup exceptions are not reported. That makes debugging much
> easier.
> A big part of that is to have a single consistent place that logs exceptions,
> and that has a view of whether the operation is still running, or whether it
> has been canceled.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
