[
https://issues.apache.org/jira/browse/FLINK-21132?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17272503#comment-17272503
]
Kezhu Wang commented on FLINK-21132:
------------------------------------
[~pnowojski] For option#2, which is the current option, head operator indeed
receives end of input currently except for FLIP-27 source(see FLINK-21133). For
legacy source head operator, {{StreamSource}} will fire end of input and
{{Task}} will write {{EndOfPartitionEvent}} after successful run, this matches
my previous description. For no source head operator, after received all
{{EndOfPartitionEvent}} for every subpartitions in input,
{{StreamTaskNetworkInput}} will report {{InputStatus.END_OF_INPUT}} and
{{StreamOneInputProcessor}} will fire end of input, this matches what
[apache/iceberg#2033|https://github.com/apache/iceberg/issues/2033] happens in
later discussion. So, basically, FLINK-14228 is well implemented, it just has
assumption overrode by/conflict with FLIP-34.
Personally, I prefer option#1 since save-with-savepoint is just an utility, it
does not contribute to streaming data flow. I think maybe it could be
implemented in a way that FLINK-21133 solved without additional work from
source side.
> BoundedOneInput.endInput is called when taking synchronous savepoint
> --------------------------------------------------------------------
>
> Key: FLINK-21132
> URL: https://issues.apache.org/jira/browse/FLINK-21132
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Task
> Affects Versions: 1.10.2, 1.10.3, 1.11.3, 1.12.1
> Reporter: Kezhu Wang
> Priority: Major
>
> [~elkhand](?) reported on project iceberg that {{BoundedOneInput.endInput}}
> was
> [called|https://github.com/apache/iceberg/issues/2033#issuecomment-765864038]
> when [stopping job with
> savepoint|https://github.com/apache/iceberg/issues/2033#issuecomment-765557995].
> I think it is a bug of Flink and was introduced in FLINK-14230. The
> [changes|https://github.com/apache/flink/pull/9854/files#diff-0c5fe245445b932fa83fdaf5c4802dbe671b73e8d76f39058c6eaaaffd9639faL577]
> rely on {{StreamTask.afterInvoke}} and {{OperatorChain.closeOperators}} will
> only be invoked after *end of input*. But that is not true long before after
> [FLIP-34: Terminate/Suspend Job with
> Savepoint|https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=103090212].
> Task could enter state called
> [*finished*|https://github.com/apache/flink/blob/3a8e06cd16480eacbbf0c10f36b8c79a6f741814/flink-streaming-java/src/main/java/org/apache/flink/streaming/runtime/tasks/StreamTask.java#L467]
> after synchronous savepoint, that is an expected job suspension and stopping.
> [~sunhaibotb] [~pnowojski] [~roman_khachatryan] Could you help confirm this ?
> For full context, see
> [apache/iceberg#2033|https://github.com/apache/iceberg/issues/2033]. I have
> pushed branch
> [synchronous-savepoint-conflict-with-bounded-end-input-case|https://github.com/kezhuw/flink/commits/synchronous-savepoint-conflict-with-bounded-end-input-case]
> in my repository. Test case
> {{SavepointITCase.testStopSavepointWithBoundedInput}} failed due to
> {{BoundedOneInput.endInput}} called.
> I am also aware of [FLIP-147: Support Checkpoints After Tasks
> Finished|https://cwiki.apache.org/confluence/x/mw-ZCQ], maybe the three
> should align on what *finished* means exactly. [~kkl0u] [~chesnay]
> [~gaoyunhaii]
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
