[ https://issues.apache.org/jira/browse/BEAM-2393?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16402220#comment-16402220 ]
Aljoscha Krettek commented on BEAM-2393: ---------------------------------------- Ah, I think that explains it. This call in the source: https://github.com/apache/beam/blob/6148a6d063a0503ee435ab5084fcba3fb864b26f/runners/flink/src/main/java/org/apache/beam/runners/flink/translation/wrappers/streaming/io/UnboundedSourceWrapper.java#L333 will at some point invoke the ParDo because the operations are chained together, i.e. they are being executed in the same thread and "sending" elements downstream is just method calls. This means that the source will block for about 20s in that synchronized block, which keeps the checkpointing logic from acquiring the lock. > BoundedSource is not fault-tolerant in FlinkRunner Streaming mode > ----------------------------------------------------------------- > > Key: BEAM-2393 > URL: https://issues.apache.org/jira/browse/BEAM-2393 > Project: Beam > Issue Type: Bug > Components: runner-flink > Reporter: Jingsong Lee > Assignee: Grzegorz KoĊakowski > Priority: Major > > {{BoundedSourceWrapper}} does not implement snapshot() and restore(), when > the failure to restart, it will send duplicate data. -- This message was sent by Atlassian JIRA (v7.6.3#76005)