[jira] [Commented] (FLINK-7883) Make savepoints atomic with respect to state and side effects
[ https://issues.apache.org/jira/browse/FLINK-7883?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17324753#comment-17324753 ] Konstantin Knauf commented on FLINK-7883: - [~cheleb] I think, this ticket should be closed. We by now have self-contained savepoints with the "stop" command. [~AHeise] what do you think? > Make savepoints atomic with respect to state and side effects > - > > Key: FLINK-7883 > URL: https://issues.apache.org/jira/browse/FLINK-7883 > Project: Flink > Issue Type: Improvement > Components: API / DataStream, Connectors / Kafka, Runtime / State > Backends >Affects Versions: 1.3.2, 1.4.0 >Reporter: Antoine Philippot >Priority: Major > > For a cancel with savepoint command, the JobManager trigger the cancel call > once the savepoint is finished, but during the savepoint execution, kafka > source continue to poll new messages which will not be part of the savepoint > and will be replayed on the next application start. > A solution could be to stop fetching the source stream task before triggering > the savepoint. > I suggest to add an interface {{StoppableFetchingSourceFunction}} with a > method {{stopFetching}} that existant SourceFunction implementations could > implement. > We can add a {{stopFetchingSource}} property in > {{CheckpointOptions}} class to pass the desired behaviour from > {{JobManager.handleMessage(CancelJobWithSavepoint)}} to > {{SourceStreamTask.triggerCheckpoint}} -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (FLINK-7883) Make savepoints atomic with respect to state and side effects
[ https://issues.apache.org/jira/browse/FLINK-7883?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17324335#comment-17324335 ] Olivier NOUGUIER commented on FLINK-7883: - Any update ? > Make savepoints atomic with respect to state and side effects > - > > Key: FLINK-7883 > URL: https://issues.apache.org/jira/browse/FLINK-7883 > Project: Flink > Issue Type: Improvement > Components: API / DataStream, Connectors / Kafka, Runtime / State > Backends >Affects Versions: 1.3.2, 1.4.0 >Reporter: Antoine Philippot >Priority: Major > > For a cancel with savepoint command, the JobManager trigger the cancel call > once the savepoint is finished, but during the savepoint execution, kafka > source continue to poll new messages which will not be part of the savepoint > and will be replayed on the next application start. > A solution could be to stop fetching the source stream task before triggering > the savepoint. > I suggest to add an interface {{StoppableFetchingSourceFunction}} with a > method {{stopFetching}} that existant SourceFunction implementations could > implement. > We can add a {{stopFetchingSource}} property in > {{CheckpointOptions}} class to pass the desired behaviour from > {{JobManager.handleMessage(CancelJobWithSavepoint)}} to > {{SourceStreamTask.triggerCheckpoint}} -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (FLINK-7883) Make savepoints atomic with respect to state and side effects
[ https://issues.apache.org/jira/browse/FLINK-7883?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17096689#comment-17096689 ] Konstantin Knauf commented on FLINK-7883: - [~kkloudas] Should this ticket be closed by now? > Make savepoints atomic with respect to state and side effects > - > > Key: FLINK-7883 > URL: https://issues.apache.org/jira/browse/FLINK-7883 > Project: Flink > Issue Type: Improvement > Components: API / DataStream, Connectors / Kafka, Runtime / State > Backends >Affects Versions: 1.3.2, 1.4.0 >Reporter: Antoine Philippot >Priority: Major > > For a cancel with savepoint command, the JobManager trigger the cancel call > once the savepoint is finished, but during the savepoint execution, kafka > source continue to poll new messages which will not be part of the savepoint > and will be replayed on the next application start. > A solution could be to stop fetching the source stream task before triggering > the savepoint. > I suggest to add an interface {{StoppableFetchingSourceFunction}} with a > method {{stopFetching}} that existant SourceFunction implementations could > implement. > We can add a {{stopFetchingSource}} property in > {{CheckpointOptions}} class to pass the desired behaviour from > {{JobManager.handleMessage(CancelJobWithSavepoint)}} to > {{SourceStreamTask.triggerCheckpoint}} -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (FLINK-7883) Make savepoints atomic with respect to state and side effects
[ https://issues.apache.org/jira/browse/FLINK-7883?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16843189#comment-16843189 ] Vahid Hashemian commented on FLINK-7883: We are also very interested in this enhancement that lets bounded consumption by {{FlinkKafkaConsumer}}. > Make savepoints atomic with respect to state and side effects > - > > Key: FLINK-7883 > URL: https://issues.apache.org/jira/browse/FLINK-7883 > Project: Flink > Issue Type: Improvement > Components: API / DataStream, Connectors / Kafka, Runtime / State > Backends >Affects Versions: 1.3.2, 1.4.0 >Reporter: Antoine Philippot >Priority: Major > > For a cancel with savepoint command, the JobManager trigger the cancel call > once the savepoint is finished, but during the savepoint execution, kafka > source continue to poll new messages which will not be part of the savepoint > and will be replayed on the next application start. > A solution could be to stop fetching the source stream task before triggering > the savepoint. > I suggest to add an interface {{StoppableFetchingSourceFunction}} with a > method {{stopFetching}} that existant SourceFunction implementations could > implement. > We can add a {{stopFetchingSource}} property in > {{CheckpointOptions}} class to pass the desired behaviour from > {{JobManager.handleMessage(CancelJobWithSavepoint)}} to > {{SourceStreamTask.triggerCheckpoint}} -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (FLINK-7883) Make savepoints atomic with respect to state and side effects
[ https://issues.apache.org/jira/browse/FLINK-7883?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16697931#comment-16697931 ] Steven Zhen Wu commented on FLINK-7883: --- We would love to see this happening. it is the "graceful" shutdown that we need to reduce/minimize duplicates if we are going to enable aggressive/frequent rescale events. otherwise, we are going to see frequent and significant duplicates. > Make savepoints atomic with respect to state and side effects > - > > Key: FLINK-7883 > URL: https://issues.apache.org/jira/browse/FLINK-7883 > Project: Flink > Issue Type: Improvement > Components: DataStream API, Kafka Connector, State Backends, > Checkpointing >Affects Versions: 1.3.2, 1.4.0 >Reporter: Antoine Philippot >Priority: Major > > For a cancel with savepoint command, the JobManager trigger the cancel call > once the savepoint is finished, but during the savepoint execution, kafka > source continue to poll new messages which will not be part of the savepoint > and will be replayed on the next application start. > A solution could be to stop fetching the source stream task before triggering > the savepoint. > I suggest to add an interface {{StoppableFetchingSourceFunction}} with a > method {{stopFetching}} that existant SourceFunction implementations could > implement. > We can add a {{stopFetchingSource}} property in > {{CheckpointOptions}} class to pass the desired behaviour from > {{JobManager.handleMessage(CancelJobWithSavepoint)}} to > {{SourceStreamTask.triggerCheckpoint}} -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (FLINK-7883) Make savepoints atomic with respect to state and side effects
[ https://issues.apache.org/jira/browse/FLINK-7883?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16565552#comment-16565552 ] Antoine Philippot commented on FLINK-7883: -- Yes [~aljoscha], I'm agree, as long as the solution avoid duplicate reprocessing of events. My solution with StoppableFunction was the easiest way for me to achieve what I wanted ;) > Make savepoints atomic with respect to state and side effects > - > > Key: FLINK-7883 > URL: https://issues.apache.org/jira/browse/FLINK-7883 > Project: Flink > Issue Type: Improvement > Components: DataStream API, Kafka Connector, State Backends, > Checkpointing >Affects Versions: 1.3.2, 1.4.0 >Reporter: Antoine Philippot >Priority: Major > > For a cancel with savepoint command, the JobManager trigger the cancel call > once the savepoint is finished, but during the savepoint execution, kafka > source continue to poll new messages which will not be part of the savepoint > and will be replayed on the next application start. > A solution could be to stop fetching the source stream task before triggering > the savepoint. > I suggest to add an interface {{StoppableFetchingSourceFunction}} with a > method {{stopFetching}} that existant SourceFunction implementations could > implement. > We can add a {{stopFetchingSource}} property in > {{CheckpointOptions}} class to pass the desired behaviour from > {{JobManager.handleMessage(CancelJobWithSavepoint)}} to > {{SourceStreamTask.triggerCheckpoint}} -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (FLINK-7883) Make savepoints atomic with respect to state and side effects
[ https://issues.apache.org/jira/browse/FLINK-7883?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16565446#comment-16565446 ] Aljoscha Krettek commented on FLINK-7883: - For a start, I renamed the issue. I hope it captures what we want to achive. I don't yet know to solve it but I think quiescing the pipeline and then taking the savepoint might be an option. Regarding {{StoppableFunction}}: I think adding this interface was misguided and we should not try and use it for more stuff. The idea of introducing a new source interface has been floated several times on the ML and in various issues. I think such an interface would/should not need the distinction between cancel and stop if it allowed the system (i.e. Flink) to determine when to get new data from the source instead of the source owning the pull loop. WDYT? > Make savepoints atomic with respect to state and side effects > - > > Key: FLINK-7883 > URL: https://issues.apache.org/jira/browse/FLINK-7883 > Project: Flink > Issue Type: Improvement > Components: DataStream API, Kafka Connector, State Backends, > Checkpointing >Affects Versions: 1.3.2, 1.4.0 >Reporter: Antoine Philippot >Priority: Major > > For a cancel with savepoint command, the JobManager trigger the cancel call > once the savepoint is finished, but during the savepoint execution, kafka > source continue to poll new messages which will not be part of the savepoint > and will be replayed on the next application start. > A solution could be to stop fetching the source stream task before triggering > the savepoint. > I suggest to add an interface {{StoppableFetchingSourceFunction}} with a > method {{stopFetching}} that existant SourceFunction implementations could > implement. > We can add a {{stopFetchingSource}} property in > {{CheckpointOptions}} class to pass the desired behaviour from > {{JobManager.handleMessage(CancelJobWithSavepoint)}} to > {{SourceStreamTask.triggerCheckpoint}} -- This message was sent by Atlassian JIRA (v7.6.3#76005)