[jira] [Commented] (FLINK-13698) Rework threading model of CheckpointCoordinator
[
https://issues.apache.org/jira/browse/FLINK-13698?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17333089#comment-17333089
]
Stephan Ewen commented on FLINK-13698:
--
I think it is not this ticket that we need, or at least not what is described
here.
We need to change the {{CompletedCheckpointStore}} to do asynchronous loading
of checkpoint metadata. That is an independent issue from this ticket.
> Rework threading model of CheckpointCoordinator
> ---
>
> Key: FLINK-13698
> URL: https://issues.apache.org/jira/browse/FLINK-13698
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Checkpointing
>Affects Versions: 1.10.0
>Reporter: Piotr Nowojski
>Priority: Critical
> Fix For: 1.14.0
>
>
> Currently {{CheckpointCoordinator}} and {{CheckpointFailureManager}} code is
> executed by multiple different threads (mostly {{ioExecutor}}, but not only).
> It's causing multiple concurrency issues, for example:
> https://issues.apache.org/jira/browse/FLINK-13497
> Proper fix would be to rethink threading model there. At first glance it
> doesn't seem that this code should be multi threaded, except of parts doing
> the actual IO operations, so it should be possible to run everything in one
> single ExecutionGraph's thread and just run asynchronously necessary IO
> operations with some feedback loop ("mailbox style").
> I would strongly recommend fixing this issue before adding new features in
> the \{{CheckpointCoordinator}} component.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
[jira] [Commented] (FLINK-13698) Rework threading model of CheckpointCoordinator
[
https://issues.apache.org/jira/browse/FLINK-13698?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17330080#comment-17330080
]
Till Rohrmann commented on FLINK-13698:
---
What is the state of this problem [~pnowojski]. I think we should raise the
priority of this ticket to critical because it causes cluster instabilities
which are quite serious. I think we should try to solve this problem eventually.
> Rework threading model of CheckpointCoordinator
> ---
>
> Key: FLINK-13698
> URL: https://issues.apache.org/jira/browse/FLINK-13698
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Checkpointing
>Affects Versions: 1.10.0
>Reporter: Piotr Nowojski
>Priority: Major
> Fix For: 1.13.0
>
>
> Currently {{CheckpointCoordinator}} and {{CheckpointFailureManager}} code is
> executed by multiple different threads (mostly {{ioExecutor}}, but not only).
> It's causing multiple concurrency issues, for example:
> https://issues.apache.org/jira/browse/FLINK-13497
> Proper fix would be to rethink threading model there. At first glance it
> doesn't seem that this code should be multi threaded, except of parts doing
> the actual IO operations, so it should be possible to run everything in one
> single ExecutionGraph's thread and just run asynchronously necessary IO
> operations with some feedback loop ("mailbox style").
> I would strongly recommend fixing this issue before adding new features in
> the \{{CheckpointCoordinator}} component.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
[jira] [Commented] (FLINK-13698) Rework threading model of CheckpointCoordinator
[
https://issues.apache.org/jira/browse/FLINK-13698?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17229069#comment-17229069
]
Roman Khachatryan commented on FLINK-13698:
---
>From what I know from a recent discussion with Piotr, it's not being included
>in 1.12. (he is currenlty on vacation).
> Rework threading model of CheckpointCoordinator
> ---
>
> Key: FLINK-13698
> URL: https://issues.apache.org/jira/browse/FLINK-13698
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Checkpointing
>Affects Versions: 1.10.0
>Reporter: Piotr Nowojski
>Assignee: Biao Liu
>Priority: Major
> Fix For: 1.12.0
>
>
> Currently {{CheckpointCoordinator}} and {{CheckpointFailureManager}} code is
> executed by multiple different threads (mostly {{ioExecutor}}, but not only).
> It's causing multiple concurrency issues, for example:
> https://issues.apache.org/jira/browse/FLINK-13497
> Proper fix would be to rethink threading model there. At first glance it
> doesn't seem that this code should be multi threaded, except of parts doing
> the actual IO operations, so it should be possible to run everything in one
> single ExecutionGraph's thread and just run asynchronously necessary IO
> operations with some feedback loop ("mailbox style").
> I would strongly recommend fixing this issue before adding new features in
> the \{{CheckpointCoordinator}} component.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
[jira] [Commented] (FLINK-13698) Rework threading model of CheckpointCoordinator
[
https://issues.apache.org/jira/browse/FLINK-13698?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17228487#comment-17228487
]
Stephan Ewen commented on FLINK-13698:
--
AS far as I know, there has been no progress here.
> Rework threading model of CheckpointCoordinator
> ---
>
> Key: FLINK-13698
> URL: https://issues.apache.org/jira/browse/FLINK-13698
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Checkpointing
>Affects Versions: 1.10.0
>Reporter: Piotr Nowojski
>Assignee: Biao Liu
>Priority: Major
> Fix For: 1.12.0
>
>
> Currently {{CheckpointCoordinator}} and {{CheckpointFailureManager}} code is
> executed by multiple different threads (mostly {{ioExecutor}}, but not only).
> It's causing multiple concurrency issues, for example:
> https://issues.apache.org/jira/browse/FLINK-13497
> Proper fix would be to rethink threading model there. At first glance it
> doesn't seem that this code should be multi threaded, except of parts doing
> the actual IO operations, so it should be possible to run everything in one
> single ExecutionGraph's thread and just run asynchronously necessary IO
> operations with some feedback loop ("mailbox style").
> I would strongly recommend fixing this issue before adding new features in
> the \{{CheckpointCoordinator}} component.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
[jira] [Commented] (FLINK-13698) Rework threading model of CheckpointCoordinator
[
https://issues.apache.org/jira/browse/FLINK-13698?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17228484#comment-17228484
]
Till Rohrmann commented on FLINK-13698:
---
What's the progress of this issue [~pnowojski]? Will this improvement be
included in Flink {{1.12.0}}?
> Rework threading model of CheckpointCoordinator
> ---
>
> Key: FLINK-13698
> URL: https://issues.apache.org/jira/browse/FLINK-13698
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Checkpointing
>Affects Versions: 1.10.0
>Reporter: Piotr Nowojski
>Assignee: Biao Liu
>Priority: Major
> Fix For: 1.12.0
>
>
> Currently {{CheckpointCoordinator}} and {{CheckpointFailureManager}} code is
> executed by multiple different threads (mostly {{ioExecutor}}, but not only).
> It's causing multiple concurrency issues, for example:
> https://issues.apache.org/jira/browse/FLINK-13497
> Proper fix would be to rethink threading model there. At first glance it
> doesn't seem that this code should be multi threaded, except of parts doing
> the actual IO operations, so it should be possible to run everything in one
> single ExecutionGraph's thread and just run asynchronously necessary IO
> operations with some feedback loop ("mailbox style").
> I would strongly recommend fixing this issue before adding new features in
> the \{{CheckpointCoordinator}} component.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
[jira] [Commented] (FLINK-13698) Rework threading model of CheckpointCoordinator
[
https://issues.apache.org/jira/browse/FLINK-13698?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16931119#comment-16931119
]
Piotr Nowojski commented on FLINK-13698:
Thanks for the update [~SleePy].
> Rework threading model of CheckpointCoordinator
> ---
>
> Key: FLINK-13698
> URL: https://issues.apache.org/jira/browse/FLINK-13698
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Checkpointing
>Affects Versions: 1.10.0
>Reporter: Piotr Nowojski
>Assignee: Biao Liu
>Priority: Critical
>
> Currently {{CheckpointCoordinator}} and {{CheckpointFailureManager}} code is
> executed by multiple different threads (mostly {{ioExecutor}}, but not only).
> It's causing multiple concurrency issues, for example:
> https://issues.apache.org/jira/browse/FLINK-13497
> Proper fix would be to rethink threading model there. At first glance it
> doesn't seem that this code should be multi threaded, except of parts doing
> the actual IO operations, so it should be possible to run everything in one
> single ExecutionGraph's thread and just run asynchronously necessary IO
> operations with some feedback loop ("mailbox style").
> I would strongly recommend fixing this issue before adding new features in
> the \{{CheckpointCoordinator}} component.
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
[jira] [Commented] (FLINK-13698) Rework threading model of CheckpointCoordinator
[
https://issues.apache.org/jira/browse/FLINK-13698?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16924472#comment-16924472
]
Biao Liu commented on FLINK-13698:
--
Just a progress updating, now I'm doing some POC developments. I found it's a
bit hard to separate the issue and work into subtasks. Because the changes of
subtasks are tightly coupled. So I stopped creating new subtasks to avoid
adjusting them during the further developing.
> Rework threading model of CheckpointCoordinator
> ---
>
> Key: FLINK-13698
> URL: https://issues.apache.org/jira/browse/FLINK-13698
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Checkpointing
>Affects Versions: 1.10.0
>Reporter: Piotr Nowojski
>Assignee: Biao Liu
>Priority: Critical
>
> Currently {{CheckpointCoordinator}} and {{CheckpointFailureManager}} code is
> executed by multiple different threads (mostly {{ioExecutor}}, but not only).
> It's causing multiple concurrency issues, for example:
> https://issues.apache.org/jira/browse/FLINK-13497
> Proper fix would be to rethink threading model there. At first glance it
> doesn't seem that this code should be multi threaded, except of parts doing
> the actual IO operations, so it should be possible to run everything in one
> single ExecutionGraph's thread and just run asynchronously necessary IO
> operations with some feedback loop ("mailbox style").
> I would strongly recommend fixing this issue before adding new features in
> the \{{CheckpointCoordinator}} component.
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
[jira] [Commented] (FLINK-13698) Rework threading model of CheckpointCoordinator
[
https://issues.apache.org/jira/browse/FLINK-13698?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16915430#comment-16915430
]
Biao Liu commented on FLINK-13698:
--
Sorry for a bit late response. I was trapped in some other things last week :(
Thanks for detailed reviewing [~pnowojski], [~till.rohrmann]. It helps a lot.
And I really appreciate that.
I will separate this proposal into series of subtasks.
> Rework threading model of CheckpointCoordinator
> ---
>
> Key: FLINK-13698
> URL: https://issues.apache.org/jira/browse/FLINK-13698
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Checkpointing
>Affects Versions: 1.10.0
>Reporter: Piotr Nowojski
>Assignee: Biao Liu
>Priority: Critical
>
> Currently {{CheckpointCoordinator}} and {{CheckpointFailureManager}} code is
> executed by multiple different threads (mostly {{ioExecutor}}, but not only).
> It's causing multiple concurrency issues, for example:
> https://issues.apache.org/jira/browse/FLINK-13497
> Proper fix would be to rethink threading model there. At first glance it
> doesn't seem that this code should be multi threaded, except of parts doing
> the actual IO operations, so it should be possible to run everything in one
> single ExecutionGraph's thread and just run asynchronously necessary IO
> operations with some feedback loop ("mailbox style").
> I would strongly recommend fixing this issue before adding new features in
> the \{{CheckpointCoordinator}} component.
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
[jira] [Commented] (FLINK-13698) Rework threading model of CheckpointCoordinator
[
https://issues.apache.org/jira/browse/FLINK-13698?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16906075#comment-16906075
]
Biao Liu commented on FLINK-13698:
--
Thanks for your attention [~StephanEwen]!
Piotr and me had a short discussion last week.
I believe my proposal matches your suggestions well. :)
> Rework threading model of CheckpointCoordinator
> ---
>
> Key: FLINK-13698
> URL: https://issues.apache.org/jira/browse/FLINK-13698
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Checkpointing
>Affects Versions: 1.10.0
>Reporter: Piotr Nowojski
>Assignee: Biao Liu
>Priority: Critical
>
> Currently {{CheckpointCoordinator}} and {{CheckpointFailureManager}} code is
> executed by multiple different threads (mostly {{ioExecutor}}, but not only).
> It's causing multiple concurrency issues, for example:
> https://issues.apache.org/jira/browse/FLINK-13497
> Proper fix would be to rethink threading model there. At first glance it
> doesn't seem that this code should be multi threaded, except of parts doing
> the actual IO operations, so it should be possible to run everything in one
> single ExecutionGraph's thread and just run asynchronously necessary IO
> operations with some feedback loop ("mailbox style").
> I would strongly recommend fixing this issue before adding new features in
> the \{{CheckpointCoordinator}} component.
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
[jira] [Commented] (FLINK-13698) Rework threading model of CheckpointCoordinator
[
https://issues.apache.org/jira/browse/FLINK-13698?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16906060#comment-16906060
]
Stephan Ewen commented on FLINK-13698:
--
I would suggest to have the {{CheckpointCoordinator}} and the
{{Scheduler}}/{{ExecutionGraph}} in the same thread.
That makes a lot of things much easier.
- The checkpoint timer logic should be factored out into the periodic
checkpoint trigger.
- The actual call
- All cleanup and writing of state should be delegated to the I/O executor.
-> mainly state release as a "fire and forget" execution in the I/O executor
-> writing out master-hook data and checkpoint metadata needs to happen in
the I/O executor, with a Future whose completion is again handles in the main
thread executor.
> Rework threading model of CheckpointCoordinator
> ---
>
> Key: FLINK-13698
> URL: https://issues.apache.org/jira/browse/FLINK-13698
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Checkpointing
>Affects Versions: 1.10.0
>Reporter: Piotr Nowojski
>Assignee: Biao Liu
>Priority: Critical
>
> Currently {{CheckpointCoordinator}} and {{CheckpointFailureManager}} code is
> executed by multiple different threads (mostly {{ioExecutor}}, but not only).
> It's causing multiple concurrency issues, for example:
> https://issues.apache.org/jira/browse/FLINK-13497
> Proper fix would be to rethink threading model there. At first glance it
> doesn't seem that this code should be multi threaded, except of parts doing
> the actual IO operations, so it should be possible to run everything in one
> single ExecutionGraph's thread and just run asynchronously necessary IO
> operations with some feedback loop ("mailbox style").
> I would strongly recommend fixing this issue before adding new features in
> the \{{CheckpointCoordinator}} component.
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
[jira] [Commented] (FLINK-13698) Rework threading model of CheckpointCoordinator
[
https://issues.apache.org/jira/browse/FLINK-13698?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16905943#comment-16905943
]
Biao Liu commented on FLINK-13698:
--
Hi [~pnowojski], I have attached a design document. That would be great if you
have time to review it.
> Rework threading model of CheckpointCoordinator
> ---
>
> Key: FLINK-13698
> URL: https://issues.apache.org/jira/browse/FLINK-13698
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Checkpointing
>Affects Versions: 1.10.0
>Reporter: Piotr Nowojski
>Assignee: Biao Liu
>Priority: Critical
>
> Currently {{CheckpointCoordinator}} and {{CheckpointFailureManager}} code is
> executed by multiple different threads (mostly {{ioExecutor}}, but not only).
> It's causing multiple concurrency issues, for example:
> https://issues.apache.org/jira/browse/FLINK-13497
> Proper fix would be to rethink threading model there. At first glance it
> doesn't seem that this code should be multi threaded, except of parts doing
> the actual IO operations, so it should be possible to run everything in one
> single ExecutionGraph's thread and just run asynchronously necessary IO
> operations with some feedback loop ("mailbox style").
> I would strongly recommend fixing this issue before adding new features in
> the \{{CheckpointCoordinator}} component.
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
