[jira] [Commented] (FLINK-16867) Simplify default timeout configuration
[
https://issues.apache.org/jira/browse/FLINK-16867?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17190164#comment-17190164
]
Till Rohrmann commented on FLINK-16867:
---
Even with FLINK-16866 resolved, we still have the blocking operation where we
write the submitted JobGraph to the {{JobGraphStore}}. This can also take a
little bit of time depending on the external system (filesystem). Hence, it
still might make sense to have an increased {{web.timeout}} because the 10s
might still be too aggressive even after making the job submission less
blocking.
> Simplify default timeout configuration
> --
>
> Key: FLINK-16867
> URL: https://issues.apache.org/jira/browse/FLINK-16867
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Configuration, Runtime / Coordination
>Affects Versions: 1.9.2, 1.10.0, 1.11.0
>Reporter: Till Rohrmann
>Priority: Minor
> Labels: usability
> Fix For: 1.12.0
>
>
> At the moment, Flink has several timeout options:
> * {{akka.ask.timeout}}: Timeout for intra cluster RPCs (JM <\-> RM <\-> TE)
> * {{web.timeout}}: Timeout for RPCs between REST handlers and RM, JM, TE
> At the moment, these values are separately configured. This requires the user
> to know about both configuration options and that Flink has multiple timeout
> values.
> In order to simplify setups I would suggest that {{web.timeout}} defaults to
> {{akka.ask.timeout}}, if {{web.timeout}} has not been explicitly configured.
> This has the benefits that the user only need to know about a single timeout
> value which is applied cluster wide.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
[jira] [Commented] (FLINK-16867) Simplify default timeout configuration
[
https://issues.apache.org/jira/browse/FLINK-16867?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17164519#comment-17164519
]
Steven Zhen Wu commented on FLINK-16867:
I should clarify that web.timeout becomes irrelevant to us for job submission
via REST API. It seems that only client.timeout matters for that particular
code path of submitting job.
> Simplify default timeout configuration
> --
>
> Key: FLINK-16867
> URL: https://issues.apache.org/jira/browse/FLINK-16867
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Configuration, Runtime / Coordination
>Affects Versions: 1.9.2, 1.10.0, 1.11.0
>Reporter: Till Rohrmann
>Priority: Minor
> Fix For: 1.12.0
>
>
> At the moment, Flink has several timeout options:
> * {{akka.ask.timeout}}: Timeout for intra cluster RPCs (JM <\-> RM <\-> TE)
> * {{web.timeout}}: Timeout for RPCs between REST handlers and RM, JM, TE
> At the moment, these values are separately configured. This requires the user
> to know about both configuration options and that Flink has multiple timeout
> values.
> In order to simplify setups I would suggest that {{web.timeout}} defaults to
> {{akka.ask.timeout}}, if {{web.timeout}} has not been explicitly configured.
> This has the benefits that the user only need to know about a single timeout
> value which is applied cluster wide.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
[jira] [Commented] (FLINK-16867) Simplify default timeout configuration
[
https://issues.apache.org/jira/browse/FLINK-16867?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17164206#comment-17164206
]
Till Rohrmann commented on FLINK-16867:
---
Thanks for the update. I think the {{web.timeout}} is still being used when
submitting a job via the client. There is the {{JobSubmitHandler}} which should
use it. Nevertheless, this is something we should pull straight asap.
> Simplify default timeout configuration
> --
>
> Key: FLINK-16867
> URL: https://issues.apache.org/jira/browse/FLINK-16867
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Configuration, Runtime / Coordination
>Affects Versions: 1.9.2, 1.10.0, 1.11.0
>Reporter: Till Rohrmann
>Priority: Minor
> Fix For: 1.12.0
>
>
> At the moment, Flink has several timeout options:
> * {{akka.ask.timeout}}: Timeout for intra cluster RPCs (JM <\-> RM <\-> TE)
> * {{web.timeout}}: Timeout for RPCs between REST handlers and RM, JM, TE
> At the moment, these values are separately configured. This requires the user
> to know about both configuration options and that Flink has multiple timeout
> values.
> In order to simplify setups I would suggest that {{web.timeout}} defaults to
> {{akka.ask.timeout}}, if {{web.timeout}} has not been explicitly configured.
> This has the benefits that the user only need to know about a single timeout
> value which is applied cluster wide.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
[jira] [Commented] (FLINK-16867) Simplify default timeout configuration
[
https://issues.apache.org/jira/browse/FLINK-16867?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17163667#comment-17163667
]
Steven Zhen Wu commented on FLINK-16867:
[~trohrmann] just to follow up on your comment on the other jira:
https://issues.apache.org/jira/browse/FLINK-11143?focusedCommentId=17161779=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-17161779
Now with 1.11, there is a 3rd `client.timeout` and all three have different
defaults.
* web.timeout is 600,000 milli-seconds
* akka.ask.timeout is `10 s`
* client.timeout is `1 min`
With 1.11, web.timeout becomes irrelevant for REST API of job submission. only
`client.timeout` matters.
> Simplify default timeout configuration
> --
>
> Key: FLINK-16867
> URL: https://issues.apache.org/jira/browse/FLINK-16867
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Configuration, Runtime / Coordination
>Affects Versions: 1.9.2, 1.10.0, 1.11.0
>Reporter: Till Rohrmann
>Priority: Minor
> Fix For: 1.12.0
>
>
> At the moment, Flink has several timeout options:
> * {{akka.ask.timeout}}: Timeout for intra cluster RPCs (JM <\-> RM <\-> TE)
> * {{web.timeout}}: Timeout for RPCs between REST handlers and RM, JM, TE
> At the moment, these values are separately configured. This requires the user
> to know about both configuration options and that Flink has multiple timeout
> values.
> In order to simplify setups I would suggest that {{web.timeout}} defaults to
> {{akka.ask.timeout}}, if {{web.timeout}} has not been explicitly configured.
> This has the benefits that the user only need to know about a single timeout
> value which is applied cluster wide.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
[jira] [Commented] (FLINK-16867) Simplify default timeout configuration
[
https://issues.apache.org/jira/browse/FLINK-16867?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17162747#comment-17162747
]
Till Rohrmann commented on FLINK-16867:
---
I think you are right [~rmetzger]. We can only do this once we have resolved
FLINK-16866.
> Simplify default timeout configuration
> --
>
> Key: FLINK-16867
> URL: https://issues.apache.org/jira/browse/FLINK-16867
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Configuration, Runtime / Coordination
>Affects Versions: 1.9.2, 1.10.0, 1.11.0
>Reporter: Till Rohrmann
>Priority: Minor
> Fix For: 1.12.0
>
>
> At the moment, Flink has several timeout options:
> * {{akka.ask.timeout}}: Timeout for intra cluster RPCs (JM <\-> RM <\-> TE)
> * {{web.timeout}}: Timeout for RPCs between REST handlers and RM, JM, TE
> At the moment, these values are separately configured. This requires the user
> to know about both configuration options and that Flink has multiple timeout
> values.
> In order to simplify setups I would suggest that {{web.timeout}} defaults to
> {{akka.ask.timeout}}, if {{web.timeout}} has not been explicitly configured.
> This has the benefits that the user only need to know about a single timeout
> value which is applied cluster wide.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
[jira] [Commented] (FLINK-16867) Simplify default timeout configuration
[
https://issues.apache.org/jira/browse/FLINK-16867?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17150883#comment-17150883
]
Robert Metzger commented on FLINK-16867:
I generally agree that this is a good improvement.
The problem is that the default value of "web.timeout" is 10 minutes, while
akka.ask.timeout defaults to 10 seconds.
I'm not sure if increasing the akka ask timeout is feasible, as it could lead
to a bad user experience (submitting a job when we can not connect to the JM)
> Simplify default timeout configuration
> --
>
> Key: FLINK-16867
> URL: https://issues.apache.org/jira/browse/FLINK-16867
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Configuration, Runtime / Coordination
>Affects Versions: 1.9.2, 1.10.0, 1.11.0
>Reporter: Till Rohrmann
>Priority: Minor
> Fix For: 1.11.0
>
>
> At the moment, Flink has several timeout options:
> * {{akka.ask.timeout}}: Timeout for intra cluster RPCs (JM <-> RM <-> TE)
> * {{web.timeout}}: Timeout for RPCs between REST handlers and RM, JM, TE
> At the moment, these values are separately configured. This requires the user
> to know about both configuration options and that Flink has multiple timeout
> values.
> In order to simplify setups I would suggest that {{web.timeout}} defaults to
> {{akka.ask.timeout}}, if {{web.timeout}} has been explicitly configured. This
> has the benefits that the user only need to know about a single timeout value
> which is applied cluster wide.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
