[jira] [Commented] (SPARK-16963) Change Source API so that sources do not need to keep unbounded state
[
https://issues.apache.org/jira/browse/SPARK-16963?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15612609#comment-15612609
]
Apache Spark commented on SPARK-16963:
--
User 'zsxwing' has created a pull request for this issue:
https://github.com/apache/spark/pull/15661
> Change Source API so that sources do not need to keep unbounded state
> -
>
> Key: SPARK-16963
> URL: https://issues.apache.org/jira/browse/SPARK-16963
> Project: Spark
> Issue Type: Sub-task
> Components: Streaming
>Affects Versions: 2.0.0, 2.0.1
>Reporter: Frederick Reiss
>Assignee: Frederick Reiss
> Fix For: 2.0.3, 2.1.0
>
>
> The version of the Source API in Spark 2.0.0 defines a single getBatch()
> method for fetching records from the source, with the following Scaladoc
> comments defining the semantics:
> {noformat}
> /**
> * Returns the data that is between the offsets (`start`, `end`]. When
> `start` is `None` then
> * the batch should begin with the first available record. This method must
> always return the
> * same data for a particular `start` and `end` pair.
> */
> def getBatch(start: Option[Offset], end: Offset): DataFrame
> {noformat}
> These semantics mean that a Source must retain all past history for the
> stream that it backs. Further, a Source is also required to retain this data
> across restarts of the process where the Source is instantiated, even when
> the Source is restarted on a different machine.
> These restrictions make it difficult to implement the Source API, as any
> implementation requires potentially unbounded amounts of distributed storage.
> See the mailing list thread at
> [http://apache-spark-developers-list.1001551.n3.nabble.com/Source-API-requires-unbounded-distributed-storage-td18551.html]
> for more information.
> This JIRA will cover augmenting the Source API with an additional callback
> that will allow Structured Streaming scheduler to notify the source when it
> is safe to discard buffered data.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-16963) Change Source API so that sources do not need to keep unbounded state
[
https://issues.apache.org/jira/browse/SPARK-16963?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15440598#comment-15440598
]
Frederick Reiss commented on SPARK-16963:
-
Updated the pull request to address some conflicting changes in the main branch
and to address some minor review comments. Changed the name of `getMinOffset`
to `lastCommittedOffset` per Prashant's comments. Changes are still ready for
review.
> Change Source API so that sources do not need to keep unbounded state
> -
>
> Key: SPARK-16963
> URL: https://issues.apache.org/jira/browse/SPARK-16963
> Project: Spark
> Issue Type: Improvement
> Components: Streaming
>Affects Versions: 2.0.0
>Reporter: Frederick Reiss
>
> The version of the Source API in Spark 2.0.0 defines a single getBatch()
> method for fetching records from the source, with the following Scaladoc
> comments defining the semantics:
> {noformat}
> /**
> * Returns the data that is between the offsets (`start`, `end`]. When
> `start` is `None` then
> * the batch should begin with the first available record. This method must
> always return the
> * same data for a particular `start` and `end` pair.
> */
> def getBatch(start: Option[Offset], end: Offset): DataFrame
> {noformat}
> These semantics mean that a Source must retain all past history for the
> stream that it backs. Further, a Source is also required to retain this data
> across restarts of the process where the Source is instantiated, even when
> the Source is restarted on a different machine.
> These restrictions make it difficult to implement the Source API, as any
> implementation requires potentially unbounded amounts of distributed storage.
> See the mailing list thread at
> [http://apache-spark-developers-list.1001551.n3.nabble.com/Source-API-requires-unbounded-distributed-storage-td18551.html]
> for more information.
> This JIRA will cover augmenting the Source API with an additional callback
> that will allow Structured Streaming scheduler to notify the source when it
> is safe to discard buffered data.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-16963) Change Source API so that sources do not need to keep unbounded state
[
https://issues.apache.org/jira/browse/SPARK-16963?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15434767#comment-15434767
]
Prashant Sharma commented on SPARK-16963:
-
Thank you Fred, the API, looks sufficient for the purpose of MQTT connector.
May be, getMinOffset can be renamed as `lastCommittedOffset`.
> Change Source API so that sources do not need to keep unbounded state
> -
>
> Key: SPARK-16963
> URL: https://issues.apache.org/jira/browse/SPARK-16963
> Project: Spark
> Issue Type: Improvement
> Components: Streaming
>Affects Versions: 2.0.0
>Reporter: Frederick Reiss
>
> The version of the Source API in Spark 2.0.0 defines a single getBatch()
> method for fetching records from the source, with the following Scaladoc
> comments defining the semantics:
> {noformat}
> /**
> * Returns the data that is between the offsets (`start`, `end`]. When
> `start` is `None` then
> * the batch should begin with the first available record. This method must
> always return the
> * same data for a particular `start` and `end` pair.
> */
> def getBatch(start: Option[Offset], end: Offset): DataFrame
> {noformat}
> These semantics mean that a Source must retain all past history for the
> stream that it backs. Further, a Source is also required to retain this data
> across restarts of the process where the Source is instantiated, even when
> the Source is restarted on a different machine.
> These restrictions make it difficult to implement the Source API, as any
> implementation requires potentially unbounded amounts of distributed storage.
> See the mailing list thread at
> [http://apache-spark-developers-list.1001551.n3.nabble.com/Source-API-requires-unbounded-distributed-storage-td18551.html]
> for more information.
> This JIRA will cover augmenting the Source API with an additional callback
> that will allow Structured Streaming scheduler to notify the source when it
> is safe to discard buffered data.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-16963) Change Source API so that sources do not need to keep unbounded state
[
https://issues.apache.org/jira/browse/SPARK-16963?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15431779#comment-15431779
]
Frederick Reiss commented on SPARK-16963:
-
The proposed changes in the attached PR are now ready for review. [~marmbrus]
can you please have a look at your convenience? [~prashant_] can you also
please have a look with a particular focus on whether the changes fit with the
MQTT connector?
> Change Source API so that sources do not need to keep unbounded state
> -
>
> Key: SPARK-16963
> URL: https://issues.apache.org/jira/browse/SPARK-16963
> Project: Spark
> Issue Type: Improvement
> Components: Streaming
>Affects Versions: 2.0.0
>Reporter: Frederick Reiss
>
> The version of the Source API in Spark 2.0.0 defines a single getBatch()
> method for fetching records from the source, with the following Scaladoc
> comments defining the semantics:
> {noformat}
> /**
> * Returns the data that is between the offsets (`start`, `end`]. When
> `start` is `None` then
> * the batch should begin with the first available record. This method must
> always return the
> * same data for a particular `start` and `end` pair.
> */
> def getBatch(start: Option[Offset], end: Offset): DataFrame
> {noformat}
> These semantics mean that a Source must retain all past history for the
> stream that it backs. Further, a Source is also required to retain this data
> across restarts of the process where the Source is instantiated, even when
> the Source is restarted on a different machine.
> These restrictions make it difficult to implement the Source API, as any
> implementation requires potentially unbounded amounts of distributed storage.
> See the mailing list thread at
> [http://apache-spark-developers-list.1001551.n3.nabble.com/Source-API-requires-unbounded-distributed-storage-td18551.html]
> for more information.
> This JIRA will cover augmenting the Source API with an additional callback
> that will allow Structured Streaming scheduler to notify the source when it
> is safe to discard buffered data.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-16963) Change Source API so that sources do not need to keep unbounded state
[
https://issues.apache.org/jira/browse/SPARK-16963?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15412840#comment-15412840
]
Apache Spark commented on SPARK-16963:
--
User 'frreiss' has created a pull request for this issue:
https://github.com/apache/spark/pull/14553
> Change Source API so that sources do not need to keep unbounded state
> -
>
> Key: SPARK-16963
> URL: https://issues.apache.org/jira/browse/SPARK-16963
> Project: Spark
> Issue Type: Improvement
> Components: Streaming
>Affects Versions: 2.0.0
>Reporter: Frederick Reiss
>
> The version of the Source API in Spark 2.0.0 defines a single getBatch()
> method for fetching records from the source, with the following Scaladoc
> comments defining the semantics:
> {noformat}
> /**
> * Returns the data that is between the offsets (`start`, `end`]. When
> `start` is `None` then
> * the batch should begin with the first available record. This method must
> always return the
> * same data for a particular `start` and `end` pair.
> */
> def getBatch(start: Option[Offset], end: Offset): DataFrame
> {noformat}
> These semantics mean that a Source must retain all past history for the
> stream that it backs. Further, a Source is also required to retain this data
> across restarts of the process where the Source is instantiated, even when
> the Source is restarted on a different machine.
> These restrictions make it difficult to implement the Source API, as any
> implementation requires potentially unbounded amounts of distributed storage.
> See the mailing list thread at
> [http://apache-spark-developers-list.1001551.n3.nabble.com/Source-API-requires-unbounded-distributed-storage-td18551.html]
> for more information.
> This JIRA will cover augmenting the Source API with an additional callback
> that will allow Structured Streaming scheduler to notify the source when it
> is safe to discard buffered data.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
