Skip to content

Conversation

@vanzin
Copy link
Contributor

@vanzin vanzin commented Nov 16, 2015

This way the timeout handling code can correctly close "hung" channels that are
processing streams.

…anding requests.

This way the timeout handling code can correctly close "hung" channels that are
processing streams.
@SparkQA
Copy link

SparkQA commented Nov 17, 2015

Test build #46027 has finished for PR 9747 at commit 4df514c.

  • This patch passes all tests.
  • This patch merges cleanly.
  • This patch adds the following public classes (experimental):\n * class StreamingListener(object):\n

@vanzin
Copy link
Contributor Author

vanzin commented Nov 20, 2015

/cc @rxin @zsxwing

@zsxwing
Copy link
Member

zsxwing commented Nov 20, 2015

LGTM.

BTW, StreamInterceptor should catch exception from callback.onComplete, onData, onFailure. Otherwise, it may screw up the stream.

@zsxwing
Copy link
Member

zsxwing commented Nov 20, 2015

BTW, StreamInterceptor should catch exception from callback.onComplete, onData, onFailure. Otherwise, it may screw up the stream.

@vanzin could you fix this in a separate PR?

@vanzin
Copy link
Contributor Author

vanzin commented Nov 20, 2015

BTW, StreamInterceptor should catch exception from callback.onComplete, onData, onFailure.

@vanzin vanzin closed this Nov 20, 2015
@vanzin vanzin reopened this Nov 20, 2015
@vanzin
Copy link
Contributor Author

vanzin commented Nov 20, 2015

Oops. Hit the wrong thing.

@vanzin
Copy link
Contributor Author

vanzin commented Nov 20, 2015

BTW, StreamInterceptor should catch exception from callback.onComplete, onData, onFailure. Otherwise, it may screw up the stream.

It's not as simple as catching the exceptions. StreamInterceptor would have to switch to a mode where it ignores all future data until the stream is over, if the channel is to be kept open.

Right now an exception will bubble up and end up handled by TransportChannelHandler, which will close the channel.

@zsxwing
Copy link
Member

zsxwing commented Nov 20, 2015

Right now an exception will bubble up and end up handled by TransportChannelHandler, which will close the channel.

Sounds acceptable since #9530 uses a separate channel to stream files.

@vanzin
Copy link
Contributor Author

vanzin commented Nov 23, 2015

Merging to master.

@asfgit asfgit closed this in 5231cd5 Nov 23, 2015
@vanzin vanzin deleted the SPARK-11762 branch November 23, 2015 19:41
asfgit pushed a commit that referenced this pull request Nov 25, 2015
…nding requests.

This way the timeout handling code can correctly close "hung" channels that are
processing streams.

Author: Marcelo Vanzin <[email protected]>

Closes #9747 from vanzin/SPARK-11762.

(cherry picked from commit 5231cd5)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants