[
https://issues.apache.org/jira/browse/DIRMINA-639?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Emmanuel Lecharny updated DIRMINA-639:
--------------------------------------
Description:
While expecting the writeFuture to be updated when the write has been done,
it's done only when we get out of the chain. This is a major problem as you
can't rely on this to bail a connection based on a slow client. Typically, we
may stack thousands of message into the writeQueuebuffer, as the flush is only
done when we have gone though the whole chain, and back.
The only way to get the data be blushed immediately is to add an executorFilter
on the WRITE eventType, in order to create a new thread to handle this flush,
otherwise we have to wait for the current processor to be done with the chain
processing.
was:While expecting the writeFuture to be updated when the write has been
done, it's never done. This is a major problem as you can't rely on this to
bail a connection based on a slow client.
Summary: WriteFuture are updated long after a session.write() is done
(was: WriteFuture are never updated after a session.write())
> WriteFuture are updated long after a session.write() is done
> ------------------------------------------------------------
>
> Key: DIRMINA-639
> URL: https://issues.apache.org/jira/browse/DIRMINA-639
> Project: MINA
> Issue Type: Bug
> Affects Versions: 2.0.0-M3
> Reporter: Emmanuel Lecharny
> Priority: Blocker
> Fix For: 2.0.0-M4
>
>
> While expecting the writeFuture to be updated when the write has been done,
> it's done only when we get out of the chain. This is a major problem as you
> can't rely on this to bail a connection based on a slow client. Typically, we
> may stack thousands of message into the writeQueuebuffer, as the flush is
> only done when we have gone though the whole chain, and back.
> The only way to get the data be blushed immediately is to add an
> executorFilter on the WRITE eventType, in order to create a new thread to
> handle this flush, otherwise we have to wait for the current processor to be
> done with the chain processing.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
