qpid-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF GitHub Bot (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (QPIDJMS-315) Add support for Netty KQueue transport
Date Fri, 25 Aug 2017 16:39:00 GMT

    [ https://issues.apache.org/jira/browse/QPIDJMS-315?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16141844#comment-16141844
] 

ASF GitHub Bot commented on QPIDJMS-315:
----------------------------------------

Github user michaelandrepearce commented on the issue:

    https://github.com/apache/qpid-jms/pull/12
  
    Hi @gemmellr 
    
    > I don't think the various environment etc checking being done really belongs in the
client, it seems like that stuff belongs in Netty if the isAvailable() checks it offers aren't
actually sufficient. 
    
    Agreed its not ideal we have to do this, but this was found on epoll in artemis (which
you already have epoll implemented and defaulted to true) , as such until netty improve the
check then this check is worth while for the interim unless you default epoll to false currently
the client has the same issue with epoll as found  - https://issues.apache.org/jira/browse/ARTEMIS-1099
    
    
    > Similarly it feels a little odd making it check for being on a Mac when KQueue is
really for BSD rather than OSX specific, and they might open up support for others later,
again it feels like Netty is better placed to be checking that type of thing.
    
    The classifier for the dependency for kqueue implementation is for osx build, there isn't
one currently for freeBSD or other derivaitves (osx being bsd based), as such the check for
the platform being osx, to be safe. E.g. our case is we do have java clients that are on macOS.
    
    >  also think that if added the KQueue bit would be better set off by default for now
and flipped later on. It's very new in Netty itself, whereas the epoll stuff had time to mature
before it was added. Linux is also probably the primary platform we test and use the client
on, whereas we don't really test on OS X at all (though I shall change that shortly), so sticking
to the NIO bits by default there seems wiser for now.
    
    Sure this make sense we can make it default to false.
    



> Add support for Netty KQueue transport
> --------------------------------------
>
>                 Key: QPIDJMS-315
>                 URL: https://issues.apache.org/jira/browse/QPIDJMS-315
>             Project: Qpid JMS
>          Issue Type: New Feature
>    Affects Versions: 0.24.0
>            Reporter: Michael Andre Pearce
>             Fix For: 0.25.0
>
>
> Add support for using the Netty KQueue transport when the platform supports it (x86-64
osx) and provide a transport configuration option to bypass it and return to using NIO. This
change also switches the transport to using the default Netty allocator on the transports
which by default is a pooled allocator.
> To disable kqueue use the URI can be updated to include option:
> transport.useKQueue=false



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@qpid.apache.org
For additional commands, e-mail: dev-help@qpid.apache.org


Mime
View raw message