Channels, Pipelines, and database connections

[Matt DeLuco]

Norman, thanks for the reply.

I had considered passing the pool to the ChannelHandler, but
reconsidered when I thought I might be passing it to multiple
ChannelHandlers, having to open and close a pooled connection multiple
times.

The connections are pooled, but I thought I might be better off just
getting one connection from the pool and passing it between
ChannelHandlers, returning the connection to the pool only when all
DB-related ChannelHandlers are done with it (or when the connection
closes.)

Do you still think I'm better off just passing the pool to
ChannelHandlers and potentially opening/closing db connections
multiple times per client connection to the netty server?

MD


On Thu, Oct 13, 2011 at 1:59 AM, Norman Maurer
<norman.maurer at googlemail.com> wrote:
> Hi there,
>
> comments inline...
>
> 2011/10/13 Matt DeLuco <list_usr at spacebox.net>:
>> Hello,
>>
>> I'm just getting started with my own netty server, though I've read through
>> a lot of the docs and examples.  I have some questions about the basics.
>>
>> Is a Channel and associated ChannelPipeline created for every connection
>> made to the server?
>
> Yes.
>
> Is a ChannelPipelineFactory created for each
>> connection, or just once when I pass an instance to
>> ServerBootstrap.setPipelineFactory()?
>
> The ChannelPipelineFactory is just created once and create the
> ChannelPipeline which then will be used with the connection (Channel)
>
>>
>> I'll be using netty to serve sql queries over http and using BoneCP to do
>> RDBMS connection pooling.  As it stands I've got a custom
>> ChannelPipelineFactory to which I pass an ExecutionHandler and a BoneCP
>> connection pool.  Each time getPipeline() is called I add the
>> ExecutionHandler to the pipeline and I get a connection from the database
>> connection pool which I intend to pass to a custom ChannelHandler.  The
>> connection is assigned from the pool to a variable in getPipeline(), the
>> variable then passed to a ChannelHandler.
>
> A few comments:
>
> * Be sure you add the "same" instance of the ExecutionHandler to each
> pipeline. Its written to be used as singleton.
> * I think hand over a db connection to every ChannelHandler is not
> really what you want todo. As this would mean you would need to have
> one db connection per channel (network connection). This is kind of
> against the purpose of
>  a connection pool ;) Just pass the ConnectionPool to  your custom
> ChannelHandler ( I think you want to subclass
> SimpleUpstreamChannelHandler and overwrite the messageReceived(...)
> method) and retrieve a connection from it when needed, after doing the
> work put it back.
>
>
>
>>
>> With my current approach I can't figure out where I would close the database
>> connection to return it to the pool.
>
> See my comments above...
>
>>
>> Would it make more sense to subclass Channel or ChannelPipeline to contain a
>> reference to a database connection?  I could then have a ChannelFuture or
>> some such close the database connection when the Channel closes?
>
> See my comments above...
>
>>
>> Thanks.
>>
>>      -- Matt
>
>
> Bye,
> Norman
> _______________________________________________
> netty-users mailing list
> netty-users at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/netty-users
>