rename gflag --rabbit_host

"should be possible" != "is possible | is supported" ... as today we only use / test / support rabbitmq and our documentation tell you that you have to use rabbitmq, i don't see a reason to change this.

That's correct. But I think in future there we'll be users trying to run at least the Apache ActiveMQ in combination with the OpenStack. And if'ts possible we have to change the name of the variable in a later release, because it's confusing to configure the ActiveMQ with parameters called rabbit_*. But if we change it in a later release there are a lot of users who have to change there configurations. Or we have to introduce a new parameter for each new supported AMQP system, leading to a lof of repetitions in the code. I think the name of parameters should generally be as independent as possible from the underlying used software.

"I think the name of parameters should generally be as independent as possible from the underlying used software."

Agreed. Implementation specifics should not be in the name of the configuration option, IMHO.

-jay

We need to first answer the question: are we dependent on RabbitMQ and
are just using carrot as the API? Or are we dependent on carrot and
just using RabbitMQ as the backend? If the latter, I agree our queue
options should mirror the options defined in the 'carrot.connection'
documentation.

We shouldn't use 'amqp_*' prefix though since not all carrot backends
are AMQP based.

-Eric

On Wed, Feb 09, 2011 at 03:21:05PM -0000, Christian Berendt wrote:
> Public bug reported:
>
> Because OpenStack uses the python library carrot to support the AMQP
> messaging it should also be possible to use an other queuing system than
> RabbitMQ.
>
> ---snip---
> Several AMQP message broker implementations exists, including RabbitMQ, ZeroMQ and Apache ActiveMQ. You’ll need to have one of these installed, personally we’ve been using RabbitMQ.
> ---snap---
>
> I find it a little bit confusing to use a flag "--rabbit_host" for
> defining the used queuing server and I would prefer something like "--
> amqp_host" or "--amqp_server". Feel free...
>
> ** Affects: nova
> Importance: Undecided
> Status: New
>
> --
> You received this bug notification because you are a member of Nova Bug
> Team, which is subscribed to OpenStack Compute (nova).
> https://bugs.launchpad.net/bugs/715831
>
> Title:
> rename gflag --rabbit_host
>
> Status in OpenStack Compute (Nova):
> New
>
> Bug description:
> Because OpenStack uses the python library carrot to support the AMQP
> messaging it should also be possible to use an other queuing system
> than RabbitMQ.
>
> ---snip---
> Several AMQP message broker implementations exists, including RabbitMQ, ZeroMQ and Apache ActiveMQ. You’ll need to have one of these installed, personally we’ve been using RabbitMQ.
> ---snap---
>
> I find it a little bit confusing to use a flag "--rabbit_host" for
> defining the used queuing server and I would prefer something like "--
> amqp_host" or "--amqp_server". Feel free...
>

Even if we are dependent on RabbitMQ at the moment, I think this is not the case in the next months. And if we're dependent on RabbitMQ we should think about how to remove this dependency, or not?

I thought carrot only supports AMQP. I think it's a good idea to use the options defined in carrot.connection.

I thought that at some point we were going to switch to Celery. Wouldn't this move solve this issue altogether?

> -----Original Message-----
> From: <email address hidden> [mailto:<email address hidden>] On Behalf Of
> Christian Berendt
> Sent: 09 February 2011 20:59
> To: Armando Migliaccio
> Subject: [Bug 715831] Re: rename gflag --rabbit_host
>
> Even if we are dependent on RabbitMQ at the moment, I think this is not
> the case in the next months. And if we're dependent on RabbitMQ we
> should think about how to remove this dependency, or not?
>
> I thought carrot only supports AMQP. I think it's a good idea to use the
> options defined in carrot.connection.
>
> --
> You received this bug notification because you are subscribed to
> OpenStack Compute (nova).
> https://bugs.launchpad.net/bugs/715831
>
> Title:
> rename gflag --rabbit_host
>
> Status in OpenStack Compute (Nova):
> New
>
> Bug description:
> Because OpenStack uses the python library carrot to support the AMQP
> messaging it should also be possible to use an other queuing system
> than RabbitMQ.
>
> ---snip---
> Several AMQP message broker implementations exists, including RabbitMQ,
> ZeroMQ and Apache ActiveMQ. You’ll need to have one of these installed,
> personally we’ve been using RabbitMQ.
> ---snap---
>
> I find it a little bit confusing to use a flag "--rabbit_host" for
> defining the used queuing server and I would prefer something like "--
> amqp_host" or "--amqp_server". Feel free...
>

On Thu, Feb 10, 2011 at 4:45 AM, Armando Migliaccio
<email address hidden> wrote:
> I thought that at some point we were going to switch to Celery. Wouldn't
> this move solve this issue altogether?

Yes. Since you'd configure messaging entirely outside of Nova.

That said, I agree with Thierry that this is Wishlist, though.

-jay

We now have qpid support: http://wiki.openstack.org/QpidSupport

It has added a --qpid_hostname= flag

I'd say the --rabbit_host flags are specific to the carrot and kombu rpc drivers, neither of those drivers support anything other than rabbit and we can close this bug

closing bug per last comment