Quantcast

Reliability and AMQP

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Reliability and AMQP

Goel, Tanmay

Hi,

 

So, I’m still trying to understand. There seems to be reliability missing in the protocol. The protocol supports acknowledgements from the consumer to the queue but there are no acks from the queue/broker back to the publisher in any of the schemes (store-forward, pub-sub, etc). I mean, for example, the publisher sends a series of 10K messages to the broker and one message gets dropped/lost in the network before reaching the broker. The publisher will not get an ack and the broker will never know the message was supposed to come. Will problems of this sort be taken care of in the later versions of the spec or am I missing reading something?

 

Are you guys working with the 0.9 version of the spec or do you have a copy of the 0.10 release since I couldn’t find it online?

 

Thanks.

 

Regards,

Tanmay


_______________________________________________
rabbitmq-discuss mailing list
[hidden email]
http://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Reliability and AMQP

Alexis Richardson-2
Tanmay

The kind of reliability that you refer to will appear in AMQP 0-10.
We have been working on this or over six months, along with others
from the AMQP Working Group.  It's an important future enhancement to
AMQP, and to RabbitMQ, and one that we are all excited about.

Existing users will be able to continue using 0-8 and 0-9 (which is
almost the same as 0-8), but for those who need 'acks', that will
become a built-in feature of the spec, and of RabbitMQ.

The 0-10 draft is not strictly speaking for general consumption, as
you might expect with a work in progress.  But we welcome commentary
and it is an 'open' spec, so do please let me know if you would like
to get involved.

alexis



On 8/21/07, Goel, Tanmay <[hidden email]> wrote:

>
>
>
>
> Hi,
>
>
>
> So, I'm still trying to understand. There seems to be reliability missing in
> the protocol. The protocol supports acknowledgements from the consumer to
> the queue but there are no acks from the queue/broker back to the publisher
> in any of the schemes (store-forward, pub-sub, etc). I mean, for example,
> the publisher sends a series of 10K messages to the broker and one message
> gets dropped/lost in the network before reaching the broker. The publisher
> will not get an ack and the broker will never know the message was supposed
> to come. Will problems of this sort be taken care of in the later versions
> of the spec or am I missing reading something?
>
>
>
> Are you guys working with the 0.9 version of the spec or do you have a copy
> of the 0.10 release since I couldn't find it online?
>
>
>
> Thanks.
>
>
>
> Regards,
>
> Tanmay
> _______________________________________________
> rabbitmq-discuss mailing list
> [hidden email]
> http://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
>
>


--
Alexis Richardson
+44 20 7617 7339 (UK)
+44 77 9865 2911 (cell)
+1 650 206 2517 (US)

_______________________________________________
rabbitmq-discuss mailing list
[hidden email]
http://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Reliability and AMQP

Goel, Tanmay
Hi Alexis, Matthias,

Thanks for that information. I understand that 0-10 is not completely
finished. But, could you point me to the changes/proposal for reliable
messaging? Also, if there is a bullet list to the changes that have been
made/proposed for the 0-10 release, it'd be great if you could send it
to me or tell me where I can find that too. I would like to review the
scope of changes to reliable messaging, among other things like
transaction processing. For example, will the acknowledgement back to
the publisher feature provided for all types of messaging, like P2P,
Pub-sub, etc.

Thanks.

Regards,
Tanmay



-----Original Message-----
From: [hidden email] [mailto:[hidden email]]
On Behalf Of Alexis Richardson
Sent: Tuesday, August 21, 2007 4:44 PM
To: Goel, Tanmay
Cc: Matthias Radestock; [hidden email]
Subject: Re: [rabbitmq-discuss] Reliability and AMQP

Tanmay

The kind of reliability that you refer to will appear in AMQP 0-10.
We have been working on this or over six months, along with others
from the AMQP Working Group.  It's an important future enhancement to
AMQP, and to RabbitMQ, and one that we are all excited about.

Existing users will be able to continue using 0-8 and 0-9 (which is
almost the same as 0-8), but for those who need 'acks', that will
become a built-in feature of the spec, and of RabbitMQ.

The 0-10 draft is not strictly speaking for general consumption, as
you might expect with a work in progress.  But we welcome commentary
and it is an 'open' spec, so do please let me know if you would like
to get involved.

alexis



On 8/21/07, Goel, Tanmay <[hidden email]> wrote:
>
>
>
>
> Hi,
>
>
>
> So, I'm still trying to understand. There seems to be reliability
missing in
> the protocol. The protocol supports acknowledgements from the consumer
to
> the queue but there are no acks from the queue/broker back to the
publisher
> in any of the schemes (store-forward, pub-sub, etc). I mean, for
example,
> the publisher sends a series of 10K messages to the broker and one
message
> gets dropped/lost in the network before reaching the broker. The
publisher
> will not get an ack and the broker will never know the message was
supposed
> to come. Will problems of this sort be taken care of in the later
versions
> of the spec or am I missing reading something?
>
>
>
> Are you guys working with the 0.9 version of the spec or do you have a
copy

> of the 0.10 release since I couldn't find it online?
>
>
>
> Thanks.
>
>
>
> Regards,
>
> Tanmay
> _______________________________________________
> rabbitmq-discuss mailing list
> [hidden email]
> http://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
>
>


--
Alexis Richardson
+44 20 7617 7339 (UK)
+44 77 9865 2911 (cell)
+1 650 206 2517 (US)

_______________________________________________
rabbitmq-discuss mailing list
[hidden email]
http://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Reliability and AMQP

Alexis Richardson-2
Tanmay

Acks will be provided for any publishing party with an appropriate
'session' (new in 0-10) with the AMQP broker.  All brokers and in
particular exchanges will be able to manage a session.

To answer your last question: both direct and topic based routing will
be available inside a session, and the broker will be able to ack p2p
type cases and various pub/sub type cases.  Thus, delegation from an
unreliable sender to a reliable client *will* work.

I.e.: "What goes in, must come out".

Messaging from broker to receiving parties will be analogous, but will
not in 0-10 deliver 'reliable multicast' as is normally understood in
the literature.  Look forward to multicast support being introduced
soon after.

I'll check with the WG on what I can show you from the current draft
of 0-10.  As I said, it's an open spec.

alexis





On 8/22/07, Goel, Tanmay <[hidden email]> wrote:

> Hi Alexis, Matthias,
>
> Thanks for that information. I understand that 0-10 is not completely
> finished. But, could you point me to the changes/proposal for reliable
> messaging? Also, if there is a bullet list to the changes that have been
> made/proposed for the 0-10 release, it'd be great if you could send it
> to me or tell me where I can find that too. I would like to review the
> scope of changes to reliable messaging, among other things like
> transaction processing. For example, will the acknowledgement back to
> the publisher feature provided for all types of messaging, like P2P,
> Pub-sub, etc.
>
> Thanks.
>
> Regards,
> Tanmay
>
>
>
> -----Original Message-----
> From: [hidden email] [mailto:[hidden email]]
> On Behalf Of Alexis Richardson
> Sent: Tuesday, August 21, 2007 4:44 PM
> To: Goel, Tanmay
> Cc: Matthias Radestock; [hidden email]
> Subject: Re: [rabbitmq-discuss] Reliability and AMQP
>
> Tanmay
>
> The kind of reliability that you refer to will appear in AMQP 0-10.
> We have been working on this or over six months, along with others
> from the AMQP Working Group.  It's an important future enhancement to
> AMQP, and to RabbitMQ, and one that we are all excited about.
>
> Existing users will be able to continue using 0-8 and 0-9 (which is
> almost the same as 0-8), but for those who need 'acks', that will
> become a built-in feature of the spec, and of RabbitMQ.
>
> The 0-10 draft is not strictly speaking for general consumption, as
> you might expect with a work in progress.  But we welcome commentary
> and it is an 'open' spec, so do please let me know if you would like
> to get involved.
>
> alexis
>
>
>
> On 8/21/07, Goel, Tanmay <[hidden email]> wrote:
> >
> >
> >
> >
> > Hi,
> >
> >
> >
> > So, I'm still trying to understand. There seems to be reliability
> missing in
> > the protocol. The protocol supports acknowledgements from the consumer
> to
> > the queue but there are no acks from the queue/broker back to the
> publisher
> > in any of the schemes (store-forward, pub-sub, etc). I mean, for
> example,
> > the publisher sends a series of 10K messages to the broker and one
> message
> > gets dropped/lost in the network before reaching the broker. The
> publisher
> > will not get an ack and the broker will never know the message was
> supposed
> > to come. Will problems of this sort be taken care of in the later
> versions
> > of the spec or am I missing reading something?
> >
> >
> >
> > Are you guys working with the 0.9 version of the spec or do you have a
> copy
> > of the 0.10 release since I couldn't find it online?
> >
> >
> >
> > Thanks.
> >
> >
> >
> > Regards,
> >
> > Tanmay
> > _______________________________________________
> > rabbitmq-discuss mailing list
> > [hidden email]
> > http://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
> >
> >
>
>
> --
> Alexis Richardson
> +44 20 7617 7339 (UK)
> +44 77 9865 2911 (cell)
> +1 650 206 2517 (US)
>


--
Alexis Richardson
+44 20 7617 7339 (UK)
+44 77 9865 2911 (cell)
+1 650 206 2517 (US)

_______________________________________________
rabbitmq-discuss mailing list
[hidden email]
http://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Reliability and AMQP

Alexis Richardson-2
Tanmay,

Here is the URL for the pre draft of 0-10, in subversion.  There is
also a JIRA and Wiki that I can help you with if needed.

https://svn.amqp.org/amqp/trunk/amqp_spec/amqp-spec-0-10-working.html

All,

This stuff is 'open' (eg: public, indexed by google) but please be
aware this is work in progress and changing quickly as we finalise it.
 I think the whole WG would appreciate it if people did not copy and
paste obviously draft text to casual passers by.  That said, as I have
stated, we *welcome* feedback and participation.  Talk to us.

alexis



On 8/22/07, Alexis Richardson <[hidden email]> wrote:

> Tanmay
>
> Acks will be provided for any publishing party with an appropriate
> 'session' (new in 0-10) with the AMQP broker.  All brokers and in
> particular exchanges will be able to manage a session.
>
> To answer your last question: both direct and topic based routing will
> be available inside a session, and the broker will be able to ack p2p
> type cases and various pub/sub type cases.  Thus, delegation from an
> unreliable sender to a reliable client *will* work.
>
> I.e.: "What goes in, must come out".
>
> Messaging from broker to receiving parties will be analogous, but will
> not in 0-10 deliver 'reliable multicast' as is normally understood in
> the literature.  Look forward to multicast support being introduced
> soon after.
>
> I'll check with the WG on what I can show you from the current draft
> of 0-10.  As I said, it's an open spec.
>
> alexis
>
>
>
>
>
> On 8/22/07, Goel, Tanmay <[hidden email]> wrote:
> > Hi Alexis, Matthias,
> >
> > Thanks for that information. I understand that 0-10 is not completely
> > finished. But, could you point me to the changes/proposal for reliable
> > messaging? Also, if there is a bullet list to the changes that have been
> > made/proposed for the 0-10 release, it'd be great if you could send it
> > to me or tell me where I can find that too. I would like to review the
> > scope of changes to reliable messaging, among other things like
> > transaction processing. For example, will the acknowledgement back to
> > the publisher feature provided for all types of messaging, like P2P,
> > Pub-sub, etc.
> >
> > Thanks.
> >
> > Regards,
> > Tanmay
> >
> >
> >
> > -----Original Message-----
> > From: [hidden email] [mailto:[hidden email]]
> > On Behalf Of Alexis Richardson
> > Sent: Tuesday, August 21, 2007 4:44 PM
> > To: Goel, Tanmay
> > Cc: Matthias Radestock; [hidden email]
> > Subject: Re: [rabbitmq-discuss] Reliability and AMQP
> >
> > Tanmay
> >
> > The kind of reliability that you refer to will appear in AMQP 0-10.
> > We have been working on this or over six months, along with others
> > from the AMQP Working Group.  It's an important future enhancement to
> > AMQP, and to RabbitMQ, and one that we are all excited about.
> >
> > Existing users will be able to continue using 0-8 and 0-9 (which is
> > almost the same as 0-8), but for those who need 'acks', that will
> > become a built-in feature of the spec, and of RabbitMQ.
> >
> > The 0-10 draft is not strictly speaking for general consumption, as
> > you might expect with a work in progress.  But we welcome commentary
> > and it is an 'open' spec, so do please let me know if you would like
> > to get involved.
> >
> > alexis
> >
> >
> >
> > On 8/21/07, Goel, Tanmay <[hidden email]> wrote:
> > >
> > >
> > >
> > >
> > > Hi,
> > >
> > >
> > >
> > > So, I'm still trying to understand. There seems to be reliability
> > missing in
> > > the protocol. The protocol supports acknowledgements from the consumer
> > to
> > > the queue but there are no acks from the queue/broker back to the
> > publisher
> > > in any of the schemes (store-forward, pub-sub, etc). I mean, for
> > example,
> > > the publisher sends a series of 10K messages to the broker and one
> > message
> > > gets dropped/lost in the network before reaching the broker. The
> > publisher
> > > will not get an ack and the broker will never know the message was
> > supposed
> > > to come. Will problems of this sort be taken care of in the later
> > versions
> > > of the spec or am I missing reading something?
> > >
> > >
> > >
> > > Are you guys working with the 0.9 version of the spec or do you have a
> > copy
> > > of the 0.10 release since I couldn't find it online?
> > >
> > >
> > >
> > > Thanks.
> > >
> > >
> > >
> > > Regards,
> > >
> > > Tanmay
> > > _______________________________________________
> > > rabbitmq-discuss mailing list
> > > [hidden email]
> > > http://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
> > >
> > >
> >
> >
> > --
> > Alexis Richardson
> > +44 20 7617 7339 (UK)
> > +44 77 9865 2911 (cell)
> > +1 650 206 2517 (US)
> >
>
>
> --
> Alexis Richardson
> +44 20 7617 7339 (UK)
> +44 77 9865 2911 (cell)
> +1 650 206 2517 (US)
>


--
Alexis Richardson
+44 20 7617 7339 (UK)
+44 77 9865 2911 (cell)
+1 650 206 2517 (US)

_______________________________________________
rabbitmq-discuss mailing list
[hidden email]
http://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
Loading...