openstack team mailing list archive

[Eugene Kirpichov <ekirpichov@xxxxxxxxx>]

Hi Andrea,

On Thu, Jul 26, 2012 at 4:10 AM, Rosa, Andrea (HP Cloud Services)
<andrea.rosa@xxxxxx> wrote:
> Hi Eugene,
>
> Thanks for the patch.
> I have a question:
> it seems to me that this patch is a (good) starting point of a broader change to be able to use HA in active/active configuration with RMQ.
> As far as I know, with that configuration we need to add some extra logic for consumers to deal with "consumer cancellation notification" and duplicated messages due to a (potentially) re-send after a failover.
>
> Is that correct?
Well, we ran fine in production without this extra logic, but perhaps
we just didn't hit a situation where it was required.

As for consumer cancellation notifications - I need to remember when
exactly they happen in an HA setup. Maybe you're right.

As for duplicated messages - the situation here is no different
whether you have 1 or many rabbits; reconnection logic was already in
place. Perhaps something should be done here, but it seems that the
lack of this logic didn't hurt anyone so far. Maybe it is because IIRC
messages are ack'd immediately and the failure window is very small.

> Is there some plan to have a blueprint for this change?
I don't have such a plan. Should I?

>
> Regards
> --
> Andrea Rosa
>
>
>>-----Original Message-----
>>From: Eugene Kirpichov [mailto:ekirpichov@xxxxxxxxx]
>>Sent: 26 July 2012 00:46
>>To: Alessandro Tagliapietra; rbryant@xxxxxxxxxx
>>Cc: Rosa, Andrea (HP Cloud Services); OpenStack Development Mailing
>>List; openstack@xxxxxxxxxxxxxxxxxxx
>>Subject: Re: [openstack-dev] [Openstack] High Available queues in
>>rabbitmq
>>
>>Gentlemen,
>>
>>Here is my patch: https://review.openstack.org/#/c/10305/
>>It also depends on another small patch
>>https://review.openstack.org/#/c/10197
>>
>>I'd like to ask someone to review it.
>>Also, how to get these changes into nova? It seems that nova has a
>>copy-paste of openstack-common inside it, should I just mirror the
>>changes to nova once they're accepted in openstack-common?
>>
>>I'm cc'ing Russell Bryant because he originally created the
>>openstack-common module.
>>
>>On Wed, Jul 25, 2012 at 3:03 AM, Alessandro Tagliapietra
>><tagliapietra.alessandro@xxxxxxxxx> wrote:
>>> Yup, using as resource is a "old" way as
>>http://www.rabbitmq.com/ha.html
>>> Active/active makes sure that you have no downtime and it's simple as
>>you
>>> don't need to use DRBD.
>>>
>>> 2012/7/25 Rosa, Andrea (HP Cloud Services) <andrea.rosa@xxxxxx>
>>>
>>>> Sorry for my question, I have just seen from the original thread that
>>>> we are talking about HA with Active/Active solution.
>>>> --
>>>> Andrea Rosa
>>>>
>>>> >-----Original Message-----
>>>> >From: Rosa, Andrea (HP Cloud Services)
>>>> >Sent: 25 July 2012 10:45
>>>> >To: Eugene Kirpichov
>>>> >Cc: openstack-dev@xxxxxxxxxxxxxxxxxxx; Alessandro Tagliapietra;
>>>> >openstack@xxxxxxxxxxxxxxxxxxx
>>>> >Subject: Re: [openstack-dev] [Openstack] High Available queues in
>>>> >rabbitmq
>>>> >
>>>> >Hi
>>>> >
>>>> >Your patch doesn't use a Resource manager, so are you working on an
>>>> >Active/Active
>>>> >configuration using mirrored queues? Or are you working on a cluster
>>>> >configuration?
>>>> >
>>>> >I am really interested in that change, thanks for your help.
>>>> >Regards
>>>> >--
>>>> >Andrea Rosa
>>>> >
>>>> >>-----Original Message-----
>>>> >>From: openstack-bounces+andrea.rosa=hp.com@xxxxxxxxxxxxxxxxxxx
>>>> >>[mailto:openstack-bounces+andrea.rosa=hp.com@xxxxxxxxxxxxxxxxxxx]
>>On
>>>> >>Behalf Of Alessandro Tagliapietra
>>>> >>Sent: 24 July 2012 17:58
>>>> >>To: Eugene Kirpichov
>>>> >>Cc: openstack-dev@xxxxxxxxxxxxxxxxxxx;
>>openstack@xxxxxxxxxxxxxxxxxxx
>>>> >>Subject: Re: [Openstack] High Available queues in rabbitmq
>>>> >>
>>>> >>Oh, so without the need to put an IP floating between hosts.
>>>> >>Good job, thanks for helping
>>>> >>
>>>> >>Best
>>>> >>
>>>> >>Alessandro
>>>> >>
>>>> >>Il giorno 24/lug/2012, alle ore 17:49, Eugene Kirpichov ha scritto:
>>>> >>
>>>> >>> Hi Alessandro,
>>>> >>>
>>>> >>> My patch is about removing the need for pacemaker (and it's
>>pacemaker
>>>> >>> that I denoted with the term "TCP load balancer").
>>>> >>>
>>>> >>> I didn't submit the patch yesterday because I underestimated the
>>>> >>> effort to write unit tests for it and found a few issues on the
>>way.
>>>> >I
>>>> >>> hope I'll finish today.
>>>> >>>
>>>> >>> On Tue, Jul 24, 2012 at 12:00 AM, Alessandro Tagliapietra
>>>> >>> <tagliapietra.alessandro@xxxxxxxxx> wrote:
>>>> >>>> Sorry for the delay, i was out from work.
>>>> >>>> Awesome work Eugene, I don't need the patch instantly as i'm
>>still
>>>> >>building the infrastructure.
>>>> >>>> Will it will take alot of time to go in Ubuntu repositories?
>>>> >>>>
>>>> >>>> Why you said you need load balancing? You can use only the
>>master
>>>> >>node and in case the rabbitmq-server dies, switch the ip to the new
>>>> >>master with pacemaker, that's how I would do.
>>>> >>>>
>>>> >>>> Best Regards
>>>> >>>>
>>>> >>>> Alessadro
>>>> >>>>
>>>> >>>>
>>>> >>>> Il giorno 23/lug/2012, alle ore 21:49, Eugene Kirpichov ha
>>scritto:
>>>> >>>>
>>>> >>>>> +openstack-dev@
>>>> >>>>>
>>>> >>>>> To openstack-dev: this is a discussion of an upcoming patch
>>about
>>>> >>>>> native RabbitMQ H/A support in nova. I'll post the patch for
>>>> >>>>> codereview today.
>>>> >>>>>
>>>> >>>>> On Mon, Jul 23, 2012 at 12:46 PM, Eugene Kirpichov
>>>> >><ekirpichov@xxxxxxxxx> wrote:
>>>> >>>>>> Yup, that's basically the same thing that Jay suggested :)
>>Obvious
>>>> >>in
>>>> >>>>>> retrospect...
>>>> >>>>>>
>>>> >>>>>> On Mon, Jul 23, 2012 at 12:42 PM, Oleg Gelbukh
>>>> >><ogelbukh@xxxxxxxxxxxx> wrote:
>>>> >>>>>>> Eugene,
>>>> >>>>>>>
>>>> >>>>>>> I suggest just add option 'rabbit_servers' that will override
>>>> >>>>>>> 'rabbit_host'/'rabbit_port' pair, if present. This won't
>>break
>>>> >>anything, in
>>>> >>>>>>> my understanding.
>>>> >>>>>>>
>>>> >>>>>>> --
>>>> >>>>>>> Best regards,
>>>> >>>>>>> Oleg Gelbukh
>>>> >>>>>>> Mirantis, Inc.
>>>> >>>>>>>
>>>> >>>>>>>
>>>> >>>>>>> On Mon, Jul 23, 2012 at 10:58 PM, Eugene Kirpichov
>>>> >><ekirpichov@xxxxxxxxx>
>>>> >>>>>>> wrote:
>>>> >>>>>>>>
>>>> >>>>>>>> Hi,
>>>> >>>>>>>>
>>>> >>>>>>>> I'm working on a RabbitMQ H/A patch right now.
>>>> >>>>>>>>
>>>> >>>>>>>> It actually involves more than just using H/A queues (unless
>>>> >>you're
>>>> >>>>>>>> willing to add a TCP load balancer on top of your RMQ
>>cluster).
>>>> >>>>>>>> You also need to add support for multiple RabbitMQ's
>>directly to
>>>> >>nova.
>>>> >>>>>>>> This is not hard at all, and I have the patch ready and
>>tested
>>>> >in
>>>> >>>>>>>> production.
>>>> >>>>>>>>
>>>> >>>>>>>> Alessandro, if you need this urgently, I can send you the
>>patch
>>>> >>right
>>>> >>>>>>>> now before the discussion codereview for inclusion in core
>>nova.
>>>> >>>>>>>>
>>>> >>>>>>>> The only problem is, it breaks backward compatibility a bit:
>>my
>>>> >>patch
>>>> >>>>>>>> assumes you have a flag "rabbit_addresses" which should look
>>>> >like
>>>> >>>>>>>> "rmq-host1:5672,rmq-host2:5672" instead of the prior
>>rabbit_host
>>>> >>and
>>>> >>>>>>>> rabbit_port flags.
>>>> >>>>>>>>
>>>> >>>>>>>> Guys, can you advise on a way to do this without being ugly
>>and
>>>> >>>>>>>> without breaking compatibility?
>>>> >>>>>>>> Maybe have "rabbit_host", "rabbit_port" be ListOpt's? But
>>that
>>>> >>sounds
>>>> >>>>>>>> weird, as their names are in singular.
>>>> >>>>>>>> Maybe have "rabbit_host", "rabbit_port" and also
>>"rabbit_host2",
>>>> >>>>>>>> "rabbit_port2" (assuming we only have clusters of 2 nodes)?
>>>> >>>>>>>> Something else?
>>>> >>>>>>>>
>>>> >>>>>>>> On Mon, Jul 23, 2012 at 11:27 AM, Jay Pipes
>><jaypipes@xxxxxxxxx>
>>>> >>wrote:
>>>> >>>>>>>>> On 07/23/2012 09:02 AM, Alessandro Tagliapietra wrote:
>>>> >>>>>>>>>> Hi guys,
>>>> >>>>>>>>>>
>>>> >>>>>>>>>> just an idea, i'm deploying Openstack trying to make it
>>HA.
>>>> >>>>>>>>>> The missing thing is rabbitmq, which can be easily started
>>in
>>>> >>>>>>>>>> active/active mode, but it needs to declare the queues
>>adding
>>>> >>an
>>>> >>>>>>>>>> x-ha-policy entry.
>>>> >>>>>>>>>> http://www.rabbitmq.com/ha.html
>>>> >>>>>>>>>> It would be nice to add a config entry to be able to
>>declare
>>>> >>the queues
>>>> >>>>>>>>>> in that way.
>>>> >>>>>>>>>> If someone know where to edit the openstack code, else
>>i'll
>>>> >try
>>>> >>to do
>>>> >>>>>>>>>> that in the next weeks maybe.
>>>> >>>>>>>>>
>>>> >>>>>>>>>
>>>> >>>>>>>>> https://github.com/openstack/openstack-
>>>> >>common/blob/master/openstack/common/rpc/impl_kombu.py
>>>> >>>>>>>>>
>>>> >>>>>>>>> You'll need to add the config options there and the queue
>>is
>>>> >>declared
>>>> >>>>>>>>> here with the options supplied to the ConsumerBase
>>constructor:
>>>> >>>>>>>>>
>>>> >>>>>>>>>
>>>> >>>>>>>>> https://github.com/openstack/openstack-
>>>> >>common/blob/master/openstack/common/rpc/impl_kombu.py#L114
>>>> >>>>>>>>>
>>>> >>>>>>>>> Best,
>>>> >>>>>>>>> -jay
>>>> >>>>>>>>>
>>>> >>>>>>>>> _______________________________________________
>>>> >>>>>>>>> Mailing list: https://launchpad.net/~openstack
>>>> >>>>>>>>> Post to     : openstack@xxxxxxxxxxxxxxxxxxx
>>>> >>>>>>>>> Unsubscribe : https://launchpad.net/~openstack
>>>> >>>>>>>>> More help   : https://help.launchpad.net/ListHelp
>>>> >>>>>>>>
>>>> >>>>>>>>
>>>> >>>>>>>>
>>>> >>>>>>>> --
>>>> >>>>>>>> Eugene Kirpichov
>>>> >>>>>>>> http://www.linkedin.com/in/eugenekirpichov
>>>> >>>>>>>>
>>>> >>>>>>>> _______________________________________________
>>>> >>>>>>>> Mailing list: https://launchpad.net/~openstack
>>>> >>>>>>>> Post to     : openstack@xxxxxxxxxxxxxxxxxxx
>>>> >>>>>>>> Unsubscribe : https://launchpad.net/~openstack
>>>> >>>>>>>> More help   : https://help.launchpad.net/ListHelp
>>>> >>>>>>>
>>>> >>>>>>>
>>>> >>>>>>
>>>> >>>>>>
>>>> >>>>>>
>>>> >>>>>> --
>>>> >>>>>> Eugene Kirpichov
>>>> >>>>>> http://www.linkedin.com/in/eugenekirpichov
>>>> >>>>>
>>>> >>>>>
>>>> >>>>>
>>>> >>>>> --
>>>> >>>>> Eugene Kirpichov
>>>> >>>>> http://www.linkedin.com/in/eugenekirpichov
>>>> >>>>>
>>>> >>>>> _______________________________________________
>>>> >>>>> Mailing list: https://launchpad.net/~openstack
>>>> >>>>> Post to     : openstack@xxxxxxxxxxxxxxxxxxx
>>>> >>>>> Unsubscribe : https://launchpad.net/~openstack
>>>> >>>>> More help   : https://help.launchpad.net/ListHelp
>>>> >>>>
>>>> >>>
>>>> >>>
>>>> >>>
>>>> >>> --
>>>> >>> Eugene Kirpichov
>>>> >>> http://www.linkedin.com/in/eugenekirpichov
>>>> >>
>>>> >>
>>>> >>_______________________________________________
>>>> >>Mailing list: https://launchpad.net/~openstack
>>>> >>Post to     : openstack@xxxxxxxxxxxxxxxxxxx
>>>> >>Unsubscribe : https://launchpad.net/~openstack
>>>> >>More help   : https://help.launchpad.net/ListHelp
>>>> >
>>>> >_______________________________________________
>>>> >OpenStack-dev mailing list
>>>> >OpenStack-dev@xxxxxxxxxxxxxxxxxxx
>>>> >http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>>
>>>
>>
>>
>>
>>--
>>Eugene Kirpichov
>>http://www.linkedin.com/in/eugenekirpichov



-- 
Eugene Kirpichov
http://www.linkedin.com/in/eugenekirpichov