[openstack-dev] [oslo.messaging] ZeroMQ driver maintenance next steps
harlowja at outlook.com
Mon Nov 17 20:36:30 UTC 2014
Doug Hellmann wrote:
> On Nov 17, 2014, at 10:01 AM, Denis Makogon<dmakogon at mirantis.com> wrote:
>> On Mon, Nov 17, 2014 at 4:26 PM, James Page<james.page at ubuntu.com> wrote:
>>> -----BEGIN PGP SIGNED MESSAGE-----
>>> Hash: SHA256
>>> Hi Denis
>>> On 17/11/14 07:43, Denis Makogon wrote:
>>>> During Paris Design summit oslo.messaging session was raised good
>>>> question about maintaining ZeroMQ driver in upstream (see section
>>>> “dropping ZeroMQ support in oslo.messaging” at ) . As we all
>>>> know, good thoughts are comming always after. I’d like to propose
>>>> several improvements in process of maintaining and developing of
>>>> ZeroMQ driver in upstream.
>>>> Contribution focus. As we all see, that there are enough patches
>>>> that are trying to address certain problems related to ZeroMQ
>>>> Few of them trying to add functional tests, which is definitely
>>>> good, but … there’s always ‘but’, they are not “gate”-able.
>>> I'm not sure I understand you statement about them not being
>>> "gate"-able - the functional/unit tests currently proposed for the zmq
>>> driver run fine as part of the standard test suite execution - maybe
>>> the confusion is over what 'functional' actually means, but in my
>>> opinion until we have some level of testing of this driver, we can't
>>> effectively make changes and fix bugs.
>> I do agree that there's a confusion what "functional testing" means.
>> Another thing, what the best solution is? Unit tests are welcome, but they
>> are still remain to be units (they are using mocks, etc.)
>> I'd try to define what 'fuctional testing' means for me. Functional testing
>> in oslo.messaging means that we've been using real service for messaging
>> (in this case - deployed 0mq). So, the simple definition, in term os
>> OpenStack integration, we should be able to run full Tempest test suit for
>> OpenStack services that are using oslo.messaging with enabled zmq driver.
>> Am i right or not?
> That’s a good goal, but that’s not what I had in mind for in-tree functional tests.
> We should build a simple app using the Oslo libraries that we can place in the oslo.messaging source tree to use for exercising the communication patterns of the library with different drivers. Ideally that would be a single app (or set of apps) that could be used to test all drivers, with tests written against the app rather than the driver. Once we have the app and tests, we can define a new gate job to run those tests just for oslo.messaging, with a different configuration for each driver we support.
An interesting idea that might be useful that taskflow implemented/has
The examples @
get tested during unit test runs to ensure they work as expected. This
seems close to your 'simple app' (where app is replaced with example),
it might be nice to have a similar approach for oslo.messaging that has
'examples' that are these apps that get ran to probe the functionality
of oslo.messaging (as well as useful for documentation to show people
how to use it, which is the other usage taskflow has for these examples)
The hacky example tester could likely be shared (or refactored, since it
probably needs it),
>>>> My proposal for this topic is to change contribution focus from
>>>> oslo.messaging by itself to OpenStack/Infra project and DevStack
>>>> (subsequently to devstack-gate too).
>>>> I guess there would be questions “why?”. I think the answer is
>>>> pretty obvious: we have driver that is not being tested at all
>>>> within DevStack and project integration.
>>> This was discussed in the oslo.messaging summit session, and
>>> re-enabling zeromq support in devstack is definately on my todo list,
>>> but I don't think the should block landing of the currently proposed
>>> unit tests on oslo.messaging.
>>> For example https://review.openstack.org/#/c/128233/ says about adding
>> functional and units. I'm ok with units, but what about functional tests?
>> Which oslo.messaging gate job runs them?
>>>> Also i’d say that such focus re-orientation would be very useful
>>>> as source of use cases and bugs eventually. Here’s a list of what
>>>> we, as team, should do first:
>>>> Ensure that DevStack can successfully:
>>>> Install ZeroMQ.
>>>> Configure each project to work with zmq driver from
>>>> Ensure that we can run successfully simple test plan for each
>>>> project (like boot VM, fill object store container, spin up volume,
>>> A better objective would be able to run a full tempest test as
>>> conducted with the RabbitMQ driver IMHO.
>> I do agree with this too. But we should define step-by-step plan for this
>> type of testing. Since we want to see quick gate feedback adding full test
>> suit would be an overhead, at least for now.
>>>> ZeroMQ driver maintainers communityorganization. During design
>>>> session was raised question about who uses zmq driver in
>>>> I’ve seen folks from Canonical and few other companies. So, here’s
>>>> my proposals around improving process of maintaining of given
>>>> With respect to best practices of driver maintaining procedure, we
>>>> might need to set up community sub-group. What would it give to us
>>>> and to the project subsequently? It’s not pretty obvious, at least
>>>> for now, but i’d try to light out couple moments:
>>>> continuous driver stability
>>>> continuous community support (across all OpenStack Project that are
>>>> using same model: driver should have maintaining team, would it be
>>>> a company or community sub-group)
>>>> As sub-group we would need to have our own weekly meeting. Separate
>>>> meeting would keep us, as sub-group, pretty focused on zmq driver
>>>> only (but it doesn’t mean that we should not participate in regular
>>>> meetings). Same question. What it would give us and to the project?
>>>> I’d say that the only one valid answer is: we’d not disturb other
>>>> folk that are not actually interested in given topic and in zqm
>>>> drive too.
>>> I'd prefer that we continue to discuss ZMQ on the broader
>>> oslo.messaging context; I'm keen that the OpenStack community
>>> understands that we want ZMQ to be a first tier driver like qpid and
>>> rmq, and I'm not convinced that pushing discussion out to a separate
>>> sub-group enforces that message...
>> The only thing that i'm woried about is that we could eventually eat all
>> meeting time. That's why i try to build out drive maintaining/contribution
>> team for zmq drive since there were not so may hands when was raised
>> question about who uses zqm driver.
>>>> So, in the end, taking into account words above i’d like to get
>>>> feedback from all folks. I’m pretty open for discussion, and if
>>>> needed i can commit myself for driving such activities in
>>> - --
>>> James Page
>>> Ubuntu and Debian Developer
>>> james.page at ubuntu.com
>>> jamespage at debian.org
>>> -----BEGIN PGP SIGNATURE-----
>>> Version: GnuPG v1
>>> -----END PGP SIGNATURE-----
>>> OpenStack-dev mailing list
>>> OpenStack-dev at lists.openstack.org
>> Best regards,
>> Denis M.
>> OpenStack-dev mailing list
>> OpenStack-dev at lists.openstack.org
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
Sent with Postbox <http://www.getpostbox.com>
More information about the OpenStack-dev