incubator-cloudstack-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alex Huang <Alex.Hu...@citrix.com>
Subject RE: [DISCUSS] OSGi framework for plugins and more?
Date Wed, 14 Nov 2012 14:45:52 GMT
Here's my two-bits on OSGi.  I actually started a thread like this sometime ago to which Mohammad
reference.  I then did some research on what OSGi can do.  The problems I looked at using
OSGi for just isn't solvable by OSGi.

For example, 
- How to contain fault within an availability zone.
- How to do rolling upgrade and phase out the rolling upgrades over a span of days to deal
with the time that it might take.
- How to do database upgrades/downgrades for the plugins
- How to scale different components differently

OSGi does solve some problems
- How to enable and disable plugins on a production system but I'm not quite sure how reliable
that is.  Even eclipse asks you to restart eclipse after adding a plugin.

I think after looking at this, then I decided that 

- For modularity, nothing is better than compilation boundaries.  The problem with some of
the plugins is that it depends on cloud-core and cloud-server.  It shouldn't .  All plugins
must build to cloud-api only.  Since all interfaces of CloudStack is in cloud-api (if you
think about that then cloud-api is basically the OSGi bundle), that's sufficient to separate
the plugins.
- For lifecycle of plugins, it probably requires that we switch to deploy in something like
Karaf before we can achieve runtime lifecycle changes.  I'm not sure it's entire necessary
and it doesn't take care of a plugin's database versioning problem.
- To resolve the other problems, we basically need to break cloudstack into separate processes.
 Hence I've proposed the idea of disaggregating cloudstack.

--Alex

> -----Original Message-----
> From: Charles Moulliard [mailto:ch007m@gmail.com]
> Sent: Wednesday, November 14, 2012 1:41 AM
> To: cloudstack-dev
> Subject: Re: [DISCUSS] OSGi framework for plugins and more?
> 
> Hi Kelven,
> 
> As Spring/Spring DM can be used on Apache Karaf (multi-container OSGI
> runtime), that should not be a problem to use what you have done on Karaf.
> Maybe we can talk about that next on IRC channel (@ch007M)
> 
> Regards,
> 
> 
> 
> 
> On Tue, Nov 13, 2012 at 10:05 PM, Kelven Yang
> <kelven.yang@citrix.com>wrote:
> 
> > The refactoring work that I did on Spring exploration is an incremental
> > approach trying to improve and encourage modularity of CloudStack, I
> hope
> > it will help CloudStack in the future to add OSGi in the future.
> >
> > As we have a huge amount of legacy code base, it will be less risky to do
> > incremental changes, and Spring effort here is the closest match of what
> > we have in our existing code base. What we are trying to achieve is to
> > first enable internal java modules to be more component-ized, our
> > customized component framework actually provides 99% things we need,
> > instead of us to fix the missing 1%, I feel that it might be worth to just
> > switch to a more standard framework with a minimal effort. It is  that
> > simple :-).  It is also easier to get things moving, in the end, if we
> > have a well component-ized system internally, it will be easier to adopt
> > any "the best" option like OSGi
> >
> > Kelven
> >
> >
> > On 11/13/12 8:09 AM, "Mohammad Nour El-Din"
> <nour.mohammad@gmail.com>
> > wrote:
> >
> > >Hi
> > >
> > >Had typos in this message the correct one is here:
> > >
> > >hi
> > >
> > >   we already had this discussion before earlier when ACS was just
> donated
> > >to ASF.
> > >
> > >I am not an OSGi expert at all :), but I would like to know what is the
> > >problem that needs to be solved rather than discussing or jumping into
> the
> > >solution first
> > >
> > >OSGi has a lot of magnificent features idd but do we really need them all,
> > >and also what does all these features add to us in the context of the
> > >context of the cloud. to be honest I don't have a definite answer and
> > >that's why I would suggest to list what we need to achieve regarding that
> > >part and look into the alternatives and choose what is best for us, OSGi
> > >is
> > >one option for sure
> > >
> > >Feedback/input ?
> > >On Tue, Nov 13, 2012 at 8:30 AM, Mohammad Nour El-Din <
> > >nour.mohammad@gmail.com> wrote:
> > >
> > >> hi
> > >>
> > >>    we already had this discussion before earlier when ACS was just
> > >>donated
> > >> to ASF.
> > >>
> > >> I am not an OSGi expert at all :), but I would like to know what is the
> > >> problem that needs to be solved rather than discussing the solution
> > >>first
> > >>
> > >> OSGi has a lot of magnificent features idd but do we really need them
> > >>all,
> > >> and also what does all these features add to us in the context of the
> > >> context of the cloud. to be honest I don have a definite answer and
> > >>that's
> > >> why I would suggest to list what we need to achieve regarding that part
> > >>and
> > >> look into the alternatives and choose what is best for us, OSGi is one
> > >>of
> > >> them for sure
> > >>
> > >> Feedback/input ?
> > >>
> > >> Sent from my Samsung Galaxy S3
> > >> Apologies for any typos
> > >>
> > >> On Nov 13, 2012 7:56 AM, "Hugo Trippaers"
> > >><HTrippaers@schubergphilis.com>
> > >> wrote:
> > >> >
> > >> >
> > >> >
> > >> > > -----Original Message-----
> > >> > > From: Kelven Yang [mailto:kelven.yang@citrix.com]
> > >> > > Sent: Monday, November 12, 2012 11:31 PM
> > >> > > To: cloudstack-dev@incubator.apache.org
> > >> > > Subject: Re: [DISCUSS] OSGi framework for plugins and more?
> > >> > >
> > >> > >
> > >> > >
> > >> > > On 11/12/12 1:34 PM, "Chip Childers" <chip.childers@sungard.com>
> > >> wrote:
> > >> > >
> > >> > > >On Mon, Nov 12, 2012 at 4:12 AM, Hugo Trippaers
> > >> > > ><HTrippaers@schubergphilis.com> wrote:
> > >> > > >> Hey  all,
> > >> > > >>
> > >> > > >> With the recent discussions on refactoring CloudStack
and the
> > >> working
> > >> > > >>going into javelin I would like to discuss using OSGi.
The
> > >>background
> > >> > > >>is that I have been struggling with ideas on how to setup
a plugin
> > >> > > >>system for CloudStack that would allow plugins to be
separate
> > >> entities
> > >> > > >>which can be release independently from CloudStack core.
Mainly
> to
> > >> > > >>deal with the current non-asf components but for future
> expansion
> > >>as
> > >> > > well.
> > >> > > >>
> > >> > > >> While at ApacheConEU I had the chance to discuss these
ideas
> with
> > >> one
> > >> > > >>of our mentors after his talk about OSGi. I'm pretty
charmed by
> > >>OSGi
> > >> > > >>and the options it provides. It's a well thought out
system that
> > >> allow
> > >> > > >>true modularity and pluggability. With the amount of
support in
> > >>the
> > >> > > >>java industry it's a standard that feels very mature
and a safe
> > >>bet,
> > >> > > >>one I would prefer to any homegrown plugin system. It
supports
> > >> > > >>versioning of components, strict separation of classes
between
> > >> > > >>components and all kind of funky features like hot-plugging
and
> > >> > > >>hot-replace. Using OSGi would mean that people can supply
> bundles
> > >> with
> > >> > > >>functionality which are maintained separately from the
'main
> code'
> > >> > > >>without having to worry about how to integrate it with
the core.
> > >>Just
> > >> > > >>putting the module in the right directory should be enough
to get
> > >> > > CloudStack to see and use the bundle.
> > >> > > >>Upgrades happen the same way, new version of an authenticator,
> > >>just
> > >> > > >>replace the bundle and let the framework replace it with
having
> to
> > >> > > >>shutdown the server at all.
> > >> > > >>
> > >> > > >> As we are discussing making CloudStack more modular,
I would
> > >>like to
> > >> > > >>propose to start using OSGi for this. It is a bit of
a learning
> > >>curve
> > >> > > >>to start with, but one we can get help with from our
mentors. I'm
> > >> > > >>already working on setting up a proof of concept for
a plugin
> > >>system
> > >> > > >>using OSGi together with a colleague to show how it works,
code
> is
> > >> > > >>always better than words.
> > >> > > >>
> > >> > > >> So what are your thoughts?
> > >> > > >>
> > >> > > >> Cheers,
> > >> > > >>
> > >> > > >> Hugo
> > >> > > >>
> > >> > > >
> > >> > > >I'm not familiar enough with OSGi to understand the tradeoffs
of
> > >>that
> > >> > > >vs other frameworks, but I'd suggest that Kelvin weigh in
here.
> > >>The
> > >> > > >work that he's doing on the Javelin branch is similar, and
there
> > >>might
> > >> > > >be an argument for Spring instead.
> > >> > > >
> > >> > > >Kelvin, I know you just responded on the other thread about
the
> > >> > > >relative timing of a switch.  Want to weigh in on the OSGi
> > >>approach's
> > >> > > >technical merit vs. other options?
> > >> > >
> > >> > > It will be nice to see the OSGi technical merit vs. other option
in
> > >> details.
> > >> > > Laying out these basic but fundamental frames may not benefit
a lot
> > >>in
> > >> the
> > >> > > short term, but we may get fully paid in long term. Spring can
only
> > >> give us
> > >> > > solution on compile-time/load-time component integration, it
> focuses
> > >> > > more on internal component wiring, OSGi seems to focus more at
> > >> runtime, I
> > >> > > think these two may be complementary to each other.
> > >> >
> > >> > To a certain extent these technologies can be used together, but not
> > >>in
> > >> this way it seems. OSGi is a framework that focusses on separation of
> > >>code
> > >> in various modules in such a way that other modules can not see and
> work
> > >> with the classes in other modules excluding via well defined services.
> > >>This
> > >> is a fundamental choice that touches the core of how CloudStack would
> be
> > >> put together. Instead of a single codebase (or at runtime a single
> > >> classloader) where modules would be loaded via (for example) the
> spring
> > >> framework based on an xml definition, the core would be an empty
> > >>framework
> > >> and modules work be plugged in and provide a certain service. This can
> > >>be
> > >> done at load time or runtime, that up to the implementers. For example
> > >>say
> > >> the core module would need a vm provisioned it would ask the service
> > >> registry if there was a service able to provide this functionality, if
> > >> there was that service would be asked to perform that task. Here is a
> > >>short
> > >> post that describes the context far better than I can:
> > >> http://elron.us/?p=95 .The real benefits are in also in the modularity,
> > >> because the framework is very strict on what bundles/interfaces are
> > >>exposed
> > >> and required, using the proper interfaces and limiting yourself to
> > >>exposed
> > >> interfaces is enforced by the framework.
> > >> >
> > >> > Hugo
> > >> > >
> > >> > >
> > >> > > >
> > >> > > >-chip
> > >> >
> > >>
> > >>
> > >
> > >
> > >--
> > >Thanks
> > >- Mohammad Nour
> > >----
> > >"Life is like riding a bicycle. To keep your balance you must keep moving"
> > >- Albert Einstein
> >
> >
> 
> 
> --
> Charles Moulliard
> Apache Committer / Sr. Enterprise Architect (RedHat)
> Twitter : @cmoulliard | Blog : http://cmoulliard.blogspot.com

Mime
View raw message