Hi David and Aled,

Thank you for your replies, all very helpful. For now I'll leave the team
with their config management solution and separately start a Brooklyn setup.

Cheers,


Jeroen

On Mon, Dec 8, 2014 at 5:12 PM, Aled Sage <[email protected]> wrote:

> Hi Jeroen,
>
> For the "deployment and wiring" pillar, I'd add that another
> differentiator is making it simpler to wire inter-component dependencies.
> Simple examples include the database URL injected into the app-servers by
> whatever mechanism the app-server expects; another example is injecting the
> addresses of "seed nodes" of a cassandra cluster into all the nodes, and
> continuing to maintain that list of seed nodes even as nodes are
> added/removed from the cluster. As the inter-component dependencies get
> more complex, it gets increasingly hard to configure that kind of thing in
> Puppet, Chef, Salt, etc. It is a design philosophy of Brooklyn to allow
> flexibility when injecting dependencies, to do it in the way that the
> components expect (as opposed to enforcing an opinion on how dependencies
> should be injected).
>
> We'd be very happy to discuss further what a Brooklyn solution would look
> like for your use-case. If you want to have a call (as well as e-mail /
> IRC) then please do let us know.
>
> Aled
>
>
>
> On 08/12/2014 11:56, Richard Downer wrote:
>
>> Hi Jeroen,
>>
>> On 7 December 2014 at 23:38, Jeroen van der Wal <[email protected]>
>> wrote:
>>
>>> I hadn't heard of Apache Brooklyn before ApacheCon and out of curiosity I
>>> attended the presentations: it was love on first sight. But falling in
>>> love
>>> is easy, before getting into a relationship you have to know each other
>>> better. And there's the family too ;-)
>>>
>> Welcome to our group :) and I'm glad that you found the ApacheCon
>> presentations valuable - it was a key aim for my talk to introduce
>> Brooklyn to more people who knew nothing about it, so I am glad that
>> it has worked on at least one person!
>>
>>  I'm PMC member of the Apache Isis project, a Java framework for rapid
>>> Domain Drive application development. A typical Isis application is
>>> perfectly fit to be provisioned through Brooklyn so in the next months
>>> I'm
>>> going to experiment with that.
>>>
>> We'd love to hear how your experiments go! Of course you can drop in
>> to here or to our IRC channel at any time to ask questions or explore
>> how this might work.
>>
>>  Now the sysadmins of my most friendly
>>> customer are getting aware that they must start implementing some DevOps
>>> practices. They're at the stage of looking at Chef or Puppet and have a
>>> steep learning curve ahead of them so I was playing the idea of
>>> introducing
>>> Brooklyn to them and hit two flies at once (a Dutch expression maybe not
>>> proper English..). They don't have an urge to go to the cloud but have a
>>> myriad of internal (Java) applications and also OTS things like OpenLDAP,
>>> JIRA, Confluence, Nexus, Jenkins, Zabbix, etc.
>>>
>>> My question is: would Brooklyn fit in landscapes similar to my customer
>>> or
>>> should I not bother them and leave them with Chef, Puppet or whatever and
>>>
>> I always explain that Brooklyn has two halves to it (the two pillars
>> of Brooklyn that I referred to in my talk) - the first half is the
>> deployment and wiring. This could be seen as being quite similar Chef
>> or Puppet - and in fact Brooklyn is able to delegate the installation
>> to Chef, allowing Chef cookbooks to be named in the blueprint and a
>> runlist given (Puppet is something we'd like to support, too).
>>
>> The second half - runtime management by policies - is what sets us
>> apart from the deployment/configuration management tools like Chef and
>> Puppet. Once the application is deployed, it is monitored
>> continuously, and the policies can take actions such as
>> restart/replace failed nodes, and add/remove nodes to match
>> requirements of demand.
>>
>> Brooklyn was built from day one to be a "cloud" application, and many
>> of the policies take full advantage of the ability to start up new
>> machines and throw them away when no longer needed. We do have the
>> facility to use a "BYON" location - "bring your own nodes", where you
>> supply a list of IP addresses and SSH credentials of ready-to-use
>> machines, and Brooklyn manages it a bit like a cloud provider. It does
>> have disadvantages compared to a pure-cloud solution though.
>>
>> If all of these points align with what your customer is thinking, then
>> Brooklyn could be a very good match. If it doesn't align, then your
>> customer could proceed with Chef (or Puppet, or an alternative) at
>> this time, and come back to Brooklyn in the future, and know that
>> their investment in Chef is re-usable in the Brooklyn world.
>>
>> Hope this helps - please feel free to ask more questions!
>>
>> Richard.
>>
>
>

Reply via email to