Open Stack

On 01/27/2014 09:32 PM, Robert Collins wrote:
> On 28 January 2014 14:59, Richard Su <rwsu at redhat.com> wrote:
>> Hi,
>>
>> I have been looking into how to add process/service monitoring to
>> tripleo. Here I want to be able to detect when an openstack dependent
>> component that is deployed on an instance has failed. And when a failure
>> has occurred I want to be notified and eventually see it in Tuskar.
>
> +1 on the goal.
>
> We have two basic routes here:
>   - use existing things
>   - build something new
>
> Service monitoring is a rich field, and there is lots of opportunity
> to do new and interesting things. However, it is also a wicked
> problem, because - well see all the prior art.
>
> Further to that, like with Chef/Puppet deployers will often have
> existing infrastructure investment we should support.
>
> My suggestion is that we take a modular approach - we define an
> interface, supply glue code to glue e.g. the assimilation monitoring
> project into the interface, and then build on the interface.
>
> There are some important questions - how rich is the interface, what
> lives in the glue code, how many backends we'll eventually support,
> but the key thing for me is that we don't go down the path of
> rewriting e.g. nagios because we're afraid of the dependency: it's
> optional to bring in any specific backend, and we can always write our
> own later.

+1 to modular. There is too much in this space to try to dictate a 
single blessed approach and still get adoption. I suggest starting by 
trying to nail down a common way of communicating what we're looking for 
without tying it to closely to a particular implementation.

> Another point is the crawl/walk/run cycle - lets just get the ability
> to click through to a native monitoring screen to start with. That
> should be abou t a thousand times faster to ring together than a
> complete custom everything.
>
> -Rob
>