That actually sounds like a lot more than I want Capistrano core to be
involved with. :) By all means, experiment, but what you're suggesting
would be better suited to a third-party plugin than for cap-core. If
anything like this finds its way into capistrano core, it'll just be a
simple way to iterate over a subset of the servers and run tasks on
subsets of that.

- Jamis

On 2/16/09 9:07 AM, Gerhardus Geldenhuis wrote:
> Hi
> No code atm I'm afraid but I will share some ideas/requirements/
> thoughts....
> 
> We do a state check of the application before removing it from the
> cluster which we can use to compare with after new deployment.
> 
> We have a level1 and level2 check for applications. Level1 check
> functional deployment in tomcat, which basically checks the
> application server.
> Level2 actually submits a request to check complete infrastructure
> functionality and parses xml( that is the intend anyway still needs to
> be written)
> 
> We want to build in a failure percentage that if you have 10 servers
> and 1 fails after 5 deployments, deployment should continue until in
> reaches a failure percentage, say 20%.
> 
> The modjk tasks I have written that act as our loadbalancers actually
> works quite well, once I have extended it a bit more I will gladly
> share the code.
> 
> I think sequential deploy will need some generic loadbalancer/cluster
> tasks build-in.
> 
> At the moment we use something like the following:
> run "wget --quiet -O - --connect-timeout=1 --timeout=1 --tries=1
> 'http://$CAPISTRANO:HOST$/jkserver-status?cmd=update&from=list&w=#
> {modjk_loadbalancergroup}&sw=#{current_tomcat}&wa=1&wf=1&wn=#
> {current_tomcat}&wr=&wc=&wd=0' 1>&2>/dev/null"
> 
> maybe a wrapper to allow easier/simplified http interaction with put
> and get might be usefull. (perhaps even some xml manipulation)
> 
> Anyway just some random thoughts.
> 
> Regards
> 
> On Feb 16, 3:44 pm, Jamis Buck <[email protected]> wrote:
>> On 2/16/09 8:37 AM, Gerhardus Geldenhuis wrote:
>>
>>> Thanks, I am going to try that right now.
>>> How do you feel about a "sequential" extention to capistrano? Any
>>> thoughts on how it could fit in with the current structure?
>> I think that'd be great. Even as something for Capistrano core. I've
>> needed, and others have needed it, so I'm sure it'd be useful.
>>
>> In general, I'd really love to get away from the ROLES, HOSTS, and
>> HOSTFILTER environment variables. I just don't know where to go instead
>> of those, and honestly, I haven't spent much time thinking about it.
>>
>> If you have any ideas, please feel free to hack on them and see what you
>> come up with.
>>
>> - Jamis
>>
>>
>>
>>> Regards
>>> On Feb 16, 3:16 pm, Jamis Buck <[email protected]> wrote:
>>>> If you want subtasks to use a particular host, you need to set the HOSTS
>>>> or HOSTFILTER environment variables. E.g.
>>>>   find_servers(...).each do |server|
>>>>     ENV['HOSTFILTER'] = server.host
>>>>     sub.task.here
>>>>     ENV['HOSTFILTER'] = nil
>>>>   end
>>>> It's extremely ugly and hacky and I don't like it, but there's not
>>>> currently another way to do it.
>>>> - Jamis
>>>> On 2/16/09 5:09 AM, Gerhardus Geldenhuis wrote:
>>>>> I have also put the relevant code onhttp://pastie.org/390600
>>>>> Regards
> > 


--~--~---------~--~----~------------~-------~--~----~
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at http://groups.google.com/group/capistrano
-~----------~----~----~----~------~----~------~--~---

Reply via email to