> Isn't your code making use of Netscaler specific API calls ? Any LB would need > to implement these same calls right ? AutoScale feature in CloudStack provides a framework to setup AutoScale configuration (there are new APIs in CloudStack like createAutoScalePolicy, createCondition, createCounter...). This configuration in CloudStack needs to be translated to a Load balancer device similar to a load balancer rule being translated to loadbalancer device specific calls. It need not be a one-to-one mapping of calls from CloudStack to a loadbalancer device. NetScaler APIs are enhanced to support this feature. It should be pretty straight forward for a load balancer to support a similar configuration
> -----Original Message----- > From: Sebastien Goasguen [mailto:run...@gmail.com] > Sent: Friday, November 16, 2012 2:08 AM > To: cloudstack-dev@incubator.apache.org > Subject: Re: Integrating autoscale branch to master? > > > On Nov 15, 2012, at 9:02 PM, Vijay Venkatachalam > <vijay.venkatacha...@citrix.com> wrote: > > > Simple answer yes, any LB device could do this. > > > > It should be done in 2 parts > > > > The open source loadbalancer part, where the loadbalancer have to be > modified, for example, to monitor for autoscaling purposes ( in addition to > the regular health monitoring). > > Isn't your code making use of Netscaler specific API calls ? Any LB would need > to implement these same calls right ? > > -Sebastien > > > > > The cloudstack part, in the loadbalancer's resource (in cloudstack), > > translate > the autoscale config to the config which the open source loadbalancer can > understand. > > > > Thanks, > > Vijay V. > > > >> -----Original Message----- > >> From: Sebastien Goasguen [mailto:run...@gmail.com] > >> Sent: Friday, November 16, 2012 1:02 AM > >> To: cloudstack-dev@incubator.apache.org > >> Subject: Re: Integrating autoscale branch to master? > >> > >> > >> On Nov 15, 2012, at 5:44 PM, Ram Ganesh <ram.gan...@citrix.com> > wrote: > >> > >>> Any thoughts? Unless we hear otherwise would like to push the > patches. > >> This feature had been code complete for a very long time. If there > >> are still concerns/opinions let us know and we can take steps to correct > them. > >>> > >> > >> Ram, any thoughts on how this could work without Netscaler ? Any > >> alternative open source load balancer we could use to implement this ? > >> > >> -Sebastien > >> > >>> Thanks, > >>> Ram > >>> > >>> > >>>> -----Original Message----- > >>>> From: Ram Ganesh [mailto:ram.gan...@citrix.com] > >>>> Sent: 15 November 2012 07:09 > >>>> To: cloudstack-dev@incubator.apache.org > >>>> Subject: RE: Integrating autoscale branch to master? > >>>> > >>>> David, > >>>> > >>>> Can we go ahead with merge of AutoScale code into master? Are there > >>>> any more open questions? > >>>> > >>>> Thanks, > >>>> Ram > >>>> > >>>>> -----Original Message----- > >>>>> From: Vijay Venkatachalam [mailto:vijay.venkatacha...@citrix.com] > >>>>> Sent: 13 November 2012 12:34 > >>>>> To: cloudstack-dev@incubator.apache.org > >>>>> Subject: RE: Integrating autoscale branch to master? > >>>>> > >>>>> > >>>>> My replies inline > >>>>> > >>>>> Thanks, > >>>>> Vijay V. > >>>>> > >>>>>> -----Original Message----- > >>>>>> From: David Nalley [mailto:da...@gnsa.us] > >>>>>> Sent: Monday, October 15, 2012 7:42 PM > >>>>>> To: cloudstack-dev@incubator.apache.org > >>>>>> Subject: Re: Integrating autoscale branch to master? > >>>>>> > >>>>>> On Fri, Oct 12, 2012 at 11:54 AM, Vijay Venkatachalam > >>>>>> <vijay.venkatacha...@citrix.com> wrote: > >>>>>>> Ok I will keep changes ready, and will merge once 4.0's news is > >>>>> declared. > >>>>>>> > >>>>>>> -Vijay V. > >>>>>>> > >>>>>> > >>>>>> Vijay, > >>>>>> > >>>>>> I haven't kept up with this recently so a couple of > >>>>> questions/assumptions: > >>>>>> > >>>>>> 1. Autoscale code will require NetScaler libraries right? > >>>>> > >>>>> There are 2 parts to autoscale code. > >>>>> A. AutoScale Manager and its services, This is part of the core. > >>>>> And has no No Netscaler jar dependency; This part is coded like > >>>>> any other NetworkServiceManager, meaning > >>>> any > >>>>> network > >>>>> element can provide autoscale service. So this part does not have > >>>>> compile time dependency with NetScaler jar. > >>>>> > >>>>> If an autoscale provider (which is most likely already an LB > >>>>> provider) does not exist > >>>>> in that network an error is thrown at run time. > >>>>> So for all oss builds (where Netscaler is not packaged and cannot > >>>> be > >>>>> added > >>>>> to the infrastructure) we should get a run-time error when > >>>>> configuring autoscale. > >>>>> > >>>>> B. NetScaler Element and Netscaler Resource (which is part of > >>>>> non-oss build today) > >>>>> has been enhanced to provide autoscale capability. Today only > >>>>> NetScaler does this, in future any network element can he > >>>> enhanced > >>>>> to provide autoscale. This part already has NetScaler jar > >>>>> dependency > >>>>> (and is considered non-oss today) and will continue to have > >>>>> NetScaler > >>>>> jar dependency. > >>>>> > >>>>> > >>>>>> 2. Is autoscale functionality modular enough that we can turn > >>>>> building it > >>>>>> on/off at will? > >>>>> > >>>>> > >>>>> Short Answer, No. > >>>>> Since AutoScale is like an addon to LB there are touch- points. > >>>>> For example, when a LoadBalancerRule is deleted the AutoScale > >>>>> entities created for it also should be deleted, hence the dependency. > >>>>> Basically there is code in LB core to delete autoscale entities on > >>>> the > >>>>> loadbalancer > >>>>> rule's delete path. Hence Part (A.) could not be modularized. Is > >>>> there > >>>>> an alternative here? > >>>>> > >>>>> Also, in the UI autoscale will appear as part of LB to the user > >>>>> and > >>>> if > >>>>> he attempts to configure > >>>>> AutoScale in a network which does not have NetScaler; he will get > >>>>> a run-time error. > >>>>> > >>>>>> 3. Has there been any change to the netscaler java library > >>>> licensing? > >>>>>> I know there was work underway, but I never heard about a > >>>> conclusion. > >>>>>> > >>>>> > >>>>> I am still chasing the legal team on this, but for the moment, we > >>>>> should continue to treat NetScaler as non-oss. > >>>>> > >>>>>> --David > >
