incubator-cloudstack-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Chiradeep Vittal <Chiradeep.Vit...@citrix.com>
Subject Re: JVM level cluster deployments
Date Fri, 18 Jan 2013 18:20:58 GMT
It may help to remove the UI from the model -- the UI does not guarantee a
stateful session.
The UI interacts with the management server using end-user API calls.
Mutating API calls generally have 2 steps:
1. Create / modify the entity in the db
2. Call the backend provisioning steps. This is the orchestration engine.
All backend provisioning happens asynchronously. The asynchronous job
queue is a database table. If a job is incomplete due to a management
server failure, other servers in the cluster are expected to roll the
provisioning forward or roll it back.

On 1/17/13 9:16 PM, "Adnan Ashraf" <mohammad.adnan.ashraf@gmail.com> wrote:

>Thanks for the feedback. so how does CloudStack provide HA/failover
>guarantees? (not just the apache CloudStack but also the Citrix enterprise
>CloudPlatform and cloudPortal/CPBM)?
>
>A service provider using CloudStack would expect nothing breaks from two
>angle: user's perspective and back-end provisioning services. During
>provisioning of a VPC (which has many atomic steps) or in the middle of VM
>creation process, the application dies (for whatever reason). This case
>could easily result in half-baked (bad data) situation and abandoned
>workflows loosing the provisioning process.
>
>What workarounds or maybe persistent state based retries or logic is in
>place for this situation?
>
>Adnan
>
>
>On Thu, Jan 17, 2013 at 7:56 PM, Chiradeep Vittal <
>Chiradeep.Vittal@citrix.com> wrote:
>
>> As far as I know no one deploys in this manner.
>>
>> On 1/17/13 6:30 AM, "Adnan Ashraf" <mohammad.adnan.ashraf@gmail.com>
>> wrote:
>>
>> >Hi,
>> >
>> >I would appreciate if anyone can provide feedback on following
>>questions:
>> >
>> >Roughly what is the number of cloudStack deployments that are in JVM
>> >(tomcat or other server) cluster mode where all session objects, http
>> >sessions, beans etc are seamlessly failed over to the other node in
>>case
>> >one of the node in the cluster deployment goes out of service and the
>>user
>> >sessions carry as if nothing happened in the background.
>> >
>> >What percent of customers choose not to deploy cloudstack in a cluster
>> >mode
>> >and why?
>> >
>> >What is the largest known cluster size in deployment today?
>> >
>> >thanks
>> >
>> >regards,
>> >Adnan
>>
>>


Mime
View raw message