Additionally we have removed the kubernetesClusterId from the
ClusterContext and added the kubernetesClusterId to the
kuberentesServices.Idea behind that is we can't have kubernetesClusterId in
ClusterContext as there can be one or more Kubernetes clusters in the case
of multi-region deployment.But ideally Kubernetes service needs to bound
with a Kubernetes cluster.

I have push the changes to the stratos-4.1.x branch . Shall we have a quick
round of smoke testing and do the 4.1.3 release?

Thanks,
Gayan


On Sat, Sep 19, 2015 at 9:30 PM, Imesh Gunaratne <im...@apache.org> wrote:

> I had an offline discussion on this with Gayan. It looks like the root
> cause of this problem is that we have added Kubernetes Cluster ID to the
> Cluster Context. This seems to be incorrect. A cluster can have multiple
> Kubernetes Clusters if its a multi-region deployment.
>
> We need to remove the property ClusterContext.kubernetesClsuterId() and do
> following to find the Kubernetes Cluster ID of a member:
>
> Partition partition = memberContext.getPartition();
>
> String kubernetesClusterId = partition.getKubernetesClusterId();
>
> KubernetesCluster kubernetesCluster =
> CloudControllerContext.getInstance().getKubernetesCluster(kubernetesClusterId);
>
>
> In addition we also identified few improvements that we can do in
> KubernetesIaas class. Gayan is now working on this fix.
>
> Thanks
>
> On Sat, Sep 19, 2015 at 1:06 PM, Imesh Gunaratne <im...@apache.org> wrote:
>
>> We might need to use cluster id and network partition id as the key of
>> this map.
>>
>> Once we are done with this fix shall we do 4.1.3 release build?
>>
>> Thanks
>>
>>
>> On Friday, September 18, 2015, Gayan Gunarathne <gay...@wso2.com> wrote:
>>
>>> When we try to un-deploy the application in a multi region deployment,
>>> It will throw the following exception.
>>>
>>> javax.ws.rs.WebApplicationException: HTTP 404 pods "tomcatr1-pod-1" not
>>> found
>>>
>>> Issue seems like,in a topology service clusterIdClusterMap has only one
>>> value even there is two clusters in two regions. So when un-deploying it is
>>> map to the same region cluster. So when un-deploying the application, first
>>> region will remove successfully but in the second region it will throw the
>>> exception of "HTTP 404 pod not found"
>>>
>>> I think we may need to recheck the logic of adding the
>>> clusterIdClusterMap in the case of multi region deployment.
>>>
>>> Thanks,
>>> Gayan
>>> --
>>>
>>> Gayan Gunarathne
>>> Technical Lead, WSO2 Inc. (http://wso2.com)
>>> Committer & PMC Member, Apache Stratos
>>> email : gay...@wso2.com  | mobile : +94 775030545 <%2B94%20766819985>
>>>
>>>
>>>
>>
>>
>> --
>> Imesh Gunaratne
>>
>> Senior Technical Lead, WSO2
>> Committer & PMC Member, Apache Stratos
>>
>>
>
>
> --
> Imesh Gunaratne
>
> Senior Technical Lead, WSO2
> Committer & PMC Member, Apache Stratos
>



-- 

Gayan Gunarathne
Technical Lead, WSO2 Inc. (http://wso2.com)
Committer & PMC Member, Apache Stratos
email : gay...@wso2.com  | mobile : +94 775030545 <%2B94%20766819985>

Reply via email to