[ 
https://issues.apache.org/jira/browse/IGNITE-4160?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Denis Magda updated IGNITE-4160:
--------------------------------
    Description: 
There is no way to set a predefined internal IP address and port number for a 
Kubernetes pod. Kubernetes manages this on its own by assigning an IP address 
dynamically.

In order to assemble an Ignite cluster under Kubernetes supervision we have to 
provide a Kubernetes specific implementation of discovery SPI. The discovery 
implementation has to use Kubernetes API to locate private IP addresses of all 
the Ignite pods (nodes) started so far and use these IP addresses at the join 
phase.

Refer to an exemplary implementation of a discovery component in other 
distributed storage:
https://github.com/pires/hazelcast-kubernetes#hazelcast-discovery-source

  was:
There is no way to set a predefined internal IP address and port number for a 
Kubernetes pod. Kubernetes manages this on its own by assigning an IP address 
and port dynamically.

In order to assemble an Ignite cluster under Kubernetes supervision we have to 
provide a Kubernetes specific implementation of discovery SPI. The discovery 
implementation has to use Kubernetes API to locate private IP addresses of all 
the Ignite pods (nodes) started so far and use these IP addresses at the join 
phase.

Refer to an exemplary implementation of a discovery component in other 
distributed storage:
https://github.com/pires/hazelcast-kubernetes#hazelcast-discovery-source


> Discovery SPI Implementation for Ignite Kubernetes Pods
> -------------------------------------------------------
>
>                 Key: IGNITE-4160
>                 URL: https://issues.apache.org/jira/browse/IGNITE-4160
>             Project: Ignite
>          Issue Type: Sub-task
>            Reporter: Denis Magda
>            Assignee: Denis Magda
>             Fix For: 2.0
>
>
> There is no way to set a predefined internal IP address and port number for a 
> Kubernetes pod. Kubernetes manages this on its own by assigning an IP address 
> dynamically.
> In order to assemble an Ignite cluster under Kubernetes supervision we have 
> to provide a Kubernetes specific implementation of discovery SPI. The 
> discovery implementation has to use Kubernetes API to locate private IP 
> addresses of all the Ignite pods (nodes) started so far and use these IP 
> addresses at the join phase.
> Refer to an exemplary implementation of a discovery component in other 
> distributed storage:
> https://github.com/pires/hazelcast-kubernetes#hazelcast-discovery-source



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to