You are viewing a plain text version of this content. The canonical link for it is here.
Posted to submarine-dev@hadoop.apache.org by "Wanqiang Ji (Jira)" <ji...@apache.org> on 2019/11/03 15:06:00 UTC
[jira] [Assigned] (SUBMARINE-214) [WIP] Add submitter-k8s module
[ https://issues.apache.org/jira/browse/SUBMARINE-214?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Wanqiang Ji reassigned SUBMARINE-214:
-------------------------------------
Assignee: Wanqiang Ji
Description:
As a PoC, we need several below steps to run the TF job on k8s.
# The submarine server can run as an independent daemon
# tf-operator deployed in a K8s cluster. Use kubeflow/tf-operator for now
# REST API in submarine server to accept a generic TF job spec
# submitter-k8s translate the spec into tf-operator format and use k8s client API to submit this to the cluster
# REST API to get tf job status from k8s and parse it to a standard format to be consumed
Add the submitter-k8s module implements the CRUD for custom resource objects, includes the UTs. Before run the UTs, you should deploy the tf-operator in a K8s cluster and replace the config at the test resources.
Summary: [WIP] Add submitter-k8s module (was: [SS] Add submitter-yarn into server-submitter)
> [WIP] Add submitter-k8s module
> ------------------------------
>
> Key: SUBMARINE-214
> URL: https://issues.apache.org/jira/browse/SUBMARINE-214
> Project: Apache Submarine
> Issue Type: Sub-task
> Components: Submarine Server
> Reporter: Wanqiang Ji
> Assignee: Wanqiang Ji
> Priority: Major
>
> As a PoC, we need several below steps to run the TF job on k8s.
> # The submarine server can run as an independent daemon
> # tf-operator deployed in a K8s cluster. Use kubeflow/tf-operator for now
> # REST API in submarine server to accept a generic TF job spec
> # submitter-k8s translate the spec into tf-operator format and use k8s client API to submit this to the cluster
> # REST API to get tf job status from k8s and parse it to a standard format to be consumed
> Add the submitter-k8s module implements the CRUD for custom resource objects, includes the UTs. Before run the UTs, you should deploy the tf-operator in a K8s cluster and replace the config at the test resources.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)