pod status of tfjob always pending after #693

[merryzhou]

Is this a BUG REPORT or FEATURE REQUEST?:
/kind bug

What happened:
we use volcano to schedule tfjobs, but all pods status are always pending, after some research ,we found that in #693： volcano no longer support scheduling.v1alpha1 and scheduling.v1alpha2 API。
but podgroup created by tf-operator is scheduling.v1alpha1 version

so will volcano no longer compatible with tfjobs?

What you expected to happen:
volcano can scheduler tfjobs 、mpijobs。。

Environment:

• Volcano Version:
  

[k82cn]

so will volcano no longer compatible with tfjobs?

We'll continue to support tfjobs: for current tfjobs, it's better to use volcano:0.3, and we'll raise PR in tf-operator to correct that.

[k82cn]

/cc @hzxuzhonghu , please help to raise a PR in kubeflow/tf-operator for that.

[YuxiJin-tobeyjin]

Not only tf-operator, other operators such as pytorch-operator, mpi-operator all have the same problems.

[hzxuzhonghu]

We plan to release 0.4 in the next week in order to support kubeflow. And by then, i can file a pr in kubeflow.

[k82cn]

That was fixed in kubeflow/common v3.0 and main branch of kubeflow/tf-operator :)