lolee_k

> Hi, can you share the configuration you used to deploy the RayJob? The configuration in the "Reproduction script" part, [hp-pi-autoscale-job.txt](https://github.com/ray-project/kuberay/files/9495509/hp-pi-autoscale-job.txt) is the rayjob-spec.

Hi, is it a bug and how are you going to fix it? @DmitriGekhtman

I need to use autoscaler in my ray-job currently. So I will fix it for my own use this days. I think there are two ways to solve it. First,...

> > But if "ShutdownAfterJobFinishes" enabled, the first finished ray-job may release the ray-cluster. This maybe another potential problem. > > Could you clarify what you mean by "release"? I...

> > I need to use autoscaler in my ray-job currently. > > Could you say more about your use-case? How do you intend to leverage autoscaling when running a...

> > > But if "ShutdownAfterJobFinishes" enabled, the first finished ray-job may release the ray-cluster. This maybe another potential problem. > > > > > > Could you clarify what...

> @loleek You can try set `clusterSelector` as label selectors which select running rayclusters, while leave `rayClusterSpec` empty. ok. I find it in master branch, which 0.3.0 not support.

> It does seems reasonable to allow a job to get started before resources are fully provisioned. Now in ray-cluster's WorkerGroupSpec, there have MinReplicas and MaxReplicas. This actually indicates that...

Here I can give a demo code which can support volcano in ray operator. https://github.com/loleek/kuberay/commit/745e12cb6b72d4299ec1b4bef65cb46d305eafc3 I get queue and priority class from head pod spec, but it works well to...

Like the description in azure，I need session window to track actions on every device.