Subject: Re: [HTCondor-users] Htcondor with python ray
Thanks for quick reply. Its more at end of on ray part rather than htcondor.
Ray actually uses all vm resources regardless of cgroup and cpu affinities set.
In static slots I tend to set it only 1 slot so that whole vm is used, but in this case my utilisation is too low as I am not able to share the resources.
I am exploring to use docker universe for isolation. But most of stuff on internet really says to use kubernetes with ray.Â
Thanks Greg for your reply. Cant really find any
stuff to integrate both.
Ray requires isolation in case of dynamic
slots,Â
What isolation is missing from dynamic slots that are present in
static slots?
If given permission by the host OS, for both static and dynamic
slots HTCondor uses Linux cgroups to isolate memory and CPU and
swap, can run each dynamic slot with a different UID if desired, and
also can be configured to isolate process table (pid namespaces) and
subdirectories on the host (via mount namespaces) . For even more
isolation, HTCondor can also run containers or virtual machines in
dynamic slots.Â
regards,
Todd
and in case of static slots it will be less
resource utilisation leading to higher costs for bigger
VMs.Â
I was hoping to get some resources like
Kuberay(combines kubernetes and ray and auto scale)
On
11/24/25 08:42, Ram Ban wrote:
> Hi all,
>
> I am running htcondor 24 with dynamic allocation of
resources on each
> machine, Is there any resources to connect htcondor with
python ray,
> to use htcondor as auto scale for ray nodes.
>
Raman:
Many people use many different workflow managements systems
(like dask,
snakemake, Task Vine, etc.) on top of HTCondor. I'm not aware
of anyone
using Ray today, but I would not be surprised to learn of
someone doing
so. Launch ray workers from an HTCondor job should be
straightforward.
-greg
_______________________________________________
HTCondor-users mailing list
To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx
with a
subject: Unsubscribe