xhejtman
xhejtman
Anything new here? It seems that the issue is still present.
> Thanks @xhejtman , and yes don't worry about the k8s driver launcher. The implementation looks good so far. perhaps code around `useJobResource()` should be improved somehow.
so, reworkerd a bit, removed process option to pod options. if it ok, I will fix tests and add few for mpijob. also I think it would be really useful...
Example how to use it: https://github.com/CERIT-SC/nf-pi-mpi Btw, there is one remaining issue. mpioperator creates launcher and workers at once and launcher immediatelly runs `mpirun` command which can fail, if workers...
Can we proceed here?
> Looks like some CI tests are failing. Can you update them or do you need help? They are fixed
> hey @bentsherman that's OK! I can't comment for @xhejtman, but for my comment above, you haven't seen the Flux Operator talk yet (it's next week, woot!) so you should...
Hello, I think this is great refactor, as it came to my mind to support volcano jobs and using a plugin, it would be nice. > The only lingering issue...
E.g., nextflow uses term `accelerator: number`, however, it also allows to specify card type that matches node annotations in kubernetes. I think there is no way to deal with all...
This could be caused by: https://github.com/rancher/fleet/commit/84b7d8aa6964eeca44170db1ac0d320e9b7705d1